[Doc] Update Qwen3-TTS docs for consistency with Omni examples (#1226)

Signed-off-by: linyueqian <linyueqian@outlook.com>
Signed-off-by: Yueqian Lin <70319226+linyueqian@users.noreply.github.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
This commit is contained in:
Yueqian LinandHongsheng Liu authored and GitHub committed 2026-02-10 15:52:44 +08:00
1 parent 9327cccde6
commit acfe99e4d9
4 files changed
+230 -86

No files matched your search

@@ -1,9 +1,11 @@
# Qwen3-TTS Online Serving # Qwen3-TTS
Source <https://github.com/vllm-project/vllm-omni/tree/main/examples/online_serving/qwen3_tts>. Source <https://github.com/vllm-project/vllm-omni/tree/main/examples/online_serving/qwen3_tts>.
This directory contains examples for running Qwen3-TTS models with vLLM-Omni's online serving API. ## 🛠️ Installation
Please refer to [README.md](https://github.com/vllm-project/vllm-omni/tree/main/README.md)
## Supported Models ## Supported Models
@@ -12,34 +14,77 @@ This directory contains examples for running Qwen3-TTS models with vLLM-Omni's o
| `Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice` | CustomVoice | Predefined speaker voices with optional style control | | `Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice` | CustomVoice | Predefined speaker voices with optional style control |
| `Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign` | VoiceDesign | Natural language voice style description | | `Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign` | VoiceDesign | Natural language voice style description |
| `Qwen/Qwen3-TTS-12Hz-1.7B-Base` | Base | Voice cloning from reference audio | | `Qwen/Qwen3-TTS-12Hz-1.7B-Base` | Base | Voice cloning from reference audio |
| `Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice` | CustomVoice | Smaller/faster variant |
| `Qwen/Qwen3-TTS-12Hz-0.6B-Base` | Base | Smaller/faster variant for voice cloning |
## Quick Start ## Run examples (Qwen3-TTS)
### 1. Start the Server ### Launch the Server
```bash ```bash
# CustomVoice model (default) # CustomVoice model (predefined speakers)
./run_server.sh vllm serve Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice \
--stage-configs-path vllm_omni/model_executor/stage_configs/qwen3_tts.yaml \
--omni \
--port 8091 \
--trust-remote-code \
--enforce-eager
# Or specify task type # VoiceDesign model
./run_server.sh CustomVoice vllm serve Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign \
./run_server.sh VoiceDesign --stage-configs-path vllm_omni/model_executor/stage_configs/qwen3_tts.yaml \
./run_server.sh Base --omni \
--port 8091 \
--trust-remote-code \
--enforce-eager
# Base model (voice cloning)
vllm serve Qwen/Qwen3-TTS-12Hz-1.7B-Base \
--stage-configs-path vllm_omni/model_executor/stage_configs/qwen3_tts.yaml \
--omni \
--port 8091 \
--trust-remote-code \
--enforce-eager
``` ```
### 2. Run the Client If you have custom stage configs file, launch the server with command below
```bash
vllm serve Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice \
--stage-configs-path /path/to/stage_configs_file \
--omni \
--port 8091 \
--trust-remote-code \
--enforce-eager
```
Alternatively, use the convenience script:
```bash
./run_server.sh # Default: CustomVoice model
./run_server.sh CustomVoice # CustomVoice model
./run_server.sh VoiceDesign # VoiceDesign model
./run_server.sh Base # Base (voice clone) model
```
### Send TTS Request
Get into the example folder
```bash
cd examples/online_serving/qwen3_tts
```
#### Send request via python
```bash ```bash
# CustomVoice: Use predefined speaker # CustomVoice: Use predefined speaker
python openai_speech_client.py \ python openai_speech_client.py \
--text "你好,我是通义千问" \ --text "你好,我是通义千问" \
--voice Vivian \ --voice vivian \
--language Chinese --language Chinese
# CustomVoice with style instruction # CustomVoice with style instruction
python openai_speech_client.py \ python openai_speech_client.py \
--text "今天天气真好" \ --text "今天天气真好" \
--voice Ryan \ --voice ryan \
--instructions "用开心的语气说" --instructions "用开心的语气说"
# VoiceDesign: Describe the voice style # VoiceDesign: Describe the voice style
@@ -58,29 +103,86 @@ python openai_speech_client.py \
--ref-text "Original transcript of the reference audio" --ref-text "Original transcript of the reference audio"
``` ```
### 3. Using curl The Python client supports the following command-line arguments:
- `--api-base`: API base URL (default: `http://localhost:8091`)
- `--model` (or `-m`): Model name/path (default: `Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice`)
- `--task-type` (or `-t`): TTS task type. Options: `CustomVoice`, `VoiceDesign`, `Base`
- `--text`: Text to synthesize (required)
- `--voice`: Speaker/voice name (default: `vivian`). Options: `vivian`, `ryan`, `aiden`, etc.
