ONNX: every voice × style baked as speaker rows; default voice trio (Faszczewska+Masiak+Bielenia); structured voices.json

README.md

@@ -36,36 +36,42 @@ git clone https://github.com/machinekind/tts-pl-playground.git && cd tts-pl-play
 python -c "from huggingface_hub import snapshot_download as d; d('machinekind/Matcha-TTS-PL', local_dir='.')"
 python playground.py --port 8771            # UI: voices, blends, style tokens, effects, sequence composer, video visualizer
 python synth_samples.py --ckpt model/matcha_pl_target.ckpt --vocoder vocoder/hifigan_pl.pt \
-    --sentences my_sentences.txt --voice "0.5*6+0.5*3" --steps 4 --temperature 0.5 --out out/
+    --sentences my_sentences.txt --voice "0.3333*1+0.3333*2+0.3334*0" --steps 4 --temperature 0.5 --out out/
 ```
-ONNX Runtime (CPU or GPU), no PyTorch needed at run time:
+ONNX Runtime (CPU or GPU), no PyTorch needed at run time (`spks` 20 = the default `trio` voice, see *Voices*):
 ```python
 import onnxruntime as ort, numpy as np
 sess = ort.InferenceSession("onnx/matcha_pl_voices_t4.onnx")
 x = phonemes  # int64 [1, T]: espeak-ng "pl" ids from the playground's polish_cleaners (see RECIPE.md §1)
-wav, n = sess.run(None, {"x": x, "x_lengths": np.array([x.shape[1]]), "scales": np.array([0.5, 0.95], np.float32), "spks": np.array([24])})
+wav, n = sess.run(None, {"x": x, "x_lengths": np.array([x.shape[1]]), "scales": np.array([0.5, 0.95], np.float32), "spks": np.array([20])})
 ```
 
 ## Voices
 
 Speaker ids 0–7 are the target readers (0 Bartosz Bielenia, 1 Katarzyna Faszczewska, 2 Wojciech Masiak, 3 Bartosz Głogowski, 4 Jan Staszczyk, 5 Marek Proszek, 6 Piotr Kopa, 7 Radosław Krzyżowski); ids 8–19 are base-training speakers kept for completeness.
-Blend voices by averaging embedding rows (`--voice "0.5*6+0.5*3"`). The ONNX graphs have these blends baked in as
-extra speaker ids:
-
-| id | name | blend (weight × speaker id) |
-|---|---|---|
-| 20 | `kopa` | 6 |
-| 21 | `glogowski` | 3 |
-| 22 | `faszczewska` | 1 |
-| 23 | `bielenia` | 0 |
-| 24 | `mix3` | 0.5*6+0.3*3+0.2*1 |
-| 25 | `kopa_glogowski` | 0.5*6+0.5*3 |
-
-`mix3` (0.5 Kopa + 0.3 Głogowski + 0.2 Faszczewska) and `kopa_glogowski` are the recommended robot voices.
+Blend voices by averaging embedding rows (`--voice "0.33*1+0.33*2+0.34*0"`). The ONNX graphs cannot take a style input,
+so every voice × style combination is baked in as an extra speaker row; `onnx/voices.json` maps
+`voices[key].id` (no style) and `voices[key].styles[style_id]` to the row to pass as `spks`:
+
+| key | voice | blend (weight × speaker id) | row, no style | rows with style 0–17 |
+|---|---|---|---|---|
+| `trio` | Faszczewska + Masiak + Bielenia (⅓ each) · default | 0.3333*1+0.3333*2+0.3334*0 | 20 | 21–38 |
+| `mix3` | mix3 · Kopa 0.5 + Głogowski 0.3 + Faszczewska 0.2 | 0.5*6+0.3*3+0.2*1 | 39 | 40–57 |
+| `kopa_glogowski` | Kopa 0.5 + Głogowski 0.5 | 0.5*6+0.5*3 | 58 | 59–76 |
+| `reader_0` | Bartosz Bielenia | 0 | 0 | 77–94 |
+| `reader_1` | Katarzyna Faszczewska | 1 | 1 | 95–112 |
+| `reader_2` | Wojciech Masiak | 2 | 2 | 113–130 |
+| `reader_3` | Bartosz Głogowski | 3 | 3 | 131–148 |
+| `reader_4` | Jan Staszczyk | 4 | 4 | 149–166 |
+| `reader_5` | Marek Proszek | 5 | 5 | 167–184 |
+| `reader_6` | Piotr Kopa | 6 | 6 | 185–202 |
+| `reader_7` | Radosław Krzyżowski | 7 | 7 | 203–220 |
+
+The default voice is `trio`, an equal blend of Katarzyna Faszczewska, Wojciech Masiak and Bartosz Bielenia.
