2Btainment commited on
Commit
ca4ec99
·
verified ·
1 Parent(s): 607c4f0

v0.7: Unified Studio (BASS/NOISE lanes, compensated stem graph) – keeps Demucs, /data bucket, Docker port 7860

Browse files
README.md CHANGED
@@ -11,7 +11,7 @@ short_description: Forensic audio lab – HPSS + Demucs stems, arrangement mixer
11
 
12
  > **Hugging Face Space note:** Public demo instance, no authentication. Demucs htdemucs v4 runs on CPU (max. 120 s audio per job); HPSS has no duration limit. Upload limit 200 MiB. Projects are stored in an attached HF Storage Bucket at `/data` if mounted, otherwise in `/tmp` (lost on restart). Do not upload private/confidential audio.
13
 
14
- # ABSOLUTION Ω — v0.6 · Arrangement Studio + OMEGA-TAG Mini
15
 
16
  **Lokales Fullstack-Audiostudio.** Enthält alle Funktionen aus v0.4 (BASS-CORE, NOISE-CORE,
17
  Virtual Bass, HPSS, optional Demucs) und zusätzlich ein farbiges Mehrspur-Arrangement.
@@ -21,14 +21,18 @@ Virtual Bass, HPSS, optional Demucs) und zusätzlich ein farbiges Mehrspur-Arran
21
  1. ZIP entpacken. **Python 3.11+** muss installiert sein.
22
  2. Windows: `start_windows.bat` ausführen. macOS/Linux: `bash start.sh`.
23
  3. Browser: **http://127.0.0.1:8000**.
24
- 4. `Studio` → eigene Musik laden oder **Demo-Sound laden**.
25
- 5. Sidebar: **Arrangement** → **Mini-KI analysieren**.
26
- 6. Automatische *Klangtextur-Hypothesen* prüfen, ändern und **Plan speichern**.
27
- 7. **Separation starten**: HPSS (harmonisch/perkussiv/Residuum) ist sofort verfügbar;
28
- Demucs nur mit separat zugänglichen lokalen Gewichten und Neural Dependencies.
29
- 8. Im Arrangement vorhandene Stems durch **M** (Mute), **S** (Solo), Gain, Stereo-
30
- Panorama, Effekte bearbeiten. Oben die farbige Mixübersicht und Timeline nutzen.
31
- 9. **Mix berechnen** → **Mix anhören** → **WAV exportieren**.
 
 
 
 
32
 
33
  **Beispiel direkt starten, ohne Startskript:**
34
  ```bash
@@ -40,7 +44,41 @@ python -m pytest -q
40
  python -m uvicorn app.server:app --host 127.0.0.1 --port 8000
41
  ```
42
 
43
- ### Neue Funktionen
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
44
 
45
  - Mehrspur-Arrangement mit einzelnen Instrumenten-/Signalstems je nach verfügbarem Modell.
46
  - Jedes *tatsächlich* verfügbare Audio-Stem erhält eigene Farbe und samplebasierte
 
11
 
12
  > **Hugging Face Space note:** Public demo instance, no authentication. Demucs htdemucs v4 runs on CPU (max. 120 s audio per job); HPSS has no duration limit. Upload limit 200 MiB. Projects are stored in an attached HF Storage Bucket at `/data` if mounted, otherwise in `/tmp` (lost on restart). Do not upload private/confidential audio.
13
 
14
+ # ABSOLUTION Ω — v0.7 · Unified Multitrack Studio · BASS/NOISE Lanes
15
 
16
  **Lokales Fullstack-Audiostudio.** Enthält alle Funktionen aus v0.4 (BASS-CORE, NOISE-CORE,
17
  Virtual Bass, HPSS, optional Demucs) und zusätzlich ein farbiges Mehrspur-Arrangement.
 
21
  1. ZIP entpacken. **Python 3.11+** muss installiert sein.
22
  2. Windows: `start_windows.bat` ausführen. macOS/Linux: `bash start.sh`.
23
  3. Browser: **http://127.0.0.1:8000**.
24
+ 4. Im **Ω STUDIO** oben **Datei importieren** oder **Demo laden**.
25
+ 5. Optional **Mini-KI analysieren** → Hypothesen prüfen und **Plan speichern**.
26
+ 6. **Separation starten** (HPSS ist immer verfügbar; Demucs benötigt lokale Modellgewichte).
27
+ 7. In der **Element-Kontrolle** auf **Kick & Bassline als Spuren** klicken.
28
+ 8. **Störungen scannen**, verdächtige Stellen manuell auswählen und **Auswahl reparieren**.
29
+ Nur nach einer solchen Freigabe entsteht eine echte, separate **STÖRSPUR**.
30
+ 9. Bestehende Stems werden dabei rechnerisch kompensiert. Alle Spuren zusammen
31
+ ergeben beim neutralen Mix wieder das Original, soweit der Ausgangsseparator
32
+ numerisch konsistent ist. **Die Störspur bleibt zunächst hörbar**.
33
+ 10. Für die gezielte Entfernung bestätigter Klickanteile die Störspur mit **M** stummschalten.
34
+ 11. Für jede Spur Gain, Mute, Solo, Stereo-Fader und Effekte einstellen →
35
+ **Mix berechnen**, anhören und WAV/ZIP exportieren.
36
 
37
  **Beispiel direkt starten, ohne Startskript:**
38
  ```bash
 
44
  python -m uvicorn app.server:app --host 127.0.0.1 --port 8000
45
  ```
46
 
47
+
48
+ ## v0.7 — Debugging: BASS-CORE + NOISE-CORE sind echte zusätzliche Arrangement-Lanes
49
+
50
+ - **KICK** (`kick_core`): tiefes, transient-orientiertes Bassband; keine sichere Kick-Isolation.
51
+ - **BASSLINE** (`bassline_core`): tiefes, tonales Bassband; keine sichere Instrument-Isolation.
52
+ - **STÖRSPUR** (`noise_difference`): genau die **explizit freigegebenen** Reparaturdifferenzen
53
+ (`Original − repariert`). Ohne freigegebene Klickreparatur gibt es diese Spur nicht.
54
+ - **Kompensierte Basisstems**: das System zieht die drei zusätzlichen Signalanteile von
55
+ den tatsächlichen Basisstems ab. Die Zusatz-Lanes **verdoppeln niemals** die Bassenergie.
56
+ - Bei HPSS werden Bassline und Kick aus `harmonic` bzw. `percussive` kompensiert, die
57
+ Stördifferenz aus `residual`. Mit Demucs werden `bass`, `drums` und `residual` verwendet.
58
+ Ohne Grundseparation wird der Originalmix als `source_body` herangezogen.
59
+ - Eine neutrale Summe aller Lanes bleibt die Originalmischung. Nichtdestuktive Bearbeitung;
60
+ ein neuer Stereo-Mix wird nur auf ausdrücklichen Wunsch gerendert.
61
+ - **Klangschutz:** Noise-Lane ist nicht stumm. `M` auf NOISE bewirkt erst nach manueller
62
+ Freigabe die tatsächliche Korrektur. `S` ermöglicht differenzbasiertes Abhören.
63
+ - Die bestehenden Bass-Optimierungs-Ausgaben (mit optional synthetischen Obertönen)
64
+ bleiben separate Audition/Exports und werden **nicht** automatisch in den neutralen
65
+ Gesamtmix addiert.
66
+ - **Kdenlive/FFmpeg Virtual Bass** bleibt separater 2.1-LFE-Export. Kein Import in
67
+ den Stereo-Mixer, weil dieser eine andere Kanalbelegung hat.
68
+
69
+ ### Technische Grenzen
70
+ - Kein Stems-„Magie“: HPSS ist eine spektrale Abschätzung; Demucs braucht geprüfte,
71
+ lokal bereitgestellte Gewichte. Mini-TAG benennt nur Hypothesen.
72
+ - NOISE-CORE extrahiert bisher nur manuell akzeptierte Reparaturdifferenzen,
73
+ nicht zuverlässig jede Art von Umgebungsrauschen.
74
+ - Der Mixer rechnet mit WAV Float32, kann bei starken Gain-/FX-Änderungen >0 dBFS
75
+ ausgeben und meldet das ausdrücklich.
76
+ - Das Bauen zusätzlicher Spuren legt WAV-Dateien im `omega_data`-Ordner ab.
77
+ Alte Graphgenerationen werden für laufende Projekte nicht automatisch gelöscht.
78
+ - Das lokale Backend hat weder Mehrnutzer-Authentifizierung noch dauerhafte Jobs.
79
+ Nicht ungeschützt ins öffentliche Internet stellen.
80
+
81
+ ### Funktionen und Komponenten
82
 
83
  - Mehrspur-Arrangement mit einzelnen Instrumenten-/Signalstems je nach verfügbarem Modell.
84
  - Jedes *tatsächlich* verfügbare Audio-Stem erhält eigene Farbe und samplebasierte
app/server.py CHANGED
@@ -27,6 +27,7 @@ from .noise import RepairParams, inspect_noise_file, repair_noise_file
27
  from .virtualbass import available as virtualbass_available, render_virtualbass
28
  from .mini_model import inspect_file as inspect_mini, MODEL_FILE
29
  from .mixer import render_mix, sampled_peaks, rubberband_available
 
30
 
31
  from .audio import (
32
  AudioError, align_and_resample, analyze_file, hpss_separate_file,
@@ -37,8 +38,6 @@ from .audio import (
37
  # Configure ABSOLUTION_MAX_UPLOAD_MIB to fit your available storage.
38
  MAX_UPLOAD_MIB = max(1, min(4096, int(os.environ.get("ABSOLUTION_MAX_UPLOAD_MIB", "1024"))))
39
  MAX_UPLOAD_BYTES = MAX_UPLOAD_MIB * 1024 * 1024
40
- # Demucs on shared CPU hardware: cap input duration (0 = unlimited).
41
- DEMUCS_MAX_SECONDS = float(os.environ.get("ABSOLUTION_DEMUCS_MAX_SECONDS", "0"))
42
  ALLOWED_EXTENSIONS = {".wav", ".flac", ".aif", ".aiff", ".mp3", ".ogg"}
43
 
44
 
@@ -53,6 +52,10 @@ def check_id(raw: str) -> str:
53
  raise HTTPException(status_code=404, detail="Project not found.") from exc
54
 
55
 
 
 
 
 
56
  class ProjectCreate(BaseModel):
57
  name: str = Field(default="Untitled project", min_length=1, max_length=80)
58
 
@@ -158,21 +161,64 @@ class Service:
158
  "id": project_id, "name": name, "created_at": timestamp(),
159
  "asset": None, "analysis": None, "stems": {},
160
  "separation": None, "report": None, "bass": None, "noise": None, "virtualbass": None,
161
- "inventory": None, "plan": [], "mixer": None,
162
  }
163
  self.write_manifest(project_id, manifest)
164
  return manifest
165
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
166
  def create_mixer_job(self, request: MixerRequest) -> dict:
167
  project_id=check_id(request.project_id)
168
  with self.lock:
169
  manifest=self.manifest(project_id)
170
- if not manifest.get("stems"):
171
- raise HTTPException(status_code=409,detail="Run actual separation first; planned labels are not audio.")
172
- unknown=set(request.tracks)-set(manifest["stems"])
 
173
  if unknown:
174
  raise HTTPException(status_code=422,detail="Unknown real stems: "+", ".join(sorted(unknown)))
175
- if len(manifest["stems"])>128:
176
  raise HTTPException(status_code=422,detail="More than 128 stems not supported for one render.")
177
  if (request.tempo!=1 or request.pitch_semitones!=0) and not rubberband_available():
178
  raise HTTPException(status_code=422,detail="FFmpeg with rubberband is required for independent tempo/pitch.")
@@ -191,7 +237,11 @@ class Service:
191
  try:
192
  self.update_job(job_id,status="running",progress=3)
193
  manifest=self.manifest(project_id);folder=self.folder(project_id)
194
- sources={name:folder/"stems"/data["master"] for name,data in manifest["stems"].items()}
 
 
 
 
195
  output=folder/"mixer"/(job_id+".wav")
196
  result=render_mix(sources,settings,output,
197
  progress=lambda r:self.update_job(job_id,progress=3+int(r*90)))
@@ -219,7 +269,7 @@ class Service:
219
  "id": "demucs", "name": "Demucs HT v4 · neural (local weights)",
220
  "ready": configured, "type": "neural_network",
221
  "stems": ["drums", "bass", "other", "vocals", "residual"],
222
- "limitations": ("Requires trusted local htdemucs weights/config (official checkpoint 955717e8) and "
223
  "demucs/torch dependencies. No automatic downloads."),
224
  },
225
  ]
@@ -309,7 +359,10 @@ class Service:
309
  with self.lock:
310
  updated = self.manifest(project_id)
311
  updated["bass"] = result
 