- `--language`: Language. Options: `Auto`, `Chinese`, `English`, `Japanese`, `Korean`, `German`, `French`, `Russian`, `Portuguese`, `Spanish`, `Italian`
- `--instructions`: Voice style/emotion instructions
- `--ref-audio`: Reference audio file path or URL for voice cloning (Base task)
- `--ref-text`: Reference audio transcript for voice cloning (Base task)
- `--response-format`: Audio output format (default: `wav`). Options: `wav`, `mp3`, `flac`, `pcm`, `aac`, `opus`
- `--output` (or `-o`): Output audio file path (default: `tts_output.wav`)
#### Send request via curl
```bash ```bash
# Simple TTS request # Simple TTS request
curl -X POST http://localhost:8000/v1/audio/speech \ curl -X POST http://localhost:8091/v1/audio/speech \
-H "Content-Type: application/json" \ -H "Content-Type: application/json" \
-d '{ -d '{
"input": "Hello, how are you?", "input": "Hello, how are you?",
"voice": "Vivian", "voice": "vivian",
"language": "English" "language": "English"
}' --output output.wav }' --output output.wav
# With style instruction # With style instruction
curl -X POST http://localhost:8000/v1/audio/speech \ curl -X POST http://localhost:8091/v1/audio/speech \
-H "Content-Type: application/json" \ -H "Content-Type: application/json" \
-d '{ -d '{
"input": "I am so excited!", "input": "I am so excited!",
"voice": "Vivian", "voice": "vivian",
"instructions": "Speak with great enthusiasm" "instructions": "Speak with great enthusiasm"
}' --output excited.wav }' --output excited.wav
# List available voices in CustomVoice models # List available voices in CustomVoice models
curl http://localhost:8000/v1/audio/voices curl http://localhost:8091/v1/audio/voices
```
### Using OpenAI SDK
```python
from openai import OpenAI
client = OpenAI(base_url="http://localhost:8091/v1", api_key="none")
response = client.audio.speech.create(
model="Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice",
voice="vivian",
input="Hello, how are you?",
)
response.stream_to_file("output.wav")
```
### Using Python httpx
```python
import httpx
response = httpx.post(
"http://localhost:8091/v1/audio/speech",
json={
"input": "Hello, how are you?",
"voice": "vivian",
"language": "English",
},
timeout=300.0,
)
with open("output.wav", "wb") as f:
f.write(response.content)
```
### FAQ
If you encounter error about backend of librosa, try to install ffmpeg with command below.
```
sudo apt update
sudo apt install ffmpeg
``` ```
## API Reference ## API Reference
@@ -89,16 +191,31 @@ curl http://localhost:8000/v1/audio/voices
``` ```
POST /v1/audio/speech POST /v1/audio/speech
Content-Type: application/json
``` ```
This endpoint follows the [OpenAI Audio Speech API](https://platform.openai.com/docs/api-reference/audio/createSpeech) format with additional Qwen3-TTS parameters. This endpoint follows the [OpenAI Audio Speech API](https://platform.openai.com/docs/api-reference/audio/createSpeech) format with additional Qwen3-TTS parameters.
### Voices Endpoint
```
GET /v1/audio/voices
```
Lists available voices for the loaded model:
```json
{
"voices": ["aiden", "dylan", "eric", "one_anna", "ryan", "serena", "sohee", "uncle_fu", "vivian"]
}
```
### Request Body ### Request Body
```json ```json
{ {
"input": "Text to synthesize", "input": "Text to synthesize",
"voice": "Vivian", "voice": "vivian",
"response_format": "wav", "response_format": "wav",
"task_type": "CustomVoice", "task_type": "CustomVoice",
"language": "Auto", "language": "Auto",
@@ -114,56 +231,38 @@ This endpoint follows the [OpenAI Audio Speech API](https://platform.openai.com/
### Response ### Response
Returns audio data in the requested format (default: WAV). Returns binary audio data with appropriate `Content-Type` header (e.g., `audio/wav`).