 
 ## Style tokens
 
-An extra embedding added to the speaker embedding (pass `styles=<id>` to `synthesise`, or bake a row for ONNX).
+An extra embedding added to the speaker embedding: pass `styles=<id>` to `synthesise` in PyTorch; in ONNX pick the baked row from `onnx/voices.json`.
 Neutral = `7`. Pitch-range tokens move the spread by about ±1.3 semitones; the question tokens give a rising
 terminal contour for yes/no questions (Polish wh-questions fall — write real punctuation, "?" drives intonation).
 

onnx/matcha_pl_voices_t2.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:592c27f56bb1f0a4d887c1c0d28073e61da0f6296c335a2b385f1a7d785fb8db
-size 140056083
+oid sha256:8bcb52a6c3a8fb0ac9b7a1e091c28b303af56b13d8686ea671574f0d1c12bb1d
+size 140106006

onnx/matcha_pl_voices_t4.onnx

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:bfe75b72a4cf7c492c06c2742b36e383d90b03c5edb7092805662ac9e47be3a8
-size 140558613
+oid sha256:e90fb19c21ec51299e20001d9d53f02319e333484a6d9bc6e9a5869fa3774c8f
+size 140608536

onnx/voices.json

@@ -1,32 +1,301 @@
 {
- "kopa": {
-  "id": 20,
-  "spec": "6",
-  "style": null
+ "default": "trio",
+ "styles": {
+  "0": "flat-range · slow",
+  "1": "flat-range · slow · question",
+  "2": "flat-range · normal",
+  "3": "flat-range · normal · question",
+  "4": "flat-range · fast",
+  "5": "flat-range · fast · question",
+  "6": "mid-range · slow",
+  "7": "mid-range · slow · question",
+  "8": "mid-range · normal",
+  "9": "mid-range · normal · question",
+  "10": "mid-range · fast",
+  "11": "mid-range · fast · question",
+  "12": "wide-range · slow",
+  "13": "wide-range · slow · question",
+  "14": "wide-range · normal",
+  "15": "wide-range · normal · question",
+  "16": "wide-range · fast",
+  "17": "wide-range · fast · question"
  },
- "glogowski": {
-  "id": 21,
-  "spec": "3",
-  "style": null
+ "voices": {
+  "trio": {
+   "label": "Faszczewska + Masiak + Bielenia (⅓ each) · default",
+   "blend": "0.3333*1+0.3333*2+0.3334*0",
+   "id": 20,
+   "styles": {
+    "0": 21,
+    "1": 22,
+    "2": 23,
+    "3": 24,
+    "4": 25,
+    "5": 26,
+    "6": 27,
+    "7": 28,
+    "8": 29,
+    "9": 30,
+    "10": 31,
+    "11": 32,
+    "12": 33,
+    "13": 34,
+    "14": 35,
+    "15": 36,
+    "16": 37,
+    "17": 38
+   }
+  },
+  "mix3": {
+   "label": "mix3 · Kopa 0.5 + Głogowski 0.3 + Faszczewska 0.2",
+   "blend": "0.5*6+0.3*3+0.2*1",
+   "id": 39,
+   "styles": {
+    "0": 40,
+    "1": 41,
+    "2": 42,
+    "3": 43,
+    "4": 44,
+    "5": 45,
+    "6": 46,
+    "7": 47,
+    "8": 48,
+    "9": 49,
+    "10": 50,
+    "11": 51,
+    "12": 52,
+    "13": 53,
+    "14": 54,
+    "15": 55,
+    "16": 56,
+    "17": 57
+   }
+  },
+  "kopa_glogowski": {
+   "label": "Kopa 0.5 + Głogowski 0.5",
+   "blend": "0.5*6+0.5*3",
+   "id": 58,
+   "styles": {
+    "0": 59,
+    "1": 60,
+    "2": 61,
+    "3": 62,
+    "4": 63,
+    "5": 64,
+    "6": 65,
+    "7": 66,
+    "8": 67,
+    "9": 68,
+    "10": 69,
+    "11": 70,
+    "12": 71,
+    "13": 72,
+    "14": 73,
+    "15": 74,
+    "16": 75,
+    "17": 76
+   }
+  },
+  "reader_0": {
+   "label": "Bartosz Bielenia",
+   "blend": "0",
+   "id": 0,
+   "styles": {
+    "0": 77,
+    "1": 78,
+    "2": 79,
+    "3": 80,
+    "4": 81,
+    "5": 82,
+    "6": 83,
+    "7": 84,
