Create config.json
Browse files- config.json +38 -0
config.json
ADDED
|
@@ -0,0 +1,38 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_name": "TruthShield VoiceGen",
|
| 3 |
+
"model_type": "vits",
|
| 4 |
+
"multilingual": true,
|
| 5 |
+
"supported_languages": [
|
| 6 |
+
"bhojpuri",
|
| 7 |
+
"bengali",
|
| 8 |
+
"english",
|
| 9 |
+
"gujarati",
|
| 10 |
+
"hindi",
|
| 11 |
+
"chhattisgarhi",
|
| 12 |
+
"kannada",
|
| 13 |
+
"magahi",
|
| 14 |
+
"maithili",
|
| 15 |
+
"marathi",
|
| 16 |
+
"telugu"
|
| 17 |
+
],
|
| 18 |
+
"speaker_embedding_model": "ECAPA-TDNN",
|
| 19 |
+
"vocoder": "HiFiGAN",
|
| 20 |
+
"sampling_rate": 22050,
|
| 21 |
+
"description": "Configuration reference for the TruthShield VoiceGen multilingual, multi-speaker TTS system. Fine-tuned weights are stored externally for evaluation as required by the Voice Tech For All Challenge.",
|
| 22 |
+
"architecture": {
|
| 23 |
+
"text_encoder": "VITS Text Encoder",
|
| 24 |
+
"duration_predictor": "VITS Duration Module",
|
| 25 |
+
"decoder": "Flow-based VITS Decoder",
|
| 26 |
+
"prosody_features": ["pitch", "energy"],
|
| 27 |
+
"speaker_embedding_dim": 192
|
| 28 |
+
},
|
| 29 |
+
"safety_layer": {
|
| 30 |
+
"enabled": true,
|
| 31 |
+
"similarity_model": "ECAPA-TDNN",
|
| 32 |
+
"outputs": ["similarity_score", "safety_verified"]
|
| 33 |
+
},
|
| 34 |
+
"api_spec_compliance": {
|
| 35 |
+
"endpoint": "GET /Get_Inference",
|
| 36 |
+
"mandatory_params": ["text", "lang", "speaker_wav"]
|
| 37 |
+
}
|
| 38 |
+
}
|