312
  self.write_manifest(project_id, updated)
 
 
313
  self.update_job(job_id, status="done", progress=100)
314
  except Exception as exc:
315
  self.update_job(job_id, status="failed", error=f"{type(exc).__name__}: {exc}")
@@ -371,6 +424,8 @@ class Service:
371
  "tracks":tracks,"report":report,
372
  }
373
  self.write_manifest(project_id,updated)
 
 
374
  self.update_job(job_id,status="done",progress=100)
375
  except Exception as exc:
376
  self.update_job(job_id,status="failed",error=f"{type(exc).__name__}: {exc}")
@@ -459,7 +514,13 @@ class Service:
459
  "preview_format": "WAV PCM 16-bit (separate listening copies)",
460
  }
461
  updated["report"] = report
 
 
 
 
 
462
  self.write_manifest(project_id, updated)
 
463
  self.update_job(job_id, status="done", progress=100)
464
  except Exception as exc:
465
  self.update_job(job_id, status="failed", error=f"{type(exc).__name__}: {exc}")
@@ -512,7 +573,7 @@ def create_app(data_root: Path | None = None) -> FastAPI:
512
  if data_root is None:
513
  data_root = writable_dir(default, Path(os.environ.get("ABSOLUTION_FALLBACK_DIR", "/tmp/omega_data")))
514
  service = Service(data_root)
515
- app = FastAPI(title="ABSOLUTION Ω", version="0.6.0",
516
  description="Local-first audio forensic vertical slice")
517
  app.state.service = service
518
  static_dir = Path(__file__).parent / "static"
@@ -532,7 +593,7 @@ def create_app(data_root: Path | None = None) -> FastAPI:
532
 
533
  @app.get("/api/v1/health")
534
  def health():
535
- return {"status": "ok", "version": "0.6.0", "arrangement_studio": True, "mini_tag_model_ready": MODEL_FILE.is_file(), "rubberband_ready": rubberband_available(), "basscore": True, "noisecore": True, "virtualbass_ready": virtualbass_available(), "job_backend": "process-local", "data_dir": str(service.data_root), "persistent_storage": str(service.data_root).startswith("/data"), "demucs_max_seconds": DEMUCS_MAX_SECONDS or None,
536
  "network_downloads": False, "auth": False,
537
  "max_upload_bytes": MAX_UPLOAD_BYTES,
538
  "duration_limit_seconds": None, "streaming_hpss": True}
@@ -746,15 +807,26 @@ def create_app(data_root: Path | None = None) -> FastAPI:
746
  if not manifest.get("asset"):raise HTTPException(status_code=404,detail="No audio")
747
  path=folder/manifest["asset"]["stored_name"]
748
  else:
749
- if stem not in manifest.get("stems",{}):
 
 
750
  raise HTTPException(status_code=404,detail="Unknown stem")
751
- path=folder/"stems"/manifest["stems"][stem]["master"]
752
  if points<64 or points>1000:
753
  raise HTTPException(status_code=422,detail="points must be 64..1000")
754
  try:return sampled_peaks(path,points)
755
  except (ValueError,RuntimeError) as exc:
756
  raise HTTPException(status_code=422,detail=str(exc)) from exc
757
 
 
 
 
 
 
 
 
 
 
758
  @app.post("/api/v1/jobs/mix",status_code=202)
759
  def start_mix(request:MixerRequest):
760
  return service.create_mixer_job(request)
@@ -775,8 +847,8 @@ def create_app(data_root: Path | None = None) -> FastAPI:
775
  @app.get("/api/v1/projects/{project_id}/stems")
776
  def get_stems(project_id: str):
777
  manifest = service.manifest(project_id)
778
- return {"stems": manifest["stems"], "report": manifest["report"],
779
- "separation": manifest["separation"]}
780
 
781
  @app.get("/api/v1/projects/{project_id}/audio/{stem}")
782
  def get_audio(project_id: str, stem: str, preview: bool = False):
@@ -787,10 +859,11 @@ def create_app(data_root: Path | None = None) -> FastAPI:
787
  raise HTTPException(status_code=404, detail="Audio not found.")
788
  path = folder / ("original_preview.wav" if preview else manifest["asset"]["stored_name"])
789
  else:
790
- if stem not in manifest["stems"]:
 
791
  raise HTTPException(status_code=404, detail="Stem not found.")
792
  key = "preview" if preview else "master"
793
- path = folder / "stems" / manifest["stems"][stem][key]
794
  if not path.is_file():
795
  raise HTTPException(status_code=404, detail="Audio not found.")
796
  return FileResponse(path, media_type="audio/wav" if preview or path.suffix == ".wav" else "application/octet-stream",
@@ -800,8 +873,9 @@ def create_app(data_root: Path | None = None) -> FastAPI:
800
  def export_stems(project_id: str):
801
  folder = service.folder(project_id)
802
  manifest = service.manifest(project_id)
803
- if not manifest["stems"]:
804
- raise HTTPException(status_code=409, detail="Separate audio before exporting.")
 
805
  target = folder / "absolution_export.zip"
806
  # Build only from controlled server-side names. Atomic replace.
807
  tmp = folder / f".{uuid.uuid4().hex}.zip"
@@ -809,8 +883,9 @@ def create_app(data_root: Path | None = None) -> FastAPI:
809
  archive.write(folder / "manifest.json", "manifest.json")
810
  original_path = folder / manifest["asset"]["stored_name"]
811
  archive.write(original_path, "source" + original_path.suffix)
812
- for item in manifest["stems"].values():
813
- archive.write(folder / "stems" / item["master"], "stems/" + item["master"])
 
814
  tmp.replace(target)
815
  return FileResponse(target, media_type="application/zip",
816
  filename=f"absolution_{project_id[:8]}.zip")
 
27
  from .virtualbass import available as virtualbass_available, render_virtualbass
28
  from .mini_model import inspect_file as inspect_mini, MODEL_FILE
29
  from .mixer import render_mix, sampled_peaks, rubberband_available
30
+ from .unified import build_unified_graph
31
 
32
  from .audio import (
33
  AudioError, align_and_resample, analyze_file, hpss_separate_file,
 
38
  # Configure ABSOLUTION_MAX_UPLOAD_MIB to fit your available storage.
39
  MAX_UPLOAD_MIB = max(1, min(4096, int(os.environ.get("ABSOLUTION_MAX_UPLOAD_MIB", "1024"))))
40
  MAX_UPLOAD_BYTES = MAX_UPLOAD_MIB * 1024 * 1024
 
 
41
  ALLOWED_EXTENSIONS = {".wav", ".flac", ".aif", ".aiff", ".mp3", ".ogg"}
42
 
43
 
 
52
  raise HTTPException(status_code=404, detail="Project not found.") from exc
53
 
54
 
55
+ # Demucs on shared CPU hardware: cap input duration (0 = unlimited).
56
+ DEMUCS_MAX_SECONDS = float(os.environ.get("ABSOLUTION_DEMUCS_MAX_SECONDS", "0"))
57
+
58
+
59
  class ProjectCreate(BaseModel):
60
  name: str = Field(default="Untitled project", min_length=1, max_length=80)
61
 
 
161
  "id": project_id, "name": name, "created_at": timestamp(),
162
  "asset": None, "analysis": None, "stems": {},
163
  "separation": None, "report": None, "bass": None, "noise": None, "virtualbass": None,
164
+ "inventory": None, "plan": [], "mixer": None, "unified": None,
165
  }
166
  self.write_manifest(project_id, manifest)
167
  return manifest
168
 
169
+ def active_tracks(self, manifest: dict) -> dict:
170
+ graph = manifest.get("unified")
171
+ if graph and graph.get("mix_consistent"):
172
+ return graph["stems"]
173
+ return manifest.get("stems") or {}
174
+
175
+ def refresh_unified(self, project_id: str) -> dict | None:
176
+ """Called only from worker or explicit user action. Never overwrites source."""
177
+ manifest = self.manifest(project_id)
178
+ graph = build_unified_graph(self.folder(project_id), manifest)
179
+ with self.lock:
180
+ latest = self.manifest(project_id)
181
+ latest["unified"] = graph
182
+ latest["mixer"] = None
183
+ self.write_manifest(project_id, latest)
184
+ return graph
185
+
186
+ def create_unified_job(self, project_id: str) -> dict:
187
+ project_id = check_id(project_id)
188
+ with self.lock:
189
+ manifest = self.manifest(project_id)
190
+ if not manifest.get("asset"):
191
+ raise HTTPException(status_code=409, detail="Import audio before building tracks.")
192
+ if any(j["project_id"]==project_id and j["status"] in ("queued","running")
193
+ for j in self.jobs.values()):
194
+ raise HTTPException(status_code=409, detail="Project already processing.")
195
+ job={"id":str(uuid.uuid4()),"project_id":project_id,
196
+ "kind":"unified", "model":"canonical_track_graph",
197
+ "status":"queued","progress":0,"created_at":timestamp(),"error":None}
198
+ self.jobs[job["id"]]=job
199
+ self.executor.submit(self.run_unified_job,job["id"])
200
+ return dict(job)
201
+
202
+ def run_unified_job(self, job_id: str):
203
+ pid=self.jobs[job_id]["project_id"]
204
+ try:
205
+ self.update_job(job_id,status="running",progress=10)
206
+ self.refresh_unified(pid)
207
+ self.update_job(job_id,status="done",progress=100)
208
+ except Exception as exc:
209
+ self.update_job(job_id,status="failed",error=f"{type(exc).__name__}: {exc}")
210
+
211
  def create_mixer_job(self, request: MixerRequest) -> dict:
212
  project_id=check_id(request.project_id)
213
  with self.lock:
214
  manifest=self.manifest(project_id)
215
+ tracks = self.active_tracks(manifest)
216
+ if not tracks:
217
+ raise HTTPException(status_code=409,detail="Run separation or build BASS/NOISE tracks first.")
218
+ unknown=set(request.tracks)-set(tracks)
219
  if unknown:
220
  raise HTTPException(status_code=422,detail="Unknown real stems: "+", ".join(sorted(unknown)))
221
+ if len(tracks)>128:
222
  raise HTTPException(status_code=422,detail="More than 128 stems not supported for one render.")
223
  if (request.tempo!=1 or request.pitch_semitones!=0) and not rubberband_available():
224
  raise HTTPException(status_code=422,detail="FFmpeg with rubberband is required for independent tempo/pitch.")
 
237
  try:
238
  self.update_job(job_id,status="running",progress=3)
239
  manifest=self.manifest(project_id);folder=self.folder(project_id)
240
+ graph = manifest.get("unified")
241
+ if graph and graph.get("mix_consistent"):
242
+ sources={name:folder/data["master"] for name,data in graph["stems"].items()}
243
+ else:
244
+ sources={name:folder/"stems"/data["master"] for name,data in manifest["stems"].items()}
245
  output=folder/"mixer"/(job_id+".wav")
246
  result=render_mix(sources,settings,output,
247
  progress=lambda r:self.update_job(job_id,progress=3+int(r*90)))
 
269
  "id": "demucs", "name": "Demucs HT v4 · neural (local weights)",
270
  "ready": configured, "type": "neural_network",
271
  "stems": ["drums", "bass", "other", "vocals", "residual"],
272
+ "limitations": ("Requires trusted local htdemucs weights/config (official checkpoint 955717e8) and optional "
273
  "demucs/torch dependencies. No automatic downloads."),
274
  },
275
  ]
 
359
  with self.lock:
360
  updated = self.manifest(project_id)
361
  updated["bass"] = result
362
+ updated["unified"] = None
363
  self.write_manifest(project_id, updated)
364
+ self.update_job(job_id, progress=94)
365
+ self.refresh_unified(project_id)
366
  self.update_job(job_id, status="done", progress=100)
367
  except Exception as exc:
368
  self.update_job(job_id, status="failed", error=f"{type(exc).__name__}: {exc}")
 
424
  "tracks":tracks,"report":report,
425
  }
426
  self.write_manifest(project_id,updated)
427
+ self.update_job(job_id,progress=94)
428
+ self.refresh_unified(project_id)
429
  self.update_job(job_id,status="done",progress=100)
430
  except Exception as exc:
431
  self.update_job(job_id,status="failed",error=f"{type(exc).__name__}: {exc}")
 