+    "8": 85,
+    "9": 86,
+    "10": 87,
+    "11": 88,
+    "12": 89,
+    "13": 90,
+    "14": 91,
+    "15": 92,
+    "16": 93,
+    "17": 94
+   }
+  },
+  "reader_1": {
+   "label": "Katarzyna Faszczewska",
+   "blend": "1",
+   "id": 1,
+   "styles": {
+    "0": 95,
+    "1": 96,
+    "2": 97,
+    "3": 98,
+    "4": 99,
+    "5": 100,
+    "6": 101,
+    "7": 102,
+    "8": 103,
+    "9": 104,
+    "10": 105,
+    "11": 106,
+    "12": 107,
+    "13": 108,
+    "14": 109,
+    "15": 110,
+    "16": 111,
+    "17": 112
+   }
+  },
+  "reader_2": {
+   "label": "Wojciech Masiak",
+   "blend": "2",
+   "id": 2,
+   "styles": {
+    "0": 113,
+    "1": 114,
+    "2": 115,
+    "3": 116,
+    "4": 117,
+    "5": 118,
+    "6": 119,
+    "7": 120,
+    "8": 121,
+    "9": 122,
+    "10": 123,
+    "11": 124,
+    "12": 125,
+    "13": 126,
+    "14": 127,
+    "15": 128,
+    "16": 129,
+    "17": 130
+   }
+  },
+  "reader_3": {
+   "label": "Bartosz Głogowski",
+   "blend": "3",
+   "id": 3,
+   "styles": {
+    "0": 131,
+    "1": 132,
+    "2": 133,
+    "3": 134,
+    "4": 135,
+    "5": 136,
+    "6": 137,
+    "7": 138,
+    "8": 139,
+    "9": 140,
+    "10": 141,
+    "11": 142,
+    "12": 143,
+    "13": 144,
+    "14": 145,
+    "15": 146,
+    "16": 147,
+    "17": 148
+   }
+  },
+  "reader_4": {
+   "label": "Jan Staszczyk",
+   "blend": "4",
+   "id": 4,
+   "styles": {
+    "0": 149,
+    "1": 150,
+    "2": 151,
+    "3": 152,
+    "4": 153,
+    "5": 154,
+    "6": 155,
+    "7": 156,
+    "8": 157,
+    "9": 158,
+    "10": 159,
+    "11": 160,
+    "12": 161,
+    "13": 162,
+    "14": 163,
+    "15": 164,
+    "16": 165,
+    "17": 166
+   }
+  },
+  "reader_5": {
+   "label": "Marek Proszek",
+   "blend": "5",
+   "id": 5,
+   "styles": {
+    "0": 167,
+    "1": 168,
+    "2": 169,
+    "3": 170,
+    "4": 171,
+    "5": 172,
+    "6": 173,
+    "7": 174,
+    "8": 175,
+    "9": 176,
+    "10": 177,
+    "11": 178,
+    "12": 179,
+    "13": 180,
+    "14": 181,
+    "15": 182,
+    "16": 183,
+    "17": 184
+   }
+  },
+  "reader_6": {
+   "label": "Piotr Kopa",
+   "blend": "6",
+   "id": 6,
+   "styles": {
+    "0": 185,
+    "1": 186,
+    "2": 187,
+    "3": 188,
+    "4": 189,
+    "5": 190,
+    "6": 191,
+    "7": 192,
+    "8": 193,
+    "9": 194,
+    "10": 195,
+    "11": 196,
+    "12": 197,
+    "13": 198,
+    "14": 199,
+    "15": 200,
+    "16": 201,
+    "17": 202
+   }
+  },
+  "reader_7": {
+   "label": "Radosław Krzyżowski",
+   "blend": "7",
+   "id": 7,
+   "styles": {
+    "0": 203,
+    "1": 204,
+    "2": 205,
+    "3": 206,
+    "4": 207,
+    "5": 208,
+    "6": 209,
+    "7": 210,
+    "8": 211,
+    "9": 212,
+    "10": 213,
+    "11": 214,
+    "12": 215,
+    "13": 216,
+    "14": 217,
+    "15": 218,
+    "16": 219,
+    "17": 220
+   }
+  }
  },
- "faszczewska": {
-  "id": 22,
-  "spec": "1",
-  "style": null
- },
- "bielenia": {
-  "id": 23,
-  "spec": "0",
-  "style": null
- },
- "mix3": {
-  "id": 24,
-  "spec": "0.5*6+0.3*3+0.2*1",
-  "style": null
- },
- "kopa_glogowski": {
-  "id": 25,
-  "spec": "0.5*6+0.5*3",
-  "style": null
- }
+ "note": "spks row = voices[key].id (no style) or voices[key].styles[style_id]; rows are speaker embedding + style embedding, baked before export"
 }