## Parameters ## Parameters
### Standard OpenAI Parameters ### OpenAI Standard Parameters
| Parameter | Type | Default | Description | | Parameter | Type | Default | Description |
|-----------|------|---------|-------------| |-----------|------|---------|-------------|
| `input` | string | required | Text to synthesize | | `input` | string | **required** | Text to synthesize |
| `voice` | string | "Vivian" | Speaker/voice name | | `model` | string | server's model | Model to use (optional, should match server if specified) |
| `voice` | string | "vivian" | Speaker name (e.g., vivian, ryan, aiden) |
| `response_format` | string | "wav" | Audio format: wav, mp3, flac, pcm, aac, opus | | `response_format` | string | "wav" | Audio format: wav, mp3, flac, pcm, aac, opus |
| `speed` | float | 1.0 | Playback speed (0.25-4.0) | | `speed` | float | 1.0 | Playback speed (0.25-4.0) |
| `model` | string | optional | Model name (optional when serving single model) |
### Qwen3-TTS Parameters ### vLLM-Omni Extension Parameters
| Parameter | Type | Default | Description | | Parameter | Type | Default | Description |
|-----------|------|---------|-------------| |-----------|------|---------|-------------|
| `task_type` | string | "CustomVoice" | Task: CustomVoice, VoiceDesign, or Base | | `task_type` | string | "CustomVoice" | Task: CustomVoice, VoiceDesign, or Base |
| `language` | string | "Auto" | Language: Auto, Chinese, English, Japanese, Korean | | `language` | string | "Auto" | Language (see supported languages below) |
| `instructions` | string | "" | Voice style/emotion instructions | | `instructions` | string | "" | Voice style/emotion instructions |
| `max_new_tokens` | int | 2048 | Maximum tokens to generate | | `max_new_tokens` | int | 2048 | Maximum tokens to generate |
**Supported languages:** Auto, Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, Italian
### Voice Clone Parameters (Base task) ### Voice Clone Parameters (Base task)
| Parameter | Type | Required | Description | | Parameter | Type | Required | Description |
|-----------|------|----------|-------------| |-----------|------|----------|-------------|
| `ref_audio` | string | Yes* | Reference audio (file path, URL, or base64) | | `ref_audio` | string | **Yes** | Reference audio (URL or base64 data URL) |
| `ref_text` | string | No | Transcript of reference audio (for ICL mode) | | `ref_text` | string | No | Transcript of reference audio (for ICL mode) |
| `x_vector_only_mode` | bool | false | Use speaker embedding only (no ICL) | | `x_vector_only_mode` | bool | No | Use speaker embedding only (no ICL) |
## Python Usage
```python
import httpx
# Simple request
response = httpx.post(
"http://localhost:8000/v1/audio/speech",
json={
"model": "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice",
"input": "Hello world",
"voice": "Vivian",
},
timeout=300.0,
)
with open("output.wav", "wb") as f:
f.write(response.content)
```
## Limitations ## Limitations
@@ -172,10 +271,11 @@ with open("output.wav", "wb") as f:
## Troubleshooting ## Troubleshooting
1. **Connection refused**: Make sure the server is running on the correct port 1. **TTS model did not produce audio output**: Ensure you're using the correct model variant for your task type (CustomVoice task → CustomVoice model, etc.)
2. **Out of memory**: Reduce `--gpu-memory-utilization` in run_server.sh 2. **Connection refused**: Make sure the server is running on the correct port
3. **Unsupported speaker**: Check supported speakers via model documentation 3. **Out of memory**: Use smaller model variant (`Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice`) or reduce `--gpu-memory-utilization`
4. **Voice clone fails**: Ensure you're using the Base model variant for voice cloning 4. **Unsupported speaker**: Use `/v1/audio/voices` to list available voices for the loaded model
5. **Voice clone fails**: Ensure you're using the Base model variant for voice cloning
## Example materials ## Example materials
+71 -27
View File
@@ -9,6 +9,8 @@ This directory contains examples for running Qwen3-TTS models with vLLM-Omni's o
| `Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice` | CustomVoice | Predefined speaker voices with optional style control | | `Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice` | CustomVoice | Predefined speaker voices with optional style control |
| `Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign` | VoiceDesign | Natural language voice style description | | `Qwen/Qwen3-TTS-12Hz-1.7B-VoiceDesign` | VoiceDesign | Natural language voice style description |
| `Qwen/Qwen3-TTS-12Hz-1.7B-Base` | Base | Voice cloning from reference audio | | `Qwen/Qwen3-TTS-12Hz-1.7B-Base` | Base | Voice cloning from reference audio |
| `Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice` | CustomVoice | Smaller/faster variant |