514
  "preview_format": "WAV PCM 16-bit (separate listening copies)",
515
  }
516
  updated["report"] = report
517
+ # Bass estimates tied to a prior separation must not survive re-separation.
518
+ if updated.get("bass") and updated["bass"].get("mode") in ("hpss", "demucs"):
519
+ updated["bass"] = None
520
+ updated["unified"] = None
521
+ updated["mixer"] = None
522
  self.write_manifest(project_id, updated)
523
+ self.refresh_unified(project_id)
524
  self.update_job(job_id, status="done", progress=100)
525
  except Exception as exc:
526
  self.update_job(job_id, status="failed", error=f"{type(exc).__name__}: {exc}")
 
573
  if data_root is None:
574
  data_root = writable_dir(default, Path(os.environ.get("ABSOLUTION_FALLBACK_DIR", "/tmp/omega_data")))
575
  service = Service(data_root)
576
+ app = FastAPI(title="ABSOLUTION Ω", version="0.7.0",
577
  description="Local-first audio forensic vertical slice")
578
  app.state.service = service
579
  static_dir = Path(__file__).parent / "static"
 
593
 
594
  @app.get("/api/v1/health")
595
  def health():
596
+ return {"status": "ok", "version": "0.7.0", "unified_stem_graph": True, "arrangement_studio": True, "mini_tag_model_ready": MODEL_FILE.is_file(), "rubberband_ready": rubberband_available(), "basscore": True, "noisecore": True, "virtualbass_ready": virtualbass_available(), "job_backend": "process-local", "data_dir": str(service.data_root), "persistent_storage": str(service.data_root).startswith("/data"), "demucs_max_seconds": DEMUCS_MAX_SECONDS or None,
597
  "network_downloads": False, "auth": False,
598
  "max_upload_bytes": MAX_UPLOAD_BYTES,
599
  "duration_limit_seconds": None, "streaming_hpss": True}
 
807
  if not manifest.get("asset"):raise HTTPException(status_code=404,detail="No audio")
808
  path=folder/manifest["asset"]["stored_name"]
809
  else:
810
+ graph=manifest.get("unified")
811
+ active=service.active_tracks(manifest)
812
+ if stem not in active:
813
  raise HTTPException(status_code=404,detail="Unknown stem")
814
+ path=folder/active[stem]["master"] if (graph and graph.get("mix_consistent")) else folder/"stems"/active[stem]["master"]
815
  if points<64 or points>1000:
816
  raise HTTPException(status_code=422,detail="points must be 64..1000")
817
  try:return sampled_peaks(path,points)
818
  except (ValueError,RuntimeError) as exc:
819
  raise HTTPException(status_code=422,detail=str(exc)) from exc
820
 
821
+ @app.post("/api/v1/jobs/unify", status_code=202)
822
+ def start_unified(project_id: str):
823
+ return service.create_unified_job(project_id)
824
+
825
+ @app.get("/api/v1/projects/{project_id}/unified")
826
+ def get_unified(project_id: str):
827
+ m = service.manifest(project_id)
828
+ return {"unified": m.get("unified"), "active_stems": service.active_tracks(m)}
829
+
830
  @app.post("/api/v1/jobs/mix",status_code=202)
831
  def start_mix(request:MixerRequest):
832
  return service.create_mixer_job(request)
 
847
  @app.get("/api/v1/projects/{project_id}/stems")
848
  def get_stems(project_id: str):
849
  manifest = service.manifest(project_id)
850
+ return {"stems": service.active_tracks(manifest), "report": manifest["report"],
851
+ "separation": manifest["separation"], "unified": manifest.get("unified")}
852
 
853
  @app.get("/api/v1/projects/{project_id}/audio/{stem}")
854
  def get_audio(project_id: str, stem: str, preview: bool = False):
 
859
  raise HTTPException(status_code=404, detail="Audio not found.")
860
  path = folder / ("original_preview.wav" if preview else manifest["asset"]["stored_name"])
861
  else:
862
+ active = service.active_tracks(manifest)
863
+ if stem not in active:
864
  raise HTTPException(status_code=404, detail="Stem not found.")
865
  key = "preview" if preview else "master"
866
+ path = folder / active[stem][key] if (manifest.get("unified") or {}).get("mix_consistent") else folder / "stems" / active[stem][key]
867
  if not path.is_file():
868
  raise HTTPException(status_code=404, detail="Audio not found.")
869
  return FileResponse(path, media_type="audio/wav" if preview or path.suffix == ".wav" else "application/octet-stream",
 
873
  def export_stems(project_id: str):
874
  folder = service.folder(project_id)
875
  manifest = service.manifest(project_id)
876
+ active = service.active_tracks(manifest)
877
+ if not active:
878
+ raise HTTPException(status_code=409, detail="Separate audio or build BASS/NOISE lanes before exporting.")
879
  target = folder / "absolution_export.zip"
880
  # Build only from controlled server-side names. Atomic replace.
881
  tmp = folder / f".{uuid.uuid4().hex}.zip"
 
883
  archive.write(folder / "manifest.json", "manifest.json")
884
  original_path = folder / manifest["asset"]["stored_name"]
885
  archive.write(original_path, "source" + original_path.suffix)
886
+ for name,item in active.items():
887
+ path = folder/item["master"] if (manifest.get("unified") or {}).get("mix_consistent") else folder/"stems"/item["master"]
888
+ archive.write(path, "stems/" + name + ".wav")
889
  tmp.replace(target)
890
  return FileResponse(target, media_type="application/zip",
891
  filename=f"absolution_{project_id[:8]}.zip")
app/static/app.js CHANGED
@@ -9,7 +9,7 @@ const state = {
9
  projectId: null, projectName: "", asset: null, analysis: null,
10
  stems: {}, report: null, separation: null, bass: null, noise: null, virtualbass: null, virtualReady: false,
11
  model: "hpss", models: [], busy: false, playing: "original",
12
- view: "studio",
13
  };
14
  let toastTimer = null;
15
  let maxUploadBytes = 1024 * 1024 * 1024;
@@ -132,7 +132,7 @@ function loadManifest(project) {
132
  state.projectName = project.name || "Unbenannt";
133
  state.asset = project.asset;
134
  state.analysis = project.analysis;
135
- state.stems = project.stems || {};
136
  state.report = project.report;
137
  state.separation = project.separation;
138
  state.bass = project.bass || null;
@@ -508,7 +508,7 @@ async function runBass() {
508
  if(now.status==="done")break;
509
  }
510
  const project=await api(`${BASE}/projects/${state.projectId}`);
511
- state.bass=project.bass;
512
  renderBass();
513
  toast("BASS-CORE abgeschlossen! Vergleiche Bassband vorher/nachher mit deinem Gehör.");
514
  } catch(error){
@@ -662,6 +662,7 @@ async function startNoiseScan() {
662
  await pollNoiseJob(job.id,"Musikalische Impulse werden vorsichtig untersucht …");
663
  const response=await api(`${BASE}/projects/${state.projectId}/noise`);
664
  state.noise=response.noise;renderNoise();
 
665
  toast("Scan beendet. Nichts wurde verändert. Bitte jeden Kandidaten einzeln prüfen.");
666
  }catch(error){toast(`NOISE-CORE: ${error.message}`,true)}
667
  finally{setBusy(false);renderNoise();}
@@ -681,6 +682,7 @@ async function startNoiseRepair() {
681
  await pollNoiseJob(job.id,"Nur freigegebene Einzelstellen werden behutsam bearbeitet …");
682
  const response=await api(`${BASE}/projects/${state.projectId}/noise`);
683
  state.noise=response.noise;renderNoise();
 
684
  toast("Reparatur gerendert. Bitte die Differenzspur anhören.");
685
  }catch(error){toast(`NOISE-CORE: ${error.message}`,true)}
686
  finally{setBusy(false);renderNoise();}
 
9
  projectId: null, projectName: "", asset: null, analysis: null,
10
  stems: {}, report: null, separation: null, bass: null, noise: null, virtualbass: null, virtualReady: false,
11
  model: "hpss", models: [], busy: false, playing: "original",
12
+ view: "arrangement",
13
  };
14
  let toastTimer = null;
15
  let maxUploadBytes = 1024 * 1024 * 1024;
 
132
  state.projectName = project.name || "Unbenannt";
133
  state.asset = project.asset;
134
  state.analysis = project.analysis;
135
+ state.stems = project.unified?.mix_consistent ? project.unified.stems : (project.stems || {});
136
  state.report = project.report;
137
  state.separation = project.separation;
138
  state.bass = project.bass || null;
 
508
  if(now.status==="done")break;
509
  }
510
  const project=await api(`${BASE}/projects/${state.projectId}`);
511
+ loadManifest(project);
512
  renderBass();
513
  toast("BASS-CORE abgeschlossen! Vergleiche Bassband vorher/nachher mit deinem Gehör.");
514
  } catch(error){
 
662
  await pollNoiseJob(job.id,"Musikalische Impulse werden vorsichtig untersucht …");
663
  const response=await api(`${BASE}/projects/${state.projectId}/noise`);
664
  state.noise=response.noise;renderNoise();
665
+ loadManifest(await api(`${BASE}/projects/${state.projectId}`));
666
  toast("Scan beendet. Nichts wurde verändert. Bitte jeden Kandidaten einzeln prüfen.");
667
  }catch(error){toast(`NOISE-CORE: ${error.message}`,true)}
668
  finally{setBusy(false);renderNoise();}
 
682
  await pollNoiseJob(job.id,"Nur freigegebene Einzelstellen werden behutsam bearbeitet …");
683
  const response=await api(`${BASE}/projects/${state.projectId}/noise`);
684
  state.noise=response.noise;renderNoise();
685
+ loadManifest(await api(`${BASE}/projects/${state.projectId}`));
686
  toast("Reparatur gerendert. Bitte die Differenzspur anhören.");
687
  }catch(error){toast(`NOISE-CORE: ${error.message}`,true)}
688
  finally{setBusy(false);renderNoise();}
app/static/arrangement.css CHANGED
@@ -169,3 +169,71 @@ main:has(>.arr-view.active){width:100%;max-width:1840px;padding:26px 24px 22px}
169
 
170
  .arr-scroll-label{display:flex;align-items:center;justify-content:end;gap:10px;padding:8px 3px;color:#89a9b6;font-size:9px;font-weight:800;letter-spacing:.1em}
171
  .arr-scroll-label input{width:min(320px,60%);accent-color:#70e4cb}
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
169
 
