[Model] Add Voxtral TTS model (#1803)

Signed-off-by: Chen-Yo Sun <chenyo.sun@mistral.ai>

docs/mkdocs/hooks/generate_api_readme.py

@@ -141,6 +141,7 @@ def scan_package(package_name: str = "vllm_omni") -> dict[str, list[str]]:
                 "vllm_omni.diffusion.models.qwen_image",
                 "vllm_omni.entrypoints.async_diffusion",
                 "vllm_omni.entrypoints.openai",
+                "vllm_omni.model_executor.models.voxtral_tts.configuration_voxtral_tts",
             ]
             if any(module_path.startswith(prefix) for prefix in excluded_prefixes):
                 continue

docs/models/supported_models.md

@@ -52,6 +52,7 @@ th {
 |`Flux2Pipeline` | FLUX.2-dev | `black-forest-labs/FLUX.2-dev` |
 |`FishSpeechSlowARForConditionalGeneration` | Fish Speech S2 Pro | `fishaudio/s2-pro` |
 |`DreamIDOmniPipeline`| DreamID-Omni | `XuGuo699/DreamID-Omni` |
+|`VoxtralTTSForConditionalGeneration` | Voxtral TTS | `mistralai/tts-model` |
 
 
 ## List of Supported Models for NPU

docs/serving/speech_api.md

@@ -4,6 +4,7 @@ vLLM-Omni provides an OpenAI-compatible API for text-to-speech (TTS) generation.
 
 - **Qwen3-TTS** (`Qwen/Qwen3-TTS-12Hz-*`) -- Qwen3-based TTS with CustomVoice, VoiceDesign, and Base (voice cloning) task types. Output: 24 kHz.
 - **Fish Speech S2 Pro** (`fishaudio/s2-pro`) -- Dual-AR TTS with DAC codec. Supports text-to-speech and voice cloning via reference audio. Output: 44.1 kHz.
+- **Voxtral TTS** (`mistralai/tts-model`) -- AR + FlowMatching TTS with preset voices. Output: 24 kHz.
 
 Each server instance runs a single model (specified at startup via `vllm serve <model> --omni`).
 
@@ -28,6 +29,14 @@ vllm-omni serve fishaudio/s2-pro \
     --trust-remote-code \
     --enforce-eager \
     --gpu-memory-utilization 0.9
+
+# Voxtral TTS
+vllm serve mistralai/tts-model \
+    --stage-configs-path vllm_omni/model_executor/stage_configs/voxtral_tts.yaml \
+    --omni \
+    --port 8091 \
+    --trust-remote-code \
+    --enforce-eager
 ```
 
 ### Generate Speech
@@ -347,6 +356,12 @@ curl -X POST http://localhost:8091/v1/audio/speech \
 
 Fish Speech uses `ref_audio` and `ref_text` for voice cloning (no `task_type` needed). The `voice` field should be set to `"default"`. See the [Fish Speech online serving example](../user_guide/examples/online_serving/fish_speech.md) for details.
 
+### Voxtral TTS
+
+| Model | Description |
+|-------|-------------|
+| `mistralai/tts-model` | 3B AR + FlowMatching TTS. Supports text-to-speech with preset voices. |
+
 ## Error Responses
 
 ### 400 Bad Request