Update app.py
Browse files
app.py
CHANGED
|
@@ -3,6 +3,7 @@ import tempfile
|
|
| 3 |
from fastapi import FastAPI, WebSocket, WebSocketDisconnect
|
| 4 |
from fastapi.middleware.cors import CORSMiddleware
|
| 5 |
from faster_whisper import WhisperModel
|
|
|
|
| 6 |
from pipeline.postprocess import process
|
| 7 |
|
| 8 |
app = FastAPI()
|
|
@@ -19,6 +20,10 @@ print("Loading Whisper model...", flush=True)
|
|
| 19 |
model = WhisperModel("large-v3", device="cpu", compute_type="int8")
|
| 20 |
print("Whisper model ready!")
|
| 21 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 22 |
@app.get("/health")
|
| 23 |
async def health():
|
| 24 |
return {"status": "ok", "message": "SubFlow backend is running"}
|
|
@@ -27,6 +32,7 @@ async def health():
|
|
| 27 |
async def transcribe(websocket: WebSocket):
|
| 28 |
await websocket.accept()
|
| 29 |
tmp_path = None
|
|
|
|
| 30 |
|
| 31 |
try:
|
| 32 |
print("Step 1: waiting for bytes")
|
|
@@ -39,11 +45,20 @@ async def transcribe(websocket: WebSocket):
|
|
| 39 |
tmp_path = tmp.name
|
| 40 |
print(f"Step 3: saved to {tmp_path}")
|
| 41 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
await websocket.send_json({"status": "transcribing"})
|
| 43 |
print("Step 4: starting transcription")
|
| 44 |
|
| 45 |
segments_raw, info = model.transcribe(
|
| 46 |
-
|
| 47 |
language="en",
|
| 48 |
word_timestamps=True,
|
| 49 |
vad_filter=True,
|
|
@@ -97,6 +112,8 @@ async def transcribe(websocket: WebSocket):
|
|
| 97 |
finally:
|
| 98 |
if tmp_path and os.path.exists(tmp_path):
|
| 99 |
os.remove(tmp_path)
|
|
|
|
|
|
|
| 100 |
try:
|
| 101 |
await websocket.close()
|
| 102 |
except Exception:
|
|
|
|
| 3 |
from fastapi import FastAPI, WebSocket, WebSocketDisconnect
|
| 4 |
from fastapi.middleware.cors import CORSMiddleware
|
| 5 |
from faster_whisper import WhisperModel
|
| 6 |
+
from df.enhance import enhance, init_df, load_audio, save_audio
|
| 7 |
from pipeline.postprocess import process
|
| 8 |
|
| 9 |
app = FastAPI()
|
|
|
|
| 20 |
model = WhisperModel("large-v3", device="cpu", compute_type="int8")
|
| 21 |
print("Whisper model ready!")
|
| 22 |
|
| 23 |
+
print("Loading DeepFilterNet model...", flush=True)
|
| 24 |
+
df_model, df_state, _ = init_df()
|
| 25 |
+
print("DeepFilterNet model ready!")
|
| 26 |
+
|
| 27 |
@app.get("/health")
|
| 28 |
async def health():
|
| 29 |
return {"status": "ok", "message": "SubFlow backend is running"}
|
|
|
|
| 32 |
async def transcribe(websocket: WebSocket):
|
| 33 |
await websocket.accept()
|
| 34 |
tmp_path = None
|
| 35 |
+
enhanced_path = None
|
| 36 |
|
| 37 |
try:
|
| 38 |
print("Step 1: waiting for bytes")
|
|
|
|
| 45 |
tmp_path = tmp.name
|
| 46 |
print(f"Step 3: saved to {tmp_path}")
|
| 47 |
|
| 48 |
+
await websocket.send_json({"status": "cleaning"})
|
| 49 |
+
print("Step 3.5: cleaning audio with DeepFilterNet")
|
| 50 |
+
|
| 51 |
+
audio_data, _ = load_audio(tmp_path, sr=df_state.sr())
|
| 52 |
+
enhanced_audio = enhance(df_model, df_state, audio_data)
|
| 53 |
+
enhanced_path = tmp_path.replace(".wav", "_enhanced.wav")
|
| 54 |
+
save_audio(enhanced_path, enhanced_audio, df_state.sr())
|
| 55 |
+
print(f"Step 3.6: cleaned audio saved to {enhanced_path}")
|
| 56 |
+
|
| 57 |
await websocket.send_json({"status": "transcribing"})
|
| 58 |
print("Step 4: starting transcription")
|
| 59 |
|
| 60 |
segments_raw, info = model.transcribe(
|
| 61 |
+
enhanced_path,
|
| 62 |
language="en",
|
| 63 |
word_timestamps=True,
|
| 64 |
vad_filter=True,
|
|
|
|
| 112 |
finally:
|
| 113 |
if tmp_path and os.path.exists(tmp_path):
|
| 114 |
os.remove(tmp_path)
|
| 115 |
+
if enhanced_path and os.path.exists(enhanced_path):
|
| 116 |
+
os.remove(enhanced_path)
|
| 117 |
try:
|
| 118 |
await websocket.close()
|
| 119 |
except Exception:
|