170
  .arr-scroll-label{display:flex;align-items:center;justify-content:end;gap:10px;padding:8px 3px;color:#89a9b6;font-size:9px;font-weight:800;letter-spacing:.1em}
171
  .arr-scroll-label input{width:min(320px,60%);accent-color:#70e4cb}
172
+
173
+ /* Ω v0.7 AURORA-inspired instrument laboratory:
174
+ original styling, no third-party fonts, image assets or component code. */
175
+ .arr-view{--arr-gold:#eccf91;--arr-teal:#79f2d5}
176
+ .arr-view::before{content:"";position:fixed;z-index:-1;inset:0;pointer-events:none;
177
+ background:radial-gradient(ellipse at 79% 16%,#18577335,transparent 45%),
178
+ radial-gradient(ellipse at 14% 86%,#1b85602b,transparent 46%)}
179
+ .arr-card,.omega-operations{
180
+ background:linear-gradient(125deg,rgba(24,41,57,.91),rgba(9,20,31,.97) 68%),
181
+ linear-gradient(90deg,#193346,#0d1a29);
182
+ box-shadow:inset 0 1px #a3e9df18,0 18px 58px #02070c55;
183
+ backdrop-filter:blur(18px);
184
+ }
185
+ .arr-view h1,.arr-card h2{letter-spacing:-.045em}
186
+ .omega-operations{margin:18px 0 22px;padding:23px;border:1px solid #426978;border-radius:15px;
187
+ position:relative;overflow:hidden}
188
+ .omega-operations::after{content:"";position:absolute;inset:0;pointer-events:none;border-radius:15px;
189
+ background:linear-gradient(132deg,#b7f5e80c,transparent 24%,transparent 80%,#ead49c0b)}
190
+ .omega-ops-head{display:flex;align-items:start;justify-content:space-between;gap:18px;margin-bottom:17px}
191
+ .omega-ops-head h2{font-size:25px;color:#e9f8f8;margin:6px 0}
192
+ .omega-ops-head h2 em{font-style:normal;color:#efcb84}
193
+ .omega-ops-head p{color:#94b5c3;font-size:12px;line-height:1.6}
194
+ .omega-procedure-grid{display:grid;grid-template-columns:1fr 1fr;gap:13px;margin-bottom:15px}
195
+ .omega-procedure{border:1px solid #36556b;background:linear-gradient(146deg,#0f2434dc,#0c1a29);
196
+ padding:18px;border-radius:12px;box-shadow:inset 0 1px #e0f4ff13}
197
+ .omega-noise-lab{border-color:#634a60}
198
+ .omega-procedure-title{display:flex;align-items:center;gap:9px;margin-bottom:9px;flex-wrap:wrap}
199
+ .omega-dot{width:8px;height:8px;border-radius:50%;background:var(--node-color);box-shadow:0 0 16px var(--node-color)}
200
+ .omega-procedure-title strong{letter-spacing:.1em;color:#e5f9f7;font-size:12px}
201
+ .omega-procedure-title small{color:#87a6bd;font-size:9px;font-weight:800;letter-spacing:.11em;margin-left:auto}
202
+ .omega-procedure p{color:#a4b9c8;font-size:12px;line-height:1.55;margin:0 0 14px}
203
+ .omega-safety{display:block;color:#839aaf;margin-top:12px;font-size:10px;line-height:1.6}
204
+ .omega-field{display:flex;gap:14px;align-items:center;flex-wrap:wrap;font-weight:700;
205
+ color:#c0e3e3;font-size:11px;letter-spacing:.03em;margin:12px 0}
206
+ .omega-field input{flex:1 1 100%;width:100%;accent-color:#83eacb}
207
+ .omega-field output{color:#81f1d4;margin-left:auto;font-variant-numeric:tabular-nums}
208
+ .omega-candidates{max-height:125px;overflow:auto;border:1px solid #33495b;border-radius:8px;
209
+ background:#071724a8;padding:10px;display:flex;flex-wrap:wrap;gap:6px;color:#8fabb9;font-size:10px}
210
+ .omega-candidate{display:flex;align-items:center;gap:5px;border:1px solid #314558;padding:6px;
211
+ border-radius:6px;color:#cee0ea;background:#122637}
212
+ .omega-candidate input{accent-color:#f49ebb}
213
+ .omega-noise-actions{display:flex;align-items:center;gap:8px;flex-wrap:wrap}
214
+ .omega-diagnostics{display:grid;grid-template-columns:repeat(4,minmax(0,1fr));gap:9px;margin-bottom:14px}
215
+ .omega-diagnostic{padding:13px;background:#0a1a28bb;border:1px solid #304b5c;border-radius:10px;
216
+ display:grid;gap:7px;min-width:0}
217
+ .omega-diagnostic span{font-size:9px;font-weight:800;color:#83a4b6;letter-spacing:.1em}
218
+ .omega-diagnostic strong{font-size:12px;overflow-wrap:anywhere;color:#b5f1e1;font-variant-numeric:tabular-nums}
219
+ .omega-track-table-wrap{overflow:auto;border:1px solid #355365;border-radius:10px}
220
+ .omega-track-table{border-collapse:collapse;width:100%;min-width:620px;font-size:11px;color:#b8d5e3}
221
+ .omega-track-table thead{background:#183444;color:#8fdccc;text-align:left;letter-spacing:.1em;font-size:9px}
222
+ .omega-track-table th,.omega-track-table td{padding:11px 12px;border-bottom:1px solid #253e50;text-align:left}
223
+ .omega-track-table tbody tr:nth-child(2n){background:#0d223064}
224
+ .omega-track-table td:first-child{font-weight:750;color:#e6f3f4}
225
+ .omega-key{display:inline-block;width:8px;height:8px;border-radius:3px;margin-right:7px;box-shadow:0 0 9px currentColor}
226
+ .omega-procedure-warning{margin-top:12px!important;color:#afc4c8;font-size:11px;line-height:1.5}
227
+ .arr-track:has(.arr-track-name strong[title="noise_difference"]){box-shadow:inset 3px 0 #f09bbd}
228
+ @media(max-width:1100px){
229
+ .omega-diagnostics{grid-template-columns:repeat(2,minmax(0,1fr))}
230
+ }
231
+ @media(max-width:680px){
232
+ .omega-operations{padding:13px}
233
+ .omega-ops-head{flex-wrap:wrap}
234
+ .omega-procedure-grid{grid-template-columns:1fr}
235
+ .omega-procedure{padding:13px}
236
+ .omega-diagnostics{gap:7px}
237
+ .omega-diagnostic{padding:10px}
238
+ .omega-diagnostic strong{font-size:11px}
239
+ }
app/static/arrangement.js CHANGED
@@ -11,7 +11,7 @@
11
  const G={
12
  inventory:null,plan:[],real:[],peaks:{},settings:{},zoom:1,scroll:0,
13
  mixed:null,projectId:null,rubberband:false,serverModel:true,rendering:false,
14
- length:0
15
  };
16
  const escape=(x)=>String(x??"").replace(/[&<>"']/g,c=>({"&":"&amp;","<":"&lt;",">":"&gt;",'"':"&quot;","'":"&#39;"}[c]));
17
  const dt=t=>{if(!Number.isFinite(t))return "—";let h=Math.floor(t/3600),m=Math.floor(t%3600/60),s=Math.floor(t%60);return h?`${h}:${String(m).padStart(2,"0")}:${String(s).padStart(2,"0")}`:`${m}:${String(s).padStart(2,"0")}`;};
@@ -23,7 +23,7 @@
23
  }
24
  function updateStatus(){
25
  byId("arrSourceName").textContent=state.asset?.filename||"Noch keine Datei — importiere im Studio oder lade die Demo.";
26
- byId("arrStemStatus").textContent=`${G.real.length} echte Stems`;
27
  byId("arrActiveTracks").textContent=`${G.real.length} reale Spuren`;
28
  byId("arrInspect").disabled=!state.asset||G.rendering;
29
  byId("arrSeparate").disabled=!state.asset||G.rendering;
@@ -39,7 +39,8 @@
39
  byId("arrRenderReport").textContent=G.real.length?"Änderungen ungerendert":"Noch kein Mix berechnet";
40
  }
41
  function reset(){
42
- G.projectId=null;G.inventory=null;G.plan=[];G.real=[];G.peaks={};G.settings={};G.mixed=null;G.length=0;
 
43
  byId("arrPlanContent").innerHTML='<div class="arr-empty-inline">Noch keine Vorinventur. Klicke auf „Mini-KI analysieren“.</div>';
44
  renderTracks();updateStatus();setUnrendered();draw();
45
  }
@@ -48,9 +49,15 @@
48
  const different=G.projectId!==projectId;
49
  G.projectId=projectId;G.inventory=manifest.inventory||null;
50
  G.plan=(manifest.plan||[]).map(p=>({...p}));
51
- G.real=Object.keys(manifest.stems||{});
 
 
 
52
  G.length=Number(manifest.analysis?.duration_seconds)||0;
53
- if(different){G.peaks={};G.settings={};G.mixed=null;}
 
 
 
54
  const settings=manifest.mixer?.settings?.tracks||{};
55
  if(different){for(const [n,p] of Object.entries(settings))G.settings[n]={...p};}
56
  if(different && manifest.mixer)G.mixed=manifest.mixer;
@@ -62,7 +69,7 @@
62
  byId("arrTempoOut").textContent=tempo+" %";
63
  byId("arrPitchOut").textContent=(pitch>=0?"+":"")+pitch+" HT";
64
  }
65
- renderPlan();renderTracks();updateStatus();
66
  if(G.mixed)setMixed(G.mixed);
67
  else setUnrendered();
68
  await Promise.all([
@@ -135,14 +142,14 @@
135
  notice("Trennplan gespeichert. Noch keine Audiospur wurde dadurch erzeugt.");
136
  }catch(e){notice("Trennplan: "+e.message,true);}
137
  }
138
- function defaultSettings(){return {gain_db:0,pan:0,mute:false,solo:false,highpass_hz:0,lowpass_hz:0,width:1,drive:0};}
139
- function getSettings(name){return G.settings[name]||(G.settings[name]=defaultSettings());}
140
- function colorFor(index){return COLORS[index%COLORS.length];}
141
  function pretty(name){
142
  const labels={vocals:"Gesang / Vocals",drums:"Schlagzeug / Drums",
143
  bass:"Bass",other:"Andere Instrumente",harmonic:"Harmonische Klanganteile",
144
  percussive:"Perkussive Klanganteile",residual:"Residuum / Restmaterial"};
145
- return labels[name]||name.replace(/[_-]/g," ");
146
  }
147
  function renderTracks(){
148
  const wrap=byId("arrTracks");
@@ -154,7 +161,7 @@
154
  const p=getSettings(name),color=colorFor(i);
155
  return `<article class="arr-track" style="--arr-color:${color}" data-index="${i}">
156
  <div class="arr-track-left">
157
- <div class="arr-track-name"><span class="arr-track-dot"></span><div><strong title="${escape(name)}">${escape(pretty(name))}</strong><small>STEM ${String(i+1).padStart(2,"0")} · REAL AUDIO</small></div>
158
  <button type="button" class="arr-track-preview" data-action="play" title="Diese Spur einzeln hören" aria-label="${escape(pretty(name))} anhören">▶</button>
159
  </div>
160
  <div class="arr-track-ctrl"><button class="arr-mute ${p.mute?"on":""}" type="button" data-action="mute" aria-pressed="${!!p.mute}" title="Mute">M</button>
@@ -179,6 +186,83 @@
179
  }).join("");
180
  updateStatus();requestAnimationFrame(draw);
181
  }
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
182
  function outputValue(param,p){
183
  switch(param){
184
  case"gain_db":return `${p>0?"+":""}${p} dB`;
@@ -307,6 +391,12 @@
307
  }
308
  function listeners(){
309
  byId("arrDemo").addEventListener("click",loadDemo);
 
 
 
 
 
 
310
  byId("arrInspect").addEventListener("click",inspect);
311
  byId("arrSavePlan").addEventListener("click",savePlan);
312
  byId("arrAddPlan").addEventListener("click",()=>{
@@ -326,7 +416,7 @@
326
  selectMethod(selected);runSeparation();
327
  });
328
  byId("arrResetMix").addEventListener("click",()=>{
329
- G.settings={};for(const n of G.real)G.settings[n]=defaultSettings();
330
  byId("arrTempo").value=100;byId("arrPitch").value=0;byId("arrTempoOut").textContent="100 %";byId("arrPitchOut").textContent="±0 HT";
331
  renderTracks();setUnrendered();
332
  });
 
11
  const G={
12
  inventory:null,plan:[],real:[],peaks:{},settings:{},zoom:1,scroll:0,
13
  mixed:null,projectId:null,rubberband:false,serverModel:true,rendering:false,
14
+ length:0,meta:{},graph:null,noise:null
15
  };
16
  const escape=(x)=>String(x??"").replace(/[&<>"']/g,c=>({"&":"&amp;","<":"&lt;",">":"&gt;",'"':"&quot;","'":"&#39;"}[c]));
17
  const dt=t=>{if(!Number.isFinite(t))return "—";let h=Math.floor(t/3600),m=Math.floor(t%3600/60),s=Math.floor(t%60);return h?`${h}:${String(m).padStart(2,"0")}:${String(s).padStart(2,"0")}`:`${m}:${String(s).padStart(2,"0")}`;};
 
23
  }
24
  function updateStatus(){
25
  byId("arrSourceName").textContent=state.asset?.filename||"Noch keine Datei — importiere im Studio oder lade die Demo.";
26
+ byId("arrStemStatus").textContent=`${G.real.length} additive Audiospuren`;
27
  byId("arrActiveTracks").textContent=`${G.real.length} reale Spuren`;
28
  byId("arrInspect").disabled=!state.asset||G.rendering;
29
  byId("arrSeparate").disabled=!state.asset||G.rendering;
 
39
  byId("arrRenderReport").textContent=G.real.length?"Änderungen ungerendert":"Noch kein Mix berechnet";
40
  }
41
  function reset(){
42
+ G.projectId=null;G.inventory=null;G.plan=[];G.real=[];G.peaks={};G.settings={};G.mixed=null;G.length=0;G.meta={};G.graph=null;G.noise=null;
43
+ renderOperations();
44
  byId("arrPlanContent").innerHTML='<div class="arr-empty-inline">Noch keine Vorinventur. Klicke auf „Mini-KI analysieren“.</div>';
45
  renderTracks();updateStatus();setUnrendered();draw();
46
  }
 
