Fabio
c36510c21e
Some checks failed
Linux Build / Linux cpu (push) Has been cancelled
Linux Build / Linux cpu-clang (push) Has been cancelled
Linux Build / Linux vulkan (push) Has been cancelled
mac-build / macOS CPU (push) Has been cancelled
Nix Build / Nix (cpu) (push) Has been cancelled
Nix Build / Nix (vulkan) (push) Has been cancelled
windows-build / Windows CPU (push) Has been cancelled
224 lines
6.5 KiB
JSON
224 lines
6.5 KiB
JSON
{
|
|
"schema_version": 1,
|
|
"family": "voxtral_tts",
|
|
"display_name": "Voxtral 4B TTS",
|
|
"description": "Mistral's Ministral-based autoregressive and flow-matching TTS model with 20 embedded preset voices.",
|
|
"category": "tts",
|
|
"status": "supported",
|
|
"tasks": [
|
|
"tts"
|
|
],
|
|
"modes": [
|
|
"offline",
|
|
"streaming"
|
|
],
|
|
"languages": [
|
|
"en",
|
|
"fr",
|
|
"es",
|
|
"de",
|
|
"it",
|
|
"pt",
|
|
"nl",
|
|
"ar",
|
|
"hi"
|
|
],
|
|
"runtime": {
|
|
"tags": [
|
|
"gguf"
|
|
]
|
|
},
|
|
"capabilities": {
|
|
"tts": [
|
|
"built_in_voices",
|
|
"long_form"
|
|
]
|
|
},
|
|
"options": {
|
|
"request": [
|
|
{
|
|
"name": "voice_id",
|
|
"type": "enum",
|
|
"description": "Embedded preset voice; the released checkpoint does not contain the encoder needed for reference-audio cloning.",
|
|
"values": [
|
|
"casual_female",
|
|
"casual_male",
|
|
"cheerful_female",
|
|
"neutral_female",
|
|
"neutral_male",
|
|
"pt_male",
|
|
"pt_female",
|
|
"nl_male",
|
|
"nl_female",
|
|
"it_male",
|
|
"it_female",
|
|
"fr_male",
|
|
"fr_female",
|
|
"es_male",
|
|
"es_female",
|
|
"de_male",
|
|
"de_female",
|
|
"ar_male",
|
|
"hi_male",
|
|
"hi_female"
|
|
],
|
|
"required": false,
|
|
"default": "casual_male"
|
|
},
|
|
{
|
|
"name": "max_tokens",
|
|
"type": "int",
|
|
"description": "Maximum generated audio frames per text chunk; each frame represents 80 ms.",
|
|
"required": false,
|
|
"min": 1,
|
|
"default": 2048
|
|
},
|
|
{
|
|
"name": "num_inference_steps",
|
|
"type": "int",
|
|
"description": "Euler integration steps for the per-frame acoustic flow model.",
|
|
"required": false,
|
|
"min": 1,
|
|
"default": 7
|
|
},
|
|
{
|
|
"name": "guidance_scale",
|
|
"type": "float",
|
|
"description": "Acoustic classifier-free guidance scale; the AR backbone does not use CFG.",
|
|
"required": false,
|
|
"min": 0,
|
|
"default": 1.2
|
|
},
|
|
{
|
|
"name": "seed",
|
|
"type": "int",
|
|
"description": "Acoustic noise seed; omitted requests select a random seed.",
|
|
"required": false,
|
|
"min": 0
|
|
},
|
|
{
|
|
"name": "text_chunk_size",
|
|
"type": "int",
|
|
"description": "Maximum Unicode codepoints per framework long-form text chunk.",
|
|
"required": false,
|
|
"min": 1,
|
|
"default": 300
|
|
},
|
|
{
|
|
"name": "text_chunk_mode",
|
|
"type": "enum",
|
|
"description": "Framework text chunking mode.",
|
|
"values": [
|
|
"default",
|
|
"tag_aware",
|
|
"japanese",
|
|
"endline"
|
|
],
|
|
"required": false,
|
|
"default": "default"
|
|
}
|
|
],
|
|
"session": [
|
|
{
|
|
"name": "weight_type",
|
|
"type": "enum",
|
|
"description": "Weight storage type; native preserves the package's tensor types.",
|
|
"values": [
|
|
"native",
|
|
"f32",
|
|
"f16",
|
|
"bf16",
|
|
"q4_0",
|
|
"q4_1",
|
|
"q5_0",
|
|
"q5_1",
|
|
"q2_k",
|
|
"q3_k",
|
|
"q4_k",
|
|
"q5_k",
|
|
"q6_k",
|
|
"q8_0",
|
|
"nvfp4"
|
|
],
|
|
"required": false,
|
|
"default": "native"
|
|
},
|
|
{
|
|
"name": "kv_cache_type",
|
|
"type": "enum",
|
|
"description": "AR key/value cache storage type.",
|
|
"values": [
|
|
"f32",
|
|
"f16",
|
|
"bf16"
|
|
],
|
|
"required": false,
|
|
"default": "f16"
|
|
}
|
|
],
|
|
"load": []
|
|
},
|
|
"package_defaults": {
|
|
"download": {
|
|
"kind": "huggingface_snapshot",
|
|
"repo": "audio-cpp/audio.cpp-gguf",
|
|
"revision": "main",
|
|
"gated": false
|
|
}
|
|
},
|
|
"packages": [
|
|
{
|
|
"id": "voxtral_4b_tts_2603_bf16",
|
|
"display_name": "Voxtral 4B TTS 2603 BF16 GGUF",
|
|
"default": true,
|
|
"format": "gguf",
|
|
"precision": "bf16",
|
|
"target_directory": "Voxtral-4B-TTS-2603-GGUF",
|
|
"files": [
|
|
"Voxtral-4B-TTS-2603-GGUF/voxtral-4b-tts-2603-bf16.gguf"
|
|
],
|
|
"strip_prefix": "Voxtral-4B-TTS-2603-GGUF"
|
|
},
|
|
{
|
|
"id": "voxtral_4b_tts_2603_q8_0",
|
|
"display_name": "Voxtral 4B TTS 2603 Q8_0 GGUF",
|
|
"default": false,
|
|
"format": "gguf",
|
|
"precision": "q8_0",
|
|
"target_directory": "Voxtral-4B-TTS-2603-GGUF",
|
|
"files": [
|
|
"Voxtral-4B-TTS-2603-GGUF/voxtral-4b-tts-2603-q8_0.gguf"
|
|
],
|
|
"strip_prefix": "Voxtral-4B-TTS-2603-GGUF"
|
|
}
|
|
],
|
|
"dependencies": [],
|
|
"ui": {
|
|
"recommended_package": "voxtral_4b_tts_2603_bf16",
|
|
"tags": [
|
|
"TTS",
|
|
"GGUF"
|
|
],
|
|
"docs": [
|
|
"docs/models/voxtral_tts.md",
|
|
"docs/gguf.md"
|
|
]
|
|
},
|
|
"sources": [
|
|
{
|
|
"format": "gguf",
|
|
"roots": {
|
|
"model": ".",
|
|
"weights": "$gguf"
|
|
},
|
|
"files": {
|
|
"config": "model:params.json",
|
|
"tokenizer_json": "model:tokenizer.json",
|
|
"tokenizer_config": "model:tokenizer_config.json"
|
|
},
|
|
"tensors": {
|
|
"model_weights": "weights:"
|
|
}
|
|
}
|
|
]
|
|
}
|