49
  const different=G.projectId!==projectId;
50
  G.projectId=projectId;G.inventory=manifest.inventory||null;
51
  G.plan=(manifest.plan||[]).map(p=>({...p}));
52
+ G.graph=manifest.unified||null;
53
+ G.meta=manifest.unified?.mix_consistent ? manifest.unified.stems : (manifest.stems||{});
54
+ G.real=Object.keys(G.meta);
55
+ G.noise=manifest.noise||null;
56
  G.length=Number(manifest.analysis?.duration_seconds)||0;
57
+ const roster=G.real.join("|");
58
+ if(different || G.lastRoster!==roster){G.peaks={};G.settings={};G.mixed=null;}
59
+ G.lastRoster=roster;
60
+ for(const n of G.real)if(!G.settings[n])G.settings[n]=defaultSettings(n);
61
  const settings=manifest.mixer?.settings?.tracks||{};
62
  if(different){for(const [n,p] of Object.entries(settings))G.settings[n]={...p};}
63
  if(different && manifest.mixer)G.mixed=manifest.mixer;
 
69
  byId("arrTempoOut").textContent=tempo+" %";
70
  byId("arrPitchOut").textContent=(pitch>=0?"+":"")+pitch+" HT";
71
  }
72
+ renderPlan();renderTracks();renderOperations();updateStatus();
73
  if(G.mixed)setMixed(G.mixed);
74
  else setUnrendered();
75
  await Promise.all([
 
142
  notice("Trennplan gespeichert. Noch keine Audiospur wurde dadurch erzeugt.");
143
  }catch(e){notice("Trennplan: "+e.message,true);}
144
  }
145
+ function defaultSettings(name=""){return {gain_db:0,pan:0,mute:false,solo:false,highpass_hz:0,lowpass_hz:0,width:1,drive:0};}
146
+ function getSettings(name){return G.settings[name]||(G.settings[name]=defaultSettings(name));}
147
+ function colorFor(index){const n=G.real[index];return G.meta[n]?.color||COLORS[index%COLORS.length];}
148
  function pretty(name){
149
  const labels={vocals:"Gesang / Vocals",drums:"Schlagzeug / Drums",
150
  bass:"Bass",other:"Andere Instrumente",harmonic:"Harmonische Klanganteile",
151
  percussive:"Perkussive Klanganteile",residual:"Residuum / Restmaterial"};
152
+ return G.meta[name]?.label||labels[name]||name.replace(/[_-]/g," ");
153
  }
154
  function renderTracks(){
155
  const wrap=byId("arrTracks");
 
161
  const p=getSettings(name),color=colorFor(i);
162
  return `<article class="arr-track" style="--arr-color:${color}" data-index="${i}">
163
  <div class="arr-track-left">
164
+ <div class="arr-track-name"><span class="arr-track-dot"></span><div><strong title="${escape(name)}">${escape(pretty(name))}</strong><small>STEM ${String(i+1).padStart(2,"0")} · ${escape(G.meta[name]?.role==="repair_difference"?"FREIGEGEBENE STÖRDIFFERENZ":G.meta[name]?.role==="derived_bass"?"BASSBAND-SCHÄTZUNG":"KOMPENSIERTE QUELLE")}</small></div>
165
  <button type="button" class="arr-track-preview" data-action="play" title="Diese Spur einzeln hören" aria-label="${escape(pretty(name))} anhören">▶</button>
166
  </div>
167
  <div class="arr-track-ctrl"><button class="arr-mute ${p.mute?"on":""}" type="button" data-action="mute" aria-pressed="${!!p.mute}" title="Mute">M</button>
 
186
  }).join("");
187
  updateStatus();requestAnimationFrame(draw);
188
  }
189
+ function renderOperations(){
190
+ const g=G.graph,hasAudio=Boolean(state?.asset);
191
+ byId("arrGraphStatus").textContent=g?.mix_consistent?"KONSISTENT":G.real.length?"BASISSTEMS":"WARTET AUF QUELLEN";
192
+ byId("arrGraphError").textContent=g?.mix_consistent?
193
+ `${Number(g.mix_consistency_max_error).toExponential(2)} max abs`:"—";
194
+ byId("arrBassStatus").textContent=G.real.includes("kick_core")?"Kick + Bassline aktiv":"Nicht berechnet";
195
+ byId("arrNoiseStatus").textContent=G.real.includes("noise_difference")?"Freigegebene Differenz":"Nicht repariert";
196
+ for(const id of ["arrBuildBass","arrScanNoise","arrUnify"]){
197
+ byId(id).disabled=!hasAudio||G.rendering;
198
+ }
199
+ const candidates=G.noise?.scan?.events||[];
200
+ const c=byId("arrNoiseCandidates");
201
+ if(!G.noise?.scan)c.textContent="Noch kein Scan. Keine automatische Stör-Entfernung.";
202
+ else if(!candidates.length)c.textContent="Keine Klick-Hypothesen gefunden; Original unverändert.";
203
+ else {
204
+ c.innerHTML=candidates.slice(0,120).map(e=>
205
+ `<label class="omega-candidate"><input type="checkbox" value="${Number(e.sample)}"> `+
206
+ `Sample ${Number(e.sample).toLocaleString("de-DE")} · ${Number(e.time_seconds??0).toFixed(3)} s · UNGEWISS</label>`).join("");
207
+ }
208
+ byId("arrRepairNoise").disabled=!hasAudio||!candidates.length||G.rendering;
209
+ const rows=Object.entries(G.meta);
210
+ byId("arrGraphTable").innerHTML=rows.length?rows.map(([name,meta])=>
211
+ `<tr><td><span class="omega-key" style="background:${escape(meta.color||"#8ea2bd")}"></span>`+
212
+ `${escape(meta.label||name)}</td><td>${escape(meta.role||"separiert")}</td>`+
213
+ `<td>${meta.role==="repair_difference"?"FREIGEGEBEN":meta.role==="derived_bass"?"SCHÄTZUNG":"REAL AUDIO"}</td>`+
214
+ `<td>${name==="noise_difference"?"MUTE nur bei Wunsch · Original bei ON":meta.role==="base_compensated"?"Kompensiert / nicht verdoppelt":"Frei mischbar"}</td></tr>`).join(""):
215
+ `<tr><td colspan="4">Zunächst Datei importieren und BASS-CORE oder Separation starten.</td></tr>`;
216
+ }
217
+ async function waitJob(job,caption){
218
+ const pid=G.projectId;
219
+ const label=byId("arrOpsProgress");
220
+ while(true){
221
+ await new Promise(r=>setTimeout(r,500));
222
+ const status=await request("/jobs/"+encodeURIComponent(job.id));
223
+ if(status.status==="failed")throw new Error(status.error||"Verarbeitung fehlgeschlagen");
224
+ if(pid!==G.projectId)throw new Error("Projektwechsel während des Jobs.");
225
+ label.textContent=`${caption}: ${status.progress||0} %`;
226
+ if(status.status==="done")break;
227
+ }
228
+ // Use the canonical manifest loader so every view sees the same real stems.
229
+ const manifest=await request(`/projects/${encodeURIComponent(pid)}`);
230
+ loadManifest(manifest);
231
+ label.textContent=`✓ ${caption} abgeschlossen. Der Originalmix wurde nicht überschrieben.`;
232
+ }
233
+ async function runOperation(action){
234
+ if(!state.asset||G.rendering)return;
235
+ G.rendering=true;renderOperations();updateStatus();
236
+ try{
237
+ let job,caption;
238
+ if(action==="bass"){
239
+ caption="Kick/Bassline vorbereiten";
240
+ const cfg={project_id:G.projectId,source_mode:"auto",
241
+ crossover_hz:Number(byId("arrBassCutoff").value),
242
+ kick_gain_db:0,bass_gain_db:0,sub_gain_db:0,
243
+ harmonics:0,lowcut_hz:22,wet:0};
244
+ job=await request("/jobs/bass",{method:"POST",headers:{"content-type":"application/json"},
245
+ body:JSON.stringify(cfg)});
246
+ }else if(action==="scan"){
247
+ caption="Stör-Hypothesen scannen";
248
+ job=await request("/jobs/noise/scan",{method:"POST",headers:{"content-type":"application/json"},
249
+ body:JSON.stringify({project_id:G.projectId})});
250
+ }else if(action==="repair"){
251
+ const picks=[...byId("arrNoiseCandidates").querySelectorAll("input:checked")].map(x=>Number(x.value));
252
+ if(!picks.length)throw new Error("Bitte zuerst tatsächliche Klickkandidaten auswählen.");
253
+ if(!confirm(`${picks.length} Klickstelle(n) reparieren? Die Differenz kann musikalische Signalanteile enthalten.`))return;
254
+ caption="Freigegebene Störspur erzeugen";
255
+ job=await request("/jobs/noise/repair",{method:"POST",headers:{"content-type":"application/json"},
256
+ body:JSON.stringify({project_id:G.projectId,approved_samples:picks,strength:Number(byId("arrNoiseStrength").value)/100})});
257
+ }else{
258
+ caption="Spurgraph rekonstruieren";
259
+ job=await request(`/jobs/unify?project_id=${encodeURIComponent(G.projectId)}`,{method:"POST"});
260
+ }
261
+ await waitJob(job,caption);
262
+ notice(caption+" abgeschlossen — Basis-Stems und Differenzen zusammengeführt.");
263
+ }catch(e){byId("arrOpsProgress").textContent="Fehler: "+e.message;notice("Studio: "+e.message,true);}
264
+ finally{G.rendering=false;renderOperations();updateStatus();}
265
+ }
266
  function outputValue(param,p){
267
  switch(param){
268
  case"gain_db":return `${p>0?"+":""}${p} dB`;
 
391
  }
392
  function listeners(){
393
  byId("arrDemo").addEventListener("click",loadDemo);
394
+ byId("arrImport").addEventListener("click",()=>byId("fileInput").click());
395
+ byId("arrBassCutoff").addEventListener("input",e=>byId("arrBassCutoffOut").textContent=e.target.value+" Hz");
396
+ byId("arrNoiseStrength").addEventListener("input",e=>byId("arrNoiseStrengthOut").textContent=e.target.value+" %");
397
+ for(const [id,mode] of [["arrBuildBass","bass"],["arrScanNoise","scan"],
398
+ ["arrRepairNoise","repair"],["arrUnify","unify"]])
399
+ byId(id).addEventListener("click",()=>runOperation(mode));
400
  byId("arrInspect").addEventListener("click",inspect);
401
  byId("arrSavePlan").addEventListener("click",savePlan);
402
  byId("arrAddPlan").addEventListener("click",()=>{
 
416
  selectMethod(selected);runSeparation();
417
  });
418
  byId("arrResetMix").addEventListener("click",()=>{
419
+ G.settings={};for(const n of G.real)G.settings[n]=defaultSettings(n);
420
  byId("arrTempo").value=100;byId("arrPitch").value=0;byId("arrTempoOut").textContent="100 %";byId("arrPitchOut").textContent="±0 HT";
421
  renderTracks();setUnrendered();
422
  });
app/static/index.html CHANGED
@@ -22,8 +22,8 @@
22
  </a>
23
  <div class="nav-group-label">WORKSPACE</div>
24
  <nav class="nav-stack" aria-label="Bereiche">
25
- <button class="nav-link active" data-view="studio" aria-current="page"><span class="nav-icon">◈</span> Studio <span class="nav-number">01</span></button>
26
- <button class="nav-link" data-view="arrangement"><span class="nav-icon">▥</span> Arrangement <span class="nav-number">NEU</span></button>
27
  <button class="nav-link" data-view="bass"><span class="nav-icon">♮</span> BASS-CORE <span class="nav-number">02</span></button>
28
  <button class="nav-link" data-view="noise"><span class="nav-icon">⌁</span> NOISE-CORE <span class="nav-number">03</span></button>
29
  <button class="nav-link" data-view="stems"><span class="nav-icon">≋</span> Meine Stems <span class="nav-count" id="stem-count">0</span></button>
@@ -41,13 +41,13 @@
41
  <span class="privacy-icon">◉</span>
42
  <div><strong>Lokale Verarbeitung</strong><small>Kein Cloud-Upload durch diese App</small></div>
43
  </div>
44
- <div class="sidebar-foot">ABSOLUTION Ω · v0.6 ARRANGEMENT<br>Forschungsversion · Single User</div>
45
  </aside>
46
 
47
  <div class="content-shell">
48
  <header class="topbar">
49
  <button id="menuToggle" class="menu-toggle icon-button" aria-label="Menü öffnen" aria-expanded="false">☰</button>
50
- <div class="crumb"><span class="crumb-muted">WORKSPACE</span><span class="crumb-sep">/</span><span id="crumbCurrent">STUDIO</span></div>
51
  <div class="top-right">
52
  <span class="api-indicator" id="apiIndicator"><i></i><span id="apiLabel">Engine wird geprüft</span></span>
53
  <button class="top-help" data-goto="guide" type="button">Hilfe <span>↗</span></button>
@@ -55,14 +55,15 @@
55
  </header>
56
 
57
  <main id="main-content" tabindex="-1">
58
- <section id="view-arrangement" class="view arr-view" aria-label="Mehrspur-Arrangement Studio">
59
  <div class="arr-header">
60
  <div class="arr-title">
61
- <div class="arr-kicker"><span class="arr-light"></span> ABSOLUTION Ω / MULTITRACK WORKSTATION <span class="arr-beta">v0.6</span></div>
62
  <h1>Alle Frequenzen. <em>Alle Ebenen.</em></h1>
63
  <p>KI-Vorinventur → editierbarer Trennplan → echte Stems → präziser Mix. Originaltreue steht über aggressiver Verarbeitung.</p>
64
  </div>
65
  <div class="arr-header-actions">
 
66
  <button class="arr-btn arr-btn-ghost" id="arrDemo">♫ Demo laden</button>
67
  <button class="arr-btn arr-btn-primary" id="arrInspect">✧ Mini-KI analysieren</button>
68
  </div>
@@ -92,6 +93,47 @@
92
  <div class="arr-help-text" id="arrPlanHint">Der Trennplan kann mehr Ziele enthalten als das gewählte Modell separat erzeugen kann. Nicht verfügbare Instrumente bleiben Planungseinträge, keine Audiodateien.</div>
93
  </section>
94
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
95
  <section class="arr-console arr-card">
96
  <div class="arr-console-top">
97
  <div class="arr-console-heading"><span class="arr-micro">MULTI-STEM WORKSPACE / MIXER</span><h2>Arrangement &amp; Frequenzspuren</h2></div>
@@ -138,7 +180,7 @@
138
  <p class="arr-disclaimer">Präzisionshinweis: Farbige Wellenformen sind Visualisierungen realer, geschätzter Stems; die Mini-KI liefert nur Textur-Hypothesen. Der Offline-Mix kann mit Filtern, Sättigung, Pan und Transposition hörbar vom Original abweichen und wird deshalb separat gerendert.</p>
139
  </section>
140
 
141
- <section id="view-studio" class="view active" aria-label="Audio-Studio">
142
  <div class="hero">
143
  <div class="hero-copy">
144
  <div class="kicker"><span class="kicker-dot"></span> INTELLIGENT AUDIO WORKSPACE <span class="kicker-index">/ 001</span></div>
 
22
  </a>
23
  <div class="nav-group-label">WORKSPACE</div>
24
  <nav class="nav-stack" aria-label="Bereiche">
25
+ <button class="nav-link" data-view="studio"><span class="nav-icon">◈</span> Audio-Import <span class="nav-number">01</span></button>
26
+ <button class="nav-link active" data-view="arrangement" aria-current="page"><span class="nav-icon">▥</span> Ω STUDIO <span class="nav-number">NEU</span></button>
27
  <button class="nav-link" data-view="bass"><span class="nav-icon">♮</span> BASS-CORE <span class="nav-number">02</span></button>
28
  <button class="nav-link" data-view="noise"><span class="nav-icon">⌁</span> NOISE-CORE <span class="nav-number">03</span></button>
29
  <button class="nav-link" data-view="stems"><span class="nav-icon">≋</span> Meine Stems <span class="nav-count" id="stem-count">0</span></button>
 
41
  <span class="privacy-icon">◉</span>
42
  <div><strong>Lokale Verarbeitung</strong><small>Kein Cloud-Upload durch diese App</small></div>
43
  </div>
44
+ <div class="sidebar-foot">ABSOLUTION Ω · v0.7 UNIFIED STUDIO<br>Forschungsversion · Single User</div>
45
  </aside>
46
 
47
  <div class="content-shell">
48
  <header class="topbar">
49
  <button id="menuToggle" class="menu-toggle icon-button" aria-label="Menü öffnen" aria-expanded="false">☰</button>
50
+ <div class="crumb"><span class="crumb-muted">WORKSPACE</span><span class="crumb-sep">/</span><span id="crumbCurrent">Ω STUDIO</span></div>
51
  <div class="top-right">
52
  <span class="api-indicator" id="apiIndicator"><i></i><span id="apiLabel">Engine wird geprüft</span></span>
53
  <button class="top-help" data-goto="guide" type="button">Hilfe <span>↗</span></button>
 
55
  </header>
56
 
57
  <main id="main-content" tabindex="-1">
58
+ <section id="view-arrangement" class="view active arr-view" aria-label="Mehrspur-Arrangement Studio">
59
  <div class="arr-header">
60
  <div class="arr-title">
61
+ <div class="arr-kicker"><span class="arr-light"></span> ABSOLUTION Ω / MULTITRACK WORKSTATION <span class="arr-beta">v0.7</span></div>
62
  <h1>Alle Frequenzen. <em>Alle Ebenen.</em></h1>
63
  <p>KI-Vorinventur → editierbarer Trennplan → echte Stems → präziser Mix. Originaltreue steht über aggressiver Verarbeitung.</p>
64
  </div>
65
  <div class="arr-header-actions">
66
+ <button class="arr-btn arr-btn-ghost" id="arrImport">+ Datei importieren</button>
67
  <button class="arr-btn arr-btn-ghost" id="arrDemo">♫ Demo laden</button>
68
  <button class="arr-btn arr-btn-primary" id="arrInspect">✧ Mini-KI analysieren</button>
69
  </div>
 
93
  <div class="arr-help-text" id="arrPlanHint">Der Trennplan kann mehr Ziele enthalten als das gewählte Modell separat erzeugen kann. Nicht verfügbare Instrumente bleiben Planungseinträge, keine Audiodateien.</div>
94
  </section>
95
 
96
+ <section class="omega-operations arr-card" aria-label="BASS-CORE und NOISE-CORE im selben Studio">
97
+ <div class="omega-ops-head">
98
+ <div><span class="arr-micro">AURORA INSPIRED · AUDIO SURGICAL DECK</span>
99
+ <h2>Element-Kontrolle <em>Ω</em></h2>
100
+ <p>Separate Zusatz-Lanes ohne Signalverdopplung: alle Quellen werden beim Synchronisieren kompensiert.</p></div>
101
+ <button type="button" id="arrUnify" class="arr-btn arr-btn-ghost">↻ Spurgraph prüfen</button>
102
+ </div>
103
+ <div class="omega-procedure-grid">
104
+ <div class="omega-procedure omega-bass-lab">
105
+ <div class="omega-procedure-title"><span class="omega-dot" style="--node-color:#6feccc"></span>
106
+ <strong>BASS-CORE</strong><small>SUB · KICK · BASSLINE</small></div>
107
+ <p>Schätzt Kick- und Bassline-Tiefbänder. Diese werden aus den Ursprungsstems subtrahiert und separat mischbar.</p>
108
+ <label class="omega-field">Trennfrequenz <output id="arrBassCutoffOut">180 Hz</output>
109
+ <input id="arrBassCutoff" type="range" min="65" max="350" step="5" value="180"></label>
110
+ <button type="button" id="arrBuildBass" class="arr-btn arr-btn-primary">+ Kick &amp; Bassline als Spuren</button>
111
+ <small class="omega-safety">DSP-Schätzung, keine garantierte Instrumenten-Isolation. Kein automatisches Bass-Enhancement.</small>
112
+ </div>
113
+ <div class="omega-procedure omega-noise-lab">
114
+ <div class="omega-procedure-title"><span class="omega-dot" style="--node-color:#f69cb8"></span>
115
+ <strong>NOISE-CORE</strong><small>FORENSIC · REPAIR DIFFERENCE</small></div>
116
+ <p>Störungen nur lokalisieren. Erst bestätigte Klickstellen erzeugen eine eigene, standardmäßig hörbare Störspur. Erst Mute entfernt den geprüften Anteil.</p>
117
+ <div class="omega-noise-actions"><button type="button" id="arrScanNoise" class="arr-btn arr-btn-ghost">⌕ Störungen scannen</button>
118
+ <button type="button" id="arrRepairNoise" class="arr-btn arr-btn-primary">✓ Auswahl reparieren</button></div>
119
+ <div id="arrNoiseCandidates" class="omega-candidates">Noch kein Scan. Ohne bestätigte Reparatur keine Störspur.</div>
120
+ <label class="omega-field">Reparaturstärke <output id="arrNoiseStrengthOut">65 %</output>
121
+ <input id="arrNoiseStrength" type="range" min="5" max="100" value="65" step="5"></label>
122
+ </div>
123
+ </div>
124
+ <div class="omega-diagnostics">
125
+ <div class="omega-diagnostic"><span>GRAPH STATUS</span><strong id="arrGraphStatus">Noch nicht berechnet</strong></div>
126
+ <div class="omega-diagnostic"><span>MIX CONSISTENCY</span><strong id="arrGraphError">—</strong></div>
127
+ <div class="omega-diagnostic"><span>KICK / BASS</span><strong id="arrBassStatus">Nicht vorhanden</strong></div>
128
+ <div class="omega-diagnostic"><span>NOISE DIFFERENCE</span><strong id="arrNoiseStatus">Nicht vorhanden</strong></div>
129
+ </div>
130
+ <div class="omega-track-table-wrap">
131
+ <table class="omega-track-table"><thead><tr><th>SPUR</th><th>QUELLE / METHODIK</th><th>STATUS</th><th>ROLLE IM MIX</th></tr></thead>
132
+ <tbody id="arrGraphTable"><tr><td colspan="4">Noch keine Audiospuren generiert.</td></tr></tbody></table>
133
+ </div>
134
+ <p class="omega-procedure-warning" id="arrOpsProgress" role="status">Neutraler Mix = Original · Störspur hörbar = Original; Mute = gezielte Reparatur · Originaldatei bleibt unverändert.</p>
135
+ </section>
136
+
137
  <section class="arr-console arr-card">
138
  <div class="arr-console-top">
139
  <div class="arr-console-heading"><span class="arr-micro">MULTI-STEM WORKSPACE / MIXER</span><h2>Arrangement &amp; Frequenzspuren</h2></div>
 
180
  <p class="arr-disclaimer">Präzisionshinweis: Farbige Wellenformen sind Visualisierungen realer, geschätzter Stems; die Mini-KI liefert nur Textur-Hypothesen. Der Offline-Mix kann mit Filtern, Sättigung, Pan und Transposition hörbar vom Original abweichen und wird deshalb separat gerendert.</p>
181
  </section>
182
 
183
+ <section id="view-studio" class="view" aria-label="Audio-Studio">
184
  <div class="hero">
185
  <div class="hero-copy">
186
  <div class="kicker"><span class="kicker-dot"></span> INTELLIGENT AUDIO WORKSPACE <span class="kicker-index">/ 001</span></div>
app/unified.py ADDED
@@ -0,0 +1,207 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """ABSOLUTION Ω — canonical, non-duplicating multitrack source graph.
2
+
3
+ Bass estimates and an explicitly approved click-difference become audible
4
+ lanes, NOT additional energy. A corresponding compensation is subtracted
5
+ from the underlying real stems. In neutral settings all lanes sum to the
6
+ original mixture up to the original separator's numeric error.
7
+
8
+ No model silently calls an estimate 'isolated instrument'.
9
+ """
10
+ from __future__ import annotations
11
+
12
+ from contextlib import ExitStack
13
+ from pathlib import Path
14
+ import math
15
+ import os
16
+ import shutil
17
+ import uuid
18
+
19
+ import numpy as np
20
+ import soundfile as sf
21
+
22
+
23
+ TRACK_COLORS = {
24
+ "harmonic": "#8ea8ff", "percussive": "#ffbb79",
25
+ "residual": "#b9c0ce", "bass": "#46d8cc", "drums": "#ffc070",
26
+ "vocals": "#d7a4ff", "other": "#8ec7ff",
27
+ "kick_core": "#fa9e61", "bassline_core": "#68f0d3",
28
+ "noise_difference": "#f38ca6", "source_body": "#aab7d5",
29
+ }
30
+
31
+
32
+ def _entry(path: Path, root: Path, label: str, kind: str, description: str) -> dict:
33
+ info = sf.info(str(path))
34
+ return {
35
+ "master": path.relative_to(root).as_posix(),
36
+ "label": label, "role": kind,
37
+ "description": description, "color": TRACK_COLORS.get(path.stem, "#b5c1d8"),
38
+ "samplerate": info.samplerate, "channels": info.channels, "frames": info.frames,
39
+ }
40
+
41
+
42
+ def build_unified_graph(project_folder: Path, manifest: dict, *, blocksize: int = 32768) -> dict | None:
43
+ """Materialize canonical mix stems, streaming bounded-memory float WAV.
44
+
45
+ Derivative lanes are only available when their producer has actually run.
46
+ If base separation is missing, the original file is used as one source.
47
+ """
48
+ asset = manifest.get("asset")
49
+ if not asset:
50
+ return None
51
+ bass = manifest.get("bass")
52
+ repair = (manifest.get("noise") or {}).get("repair")
53
+ if not bass and not repair:
54
+ return None
55
+
56
+ canonical = manifest.get("stems") or {}
57
+ base = {
58
+ name: project_folder / "stems" / item["master"]
59
+ for name, item in canonical.items()
60
+ } if canonical else {"source_body": project_folder / asset["stored_name"]}
61
+ root = project_folder.resolve()
62
+ original_path = project_folder / asset["stored_name"]
63
+ if not original_path.is_file():
64
+ raise FileNotFoundError("Original audio is missing.")
65
+ sources = dict(base)
66
+ additions = {}
67
+
68
+ if bass:
69
+ # These are RAW estimates. Never add the optimized full bass mix as a stem:
70
+ # it contains newly synthesized harmonics and overlaps the original.
71
+ bp = project_folder / "basscore" / bass["id"]
72
+ for name, key in (("kick_core", "kick_estimate"), ("bassline_core", "bassline_estimate")):
73
+ sources[name] = bp / bass["stems"][key]["master"]
74
+ additions[name] = True
75
+ if repair:
76
+ rp = project_folder / "noisecore" / repair["id"]
77
+ sources["noise_difference"] = rp / repair["tracks"]["difference"]["master"]
78
+ additions["noise_difference"] = True
79
+
80
+ info = sf.info(str(original_path))
81
+ if not (info.channels in (1, 2) and info.frames > 0):
82
+ raise ValueError("Unified graph supports valid mono/stereo PCM inputs.")
83
+ for k, path in sources.items():
84
+ other = sf.info(str(path))
85
+ if (other.frames, other.channels, other.samplerate) != (info.frames, info.channels, info.samplerate):
86
+ raise ValueError(f"Invalid length/channel/sample-rate for {k}")
87
+ # When bass comes from the canonical separator, return the two estimated
88
+ # sub-bands to their respective parent stems.
89
+ bassmode = (bass or {}).get("mode")
90
+ if not bass:
91
+ dest = {}
92
+ elif bassmode == "hpss" and {"harmonic", "percussive"}.issubset(base):
93
+ dest = {"kick_core": "percussive", "bassline_core": "harmonic"}
94
+ elif bassmode == "demucs" and {"bass", "drums"}.issubset(base):
95
+ dest = {"kick_core": "drums", "bassline_core": "bass"}
96
+ else:
97
+ dest = {n: "source_body" if "source_body" in base else
98
+ ("harmonic" if "harmonic" in base else next(iter(base)))
99
+ for n in ("kick_core", "bassline_core")}
100
+ if repair:
101
+ dest["noise_difference"] = ("residual" if "residual" in base else
102
+ "other" if "other" in base else
103
+ "source_body" if "source_body" in base else next(iter(base)))
104
+ outfolder = project_folder / "unified" / uuid.uuid4().hex
105
+ outfolder.mkdir(parents=True, exist_ok=False)
106
+ labels = {
107
+ "harmonic": "Harmonische Basis · ohne Bassline",
108
+ "percussive": "Perkussive Basis · ohne Kick-Band",
109
+ "bass": "Bass-Basis · ohne Bassline-Band",
110
+ "drums": "Drums-Basis · ohne Kick-Band",
111
+ "other": "Andere Quellen",
112
+ "vocals": "Vocals",
113
+ "residual": "Residuum · differenzkompensiert",
114
+ "source_body": "Quellmix · verbleibender Körper",
115
+ "kick_core": "KICK · geschätztes Tiefband",
116
+ "bassline_core": "BASSLINE · geschätztes Tiefband",
117
+ "noise_difference": "STÖRSPUR · freigegebene Klick-Differenz",
118
+ }
119
+ roles = {n: "derived_bass" if n in ("kick_core", "bassline_core") else
120
+ "repair_difference" if n == "noise_difference" else "base_compensated"
121
+ for n in sources}
122
+ max_reconstruction_error = 0.
123
+ audio_peak = 0.
124
+ try:
125
+ with ExitStack() as stack:
126
+ readers = {name: stack.enter_context(sf.SoundFile(str(path), "r"))
127
+ for name, path in sources.items()}
128
+ original = stack.enter_context(sf.SoundFile(str(original_path), "r"))
129
+ writers = {
130
+ name: stack.enter_context(sf.SoundFile(
131
+ str(outfolder / f"{name}.wav"), "w", samplerate=info.samplerate,
132
+ channels=info.channels, subtype="FLOAT", format="WAV"
133
+ )) for name in sources
134
+ }
135
+ for start in range(0, info.frames, blocksize):
136
+ count = min(blocksize, info.frames-start)
137
+ chunks = {
138
+ name: rd.read(count, dtype="float64", always_2d=True)
139
+ for name, rd in readers.items()
140
+ }
141
+ for name in chunks:
142
+ if len(chunks[name]) != count or not np.isfinite(chunks[name]).all():
143
+ raise ValueError(f"Corrupt source chunk {name}")
144
+ for derivative, parent in dest.items():
145
+ chunks[parent] -= chunks[derivative]
146
+ # All recorded lanes are additive. No hidden gain normalization.
147
+ total = np.zeros((count, info.channels), dtype=np.float64)
148
+ for name, chunk in chunks.items():
149
+ writers[name].write(chunk)
150
+ total += chunk
151
+ source = original.read(count, dtype="float64", always_2d=True)
152
+ err = float(np.max(np.abs(source-total)))
153
+ max_reconstruction_error = max(max_reconstruction_error, err)
154
+ audio_peak = max(audio_peak, float(np.max(np.abs(source))))
155
+ # Browser-friendly listening previews, separate from unchanged Float32 masters.
156
+ stempeaks = {}
157
+ for name in sources:
158
+ with sf.SoundFile(str(outfolder / f"{name}.wav"), "r") as inp, \
159
+ sf.SoundFile(str(outfolder / f"{name}_preview.wav"), "w",
160
+ samplerate=info.samplerate, channels=info.channels,
161
+ subtype="PCM_16", format="WAV") as preview:
162
+ peak = 0.0
163
+ while True:
164
+ part = inp.read(blocksize, dtype="float64", always_2d=True)
165
+ if not len(part):
166
+ break
167
+ peak = max(peak, float(np.max(np.abs(part))))
168
+ preview.write(np.clip(part, -1, 1))
169
+ stempeaks[name] = peak
170
+ entries = {}
171
+ for name in sources:
172
+ meta = _entry(outfolder / f"{name}.wav", project_folder, labels.get(name, name),
173
+ roles[name],
174
+ "Estimated sub-band" if roles[name] == "derived_bass"
175
+ else "Approved noise difference; muting removes these changes" if name == "noise_difference"
176
+ else "Real source compensated to prevent double-counting.")
177
+ meta["color"] = TRACK_COLORS.get(name, "#9bb5cc")
178
+ # Conservative estimated visual-only peaks; signal is always verified by IO.
179
+ meta["duration_seconds"] = round(info.frames/info.samplerate, 4)
180
+ meta["peak_dbfs"] = round(20*math.log10(max(stempeaks[name],1e-12)),3)
181
+ meta["preview"] = (outfolder / f"{name}_preview.wav").relative_to(project_folder).as_posix()
182
+ entries[name] = meta
183
+ # Keep raw material and old unified exports until the graph was produced.
184
+ tol = max(5e-5, audio_peak * 2e-5)
185
+ graph = {
186
+ "stems": entries, "based_on_asset_sha256": asset["sha256"],
187
+ "mode": "additive_compensated_sources",
188
+ "mix_consistency_max_error": max_reconstruction_error,
189
+ "mix_consistency_tolerance": tol,
190
+ "mix_consistent": bool(max_reconstruction_error <= tol),
191
+ "bass_included": bool(bass), "noise_included": bool(repair),
192
+ "noise_default_mute": False,
193
+ "warnings": [
194
+ "Kick/Bassline are estimates, not perfectly isolated instruments.",
195
+ "NOISE is an approved repair difference, not generic ambient noise extraction.",
196
+ "NOISE is audible by default to preserve the original; explicitly muting it auditions the repaired mix.",
197
+ "Gain and effect changes can exceed 0 dBFS; review export peak."
198
+ ],
199
+ }
200
+ if not graph["mix_consistent"]:
201
+ raise ValueError(
202
+ f"Graph is not mixture-consistent: error={max_reconstruction_error}, tol={tol}"
203
+ )
204
+ return graph
205
+ except Exception:
206
+ shutil.rmtree(outfolder, ignore_errors=True)
207
+ raise
docs/RELEASE_0.7.md ADDED
@@ -0,0 +1,54 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ABSOLUTION Ω v0.7 — Unified Studio / Quality Gates
2
+
3
+ ## Fehlermodus der v0.6-Architektur
4
+
5
+ `BASS-CORE` lieferte `kick_estimate` und `bassline_estimate` neben dem Basis-Mix,
6
+ `NOISE-CORE` eine Reparatur-Differenz. Blindes Addieren dieser Ausgaben zu den
7
+ bereits separierten Spuren hätte den Bass und einzelne Störimpulse verdoppelt.
8
+
9
+ ## Neue additive, kompensierte Spur-Architektur
10
+
11
+ Für die Basisstems `S_i`, Bassanteile `K`, `B` und die manuell bestätigte
12
+ Reparatur-Differenz `N = x - clean` erstellt der Builder neue Stems `S'_i`,
13
+ indem er die abgeleiteten Anteile von den jeweiligen Elternstems subtrahiert.
14
+
15
+ `sum(S'_i) + K + B + N ~= x`.
16
+
17
+ Numerischer Gate: `max_abs_error <= max(5e-5, 2e-5 * peak(x))`. Ein nicht
18
+ bestandener Gate blockiert die Veröffentlichung des neuen Graphen.
19
+
20
+ Standardeinstellung für NOISE: **hörbar / nicht stumm**. Bei Stummschaltung:
21
+ `mix_without_noise = x - N = clean` (bei neutralen übrigen Mix-Reglern).
22
+
23
+ Bassline/Kick sind frequenzbasierte Schätzungen, nicht isolierte Instrumente.
24
+ Störspur basiert ausschließlich auf manueller Reparatur, nicht ungeprüfter KI.
25
+
26
+ ## API
27
+
28
+ - `POST /api/v1/jobs/bass` erzeugt Bass-Signalanteile und baut den kanonischen Graphen
29
+ - `POST /api/v1/jobs/noise/scan` scannt nur und aktualisiert den Graphen bei alter Reparatur
30
+ - `POST /api/v1/jobs/noise/repair` erstellt freigegebene Differenz und baut Graphen neu
31
+ - `POST /api/v1/jobs/unify?project_id=<UUID>` synchronisiert vorhandene Projektdateien
32
+ - `GET /api/v1/projects/<UUID>/unified` Graphstatus, reale Spuren
33
+ - `GET /api/v1/projects/<UUID>/wave/<stem>` samplebasierte Visualisierung
34
+ - `GET /api/v1/projects/<UUID>/audio/<stem>?preview=true` echte Spur
35
+ - `POST /api/v1/jobs/mix` Offline-Float-WAV inklusive aller echten Spuren
36
+
37
+ ## Tests
38
+
39
+ `python -m pytest -q`
40
+
41
+ Regressionen prüfen neutrale Mix-Rekonstruktion, stummgeschaltete Bass- und
42
+ Noise-Spuren, unveränderte Eingabedateien, Single-Source-Modus, echte API-Jobs
43
+ einschließlich HPSS → Bass → NOISE-Reparatur → Mix/Export.
44
+
45
+ ## UI
46
+
47
+ Originales AURORA-inspiriertes CSS-Konzept: Glas, kühles Monitoring-Cyan, warme
48
+ BASS-Akzente, rosenfarbene Störspur, Instrument-/Diagnose-Tabelle.
49
+ Keine fremden Layoutdateien, Bilder oder Fonts übernommen.
50
+ Referenz: https://huggingface.co/spaces/2Btainment/einswarm-ui
51
+
52
+ Limitierung: Browser Testumgebung blockierte HTTP-Navigation zum lokalen
53
+ Server; die Oberfläche wurde offline mit produktionsgleichen Scripts und CSS
54
+ gerendert, echte Backend-Integration über FastAPI-TestClient geprüft.
pyproject.toml CHANGED
@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
4
 
5
  [project]
6
  name = "absolution-omega"
7
- version = "0.6.0"
8
  description = "Local-first audio workbench with multi-stem arrangement, synthetic mini tag model, BASS-CORE and NOISE-CORE."
9
  requires-python = ">=3.11"
10
  dependencies = [
 
4
 
5
  [project]
6
  name = "absolution-omega"
7
+ version = "0.7.0"
8
  description = "Local-first audio workbench with multi-stem arrangement, synthetic mini tag model, BASS-CORE and NOISE-CORE."
9
  requires-python = ">=3.11"
10
  dependencies = [
tests/test_unified.py ADDED
@@ -0,0 +1,193 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ """v0.7 regressions: additional BASS/NOISE lanes must never double-count energy."""
2
+ import hashlib
3
+ import io
4
+ import time
5
+ from pathlib import Path
6
+
7
+ import numpy as np
8
+ import pytest
9
+ import soundfile as sf
10
+ from fastapi.testclient import TestClient
11
+
12
+ from app.unified import build_unified_graph
13
+ from app.mixer import render_mix
14
+ from app.server import create_app
15
+
16
+
17
+ def write(path, samples, sr=8000):
18
+ path.parent.mkdir(parents=True, exist_ok=True)
19
+ sf.write(path, samples, sr, subtype="FLOAT", format="WAV")
20
+ return path
21
+
22
+
23
+ def fixtures(tmp):
24
+ sr=8000; t=np.arange(sr*2)/sr
25
+ harm=.11*np.sin(2*np.pi*90*t)
26
+ perc=np.zeros_like(harm);perc[::401]=.22
27
+ residue=.008*np.cos(2*np.pi*1733*t)
28
+ noise=np.zeros_like(harm);noise[778]=.065
29
+ stereo=lambda x:np.column_stack((x,x*.92)).astype("float32")
30
+ a,b,r,n=map(stereo,(harm,perc,residue,noise))
31
+ original=a+b+r
32
+ root=tmp/"project";root.mkdir()
33
+ write(root/"original.wav",original)
34
+ for name,x in [("harmonic",a),("percussive",b),("residual",r)]:
35
+ write(root/"stems"/f"{name}.wav",x)
36
+ # Emulate already computed BASS raw estimates. Enhanced bass mixes are excluded.
37
+ kick=b*.35
38
+ line=a*.42
39
+ write(root/"basscore"/"testbass"/"kick_estimate.wav",kick)
40
+ write(root/"basscore"/"testbass"/"bassline_estimate.wav",line)
41
+ write(root/"noisecore"/"testnoise"/"difference.wav",n)
42
+ asset_hash=hashlib.sha256((root/"original.wav").read_bytes()).hexdigest()
43
+ m={"asset":{"stored_name":"original.wav","sha256":asset_hash},
44
+ "stems":{n:{"master":f"{n}.wav"} for n in ("harmonic","percussive","residual")},
45
+ "bass":{"id":"testbass","mode":"hpss","stems":{
46
+ "kick_estimate":{"master":"kick_estimate.wav"},
47
+ "bassline_estimate":{"master":"bassline_estimate.wav"}}},
48
+ "noise":{"repair":{"id":"testnoise","tracks":{"difference":{"master":"difference.wav"}}}}}
49
+ return root,m,original,kick,line,n,asset_hash
50
+
51
+
52
+ def summed(root, graph, ignore=()):
53
+ return np.sum([sf.read(root/v["master"],always_2d=True)[0]
54
+ for name,v in graph["stems"].items() if name not in ignore], axis=0)
55
+
56
+
57
+ def test_combined_bass_noise_graph_preserves_original_and_tracks(tmp_path):
58
+ root,m,x,k,b,n,sha=fixtures(tmp_path)
59
+ graph=build_unified_graph(root,m,blocksize=1027)
60
+ assert graph["mix_consistent"]
61
+ assert graph["bass_included"] and graph["noise_included"]
62
+ assert set(graph["stems"])=={"harmonic","percussive","residual",
63
+ "kick_core","bassline_core","noise_difference"}
64
+ assert graph["mix_consistency_max_error"]<5e-7
65
+ assert np.max(np.abs(summed(root,graph)-x))<5e-7
66
+ assert np.max(np.abs(summed(root,graph,{"noise_difference"})-(x-n)))<5e-7
67
+ assert np.max(np.abs(summed(root,graph,{"kick_core"})-(x-k)))<5e-7
68
+ assert hashlib.sha256((root/"original.wav").read_bytes()).hexdigest()==sha
69
+ assert graph["noise_default_mute"] is False
70
+ assert all("preview" in v and (root/v["preview"]).is_file() for v in graph["stems"].values())
71
+
72
+
73
+ def test_mix_render_no_double_bass_or_noise(tmp_path):
74
+ root,m,x,k,b,n,_=fixtures(tmp_path)
75
+ graph=build_unified_graph(root,m)
76
+ sources={name:root/entry["master"] for name,entry in graph["stems"].items()}
77
+ out=root/"neutral.wav"
78
+ render_mix(sources,{"tracks":{}},out)
79
+ mix,_=sf.read(out,always_2d=True)
80
+ assert np.max(np.abs(mix-x))<5e-7
81
+ render_mix(sources,{"tracks":{"noise_difference":{"mute":True}}},out)
82
+ after,_=sf.read(out,always_2d=True)
83
+ assert np.max(np.abs(after-(x-n)))<5e-7
84
+ render_mix(sources,{"tracks":{"bassline_core":{"mute":True},"kick_core":{"mute":True}}},out)
85
+ after,_=sf.read(out,always_2d=True)
86
+ assert np.max(np.abs(after-(x-k-b)))<5e-7
87
+
88
+
89
+ def test_graph_for_only_noise_and_none(tmp_path):
90
+ root,m,x,k,b,n,_=fixtures(tmp_path)
91
+ m["bass"]=None
92
+ graph=build_unified_graph(root,m)
93
+ assert set(graph["stems"])=={"harmonic","percussive","residual","noise_difference"}
94
+ assert np.max(np.abs(summed(root,graph)-x))<5e-7
95
+ m["noise"]=None
96
+ assert build_unified_graph(root,m) is None
97
+
98
+
99
+ def test_graph_for_bass_without_separation(tmp_path):
100
+ root,m,x,k,b,n,_=fixtures(tmp_path)
101
+ m["stems"]={}
102
+ m["noise"]=None
103
+ m["bass"]["mode"]="dsp"
104
+ graph=build_unified_graph(root,m)
105
+ assert set(graph["stems"])=={"source_body","kick_core","bassline_core"}
106
+ assert np.max(np.abs(summed(root,graph)-x))<5e-7
107
+
108
+
109
+ def poll(client,jid):
110
+ for _ in range(250):
111
+ result=client.get("/api/v1/jobs/"+jid).json()
112
+ if result["status"] in ("done","failed"):return result
113
+ time.sleep(.04)
114
+ pytest.fail("Job timed out")
115
+
116
+
117
+ def test_unified_api_full_flow_bass_and_mixer(tmp_path):
118
+ root,m,x,k,b,n,_=fixtures(tmp_path)
119
+ # Real API path, real BASS-CORE algorithm. No model checkpoint required.
120
+ app=create_app(tmp_path/"runtime")
121
+ payload=io.BytesIO()
122
+ sf.write(payload,x,8000,format="WAV",subtype="FLOAT")
123
+ with TestClient(app) as client:
124
+ pid=client.post("/api/v1/projects",json={"name":"Unified"}).json()["id"]
125
+ response=client.post("/api/v1/assets",data={"project_id":pid},
126
+ files={"file":("demo.wav",payload.getvalue(),"audio/wav")})
127
+ assert response.status_code==201,response.text
128
+ bassjob=client.post("/api/v1/jobs/bass",json={
129
+ "project_id":pid, "source_mode":"dsp", "crossover_hz":150,
130
+ "harmonics":0, "wet":0}).json()
131
+ result=poll(client,bassjob["id"])
132
+ assert result["status"]=="done",result
133
+ manifest=client.get(f"/api/v1/projects/{pid}").json()
134
+ assert manifest["unified"]["mix_consistent"]
135
+ assert set(manifest["unified"]["stems"])=={"source_body","kick_core","bassline_core"}
136
+ for track in manifest["unified"]["stems"]:
137
+ assert client.get(f"/api/v1/projects/{pid}/wave/{track}").status_code==200
138
+ assert client.get(f"/api/v1/projects/{pid}/audio/{track}?preview=true").status_code==200
139
+ assert client.get(f"/api/v1/projects/{pid}/export").status_code==200
140
+ job=client.post("/api/v1/jobs/mix",json={"project_id":pid}).json()
141
+ assert poll(client,job["id"])["status"]=="done"
142
+ rendered=client.get(f"/api/v1/projects/{pid}/mixer/audio")
143
+ assert rendered.status_code==200
144
+ mixed,_=sf.read(io.BytesIO(rendered.content),always_2d=True)
145
+ assert np.max(np.abs(mixed-x))<2e-5
146
+
147
+
148
+ def test_full_api_hpss_bass_noise_repaired_additive(tmp_path):
149
+ """Actual HPSS + BASS-CORE + manually approved NOISE-CORE on same source."""
150
+ t=np.arange(24000)/8000
151
+ wave=.09*np.sin(2*np.pi*73*t)+.033*np.sin(2*np.pi*512*t)
152
+ stereo=np.stack((wave,wave*.85),axis=1).astype("float32")
153
+ stereo[5017,:]+=.82
154
+ bio=io.BytesIO();sf.write(bio,stereo,8000,format="WAV",subtype="FLOAT")
155
+ app=create_app(tmp_path/"complete")
156
+ with TestClient(app) as client:
157
+ pid=client.post("/api/v1/projects",json={"name":"Full"}).json()["id"]
158
+ assert client.post("/api/v1/assets",data={"project_id":pid},
159
+ files={"file":("full.wav",bio.getvalue(),"audio/wav")}).status_code==201
160
+ run=client.post("/api/v1/jobs/separate",json={"project_id":pid,"model":"hpss"})
161
+ assert run.status_code==202,run.text
162
+ assert poll(client,run.json()["id"])["status"]=="done"
163
+ run=client.post("/api/v1/jobs/bass",json={
164
+ "project_id":pid,"source_mode":"auto", "crossover_hz":140,
165
+ "harmonics":0,"wet":0})
166
+ assert poll(client,run.json()["id"])["status"]=="done"
167
+ run=client.post("/api/v1/jobs/noise/scan",json={"project_id":pid})
168
+ assert poll(client,run.json()["id"])["status"]=="done"
169
+ scan=client.get(f"/api/v1/projects/{pid}/noise").json()["noise"]["scan"]
170
+ events=[z["sample"] for z in scan["events"]]
171
+ assert 5017 in events, events
172
+ run=client.post("/api/v1/jobs/noise/repair",json={
173
+ "project_id":pid,"approved_samples":[5017], "strength":.8})
174
+ assert poll(client,run.json()["id"])["status"]=="done"
175
+ manifest=client.get(f"/api/v1/projects/{pid}").json()
176
+ graph=manifest["unified"]
177
+ assert graph and graph["mix_consistent"],graph
178
+ assert len(graph["stems"])==6
179
+ run=client.post("/api/v1/jobs/mix",json={"project_id":pid})
180
+ assert poll(client,run.json()["id"])["status"]=="done"
181
+ wav=client.get(f"/api/v1/projects/{pid}/mixer/audio")
182
+ assert wav.status_code==200
183
+ reconstructed,_=sf.read(io.BytesIO(wav.content),always_2d=True)
184
+ assert np.max(np.abs(stereo-reconstructed))<5e-5
185
+ run=client.post("/api/v1/jobs/mix",json={
186
+ "project_id":pid,"tracks":{"noise_difference":{"mute":True}}})
187
+ assert poll(client,run.json()["id"])["status"]=="done"
188
+ clean,_=sf.read(io.BytesIO(client.get(
189
+ f"/api/v1/projects/{pid}/mixer/audio").content),always_2d=True)
190
+ difference=reconstructed-clean
191
+ assert np.max(np.abs(difference))>.1
192
+ mask=np.ones(len(difference),dtype=bool);mask[5017]=False
193
+ assert np.max(np.abs(difference[mask]))<2e-5