| `Qwen/Qwen3-TTS-12Hz-0.6B-Base` | Base | Smaller/faster variant for voice cloning |
## Quick Start ## Quick Start
@@ -24,19 +26,27 @@ This directory contains examples for running Qwen3-TTS models with vLLM-Omni's o
./run_server.sh Base ./run_server.sh Base
``` ```
Or launch directly with vllm serve:
```bash
vllm serve Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice \
--stage-configs-path vllm_omni/model_executor/stage_configs/qwen3_tts.yaml \
--omni --port 8091 --trust-remote-code --enforce-eager
```
### 2. Run the Client ### 2. Run the Client
```bash ```bash
# CustomVoice: Use predefined speaker # CustomVoice: Use predefined speaker
python openai_speech_client.py \ python openai_speech_client.py \
--text "你好,我是通义千问" \ --text "你好,我是通义千问" \
--voice Vivian \ --voice vivian \
--language Chinese --language Chinese
# CustomVoice with style instruction # CustomVoice with style instruction
python openai_speech_client.py \ python openai_speech_client.py \
--text "今天天气真好" \ --text "今天天气真好" \
--voice Ryan \ --voice ryan \
--instructions "用开心的语气说" --instructions "用开心的语气说"
# VoiceDesign: Describe the voice style # VoiceDesign: Describe the voice style
@@ -59,25 +69,25 @@ python openai_speech_client.py \
```bash ```bash
# Simple TTS request # Simple TTS request
curl -X POST http://localhost:8000/v1/audio/speech \ curl -X POST http://localhost:8091/v1/audio/speech \
-H "Content-Type: application/json" \ -H "Content-Type: application/json" \
-d '{ -d '{
"input": "Hello, how are you?", "input": "Hello, how are you?",
"voice": "Vivian", "voice": "vivian",
"language": "English" "language": "English"
}' --output output.wav }' --output output.wav
# With style instruction # With style instruction
curl -X POST http://localhost:8000/v1/audio/speech \ curl -X POST http://localhost:8091/v1/audio/speech \
-H "Content-Type: application/json" \ -H "Content-Type: application/json" \
-d '{ -d '{
"input": "I am so excited!", "input": "I am so excited!",
"voice": "Vivian", "voice": "vivian",
"instructions": "Speak with great enthusiasm" "instructions": "Speak with great enthusiasm"
}' --output excited.wav }' --output excited.wav
# List available voices in CustomVoice models # List available voices in CustomVoice models
curl http://localhost:8000/v1/audio/voices curl http://localhost:8091/v1/audio/voices
``` ```
## API Reference ## API Reference
@@ -90,12 +100,26 @@ POST /v1/audio/speech
This endpoint follows the [OpenAI Audio Speech API](https://platform.openai.com/docs/api-reference/audio/createSpeech) format with additional Qwen3-TTS parameters. This endpoint follows the [OpenAI Audio Speech API](https://platform.openai.com/docs/api-reference/audio/createSpeech) format with additional Qwen3-TTS parameters.
### Voices Endpoint
```
GET /v1/audio/voices
```
Lists available voices for the loaded model:
```json
{
"voices": ["aiden", "dylan", "eric", "one_anna", "ryan", "serena", "sohee", "uncle_fu", "vivian"]
}
```
### Request Body ### Request Body
```json ```json
{ {
"input": "Text to synthesize", "input": "Text to synthesize",
"voice": "Vivian", "voice": "vivian",
"response_format": "wav", "response_format": "wav",
"task_type": "CustomVoice", "task_type": "CustomVoice",
"language": "Auto", "language": "Auto",
@@ -115,45 +139,64 @@ Returns audio data in the requested format (default: WAV).
## Parameters ## Parameters
### Standard OpenAI Parameters ### OpenAI Standard Parameters
| Parameter | Type | Default | Description | | Parameter | Type | Default | Description |
|-----------|------|---------|-------------| |-----------|------|---------|-------------|
| `input` | string | required | Text to synthesize | | `input` | string | **required** | Text to synthesize |
| `voice` | string | "Vivian" | Speaker/voice name | | `model` | string | server's model | Model to use (optional, should match server if specified) |
| `voice` | string | "vivian" | Speaker name (e.g., vivian, ryan, aiden) |
| `response_format` | string | "wav" | Audio format: wav, mp3, flac, pcm, aac, opus | | `response_format` | string | "wav" | Audio format: wav, mp3, flac, pcm, aac, opus |
| `speed` | float | 1.0 | Playback speed (0.25-4.0) | | `speed` | float | 1.0 | Playback speed (0.25-4.0) |
| `model` | string | optional | Model name (optional when serving single model) |
### Qwen3-TTS Parameters ### vLLM-Omni Extension Parameters
| Parameter | Type | Default | Description | | Parameter | Type | Default | Description |
|-----------|------|---------|-------------| |-----------|------|---------|-------------|
| `task_type` | string | "CustomVoice" | Task: CustomVoice, VoiceDesign, or Base | | `task_type` | string | "CustomVoice" | Task: CustomVoice, VoiceDesign, or Base |
| `language` | string | "Auto" | Language: Auto, Chinese, English, Japanese, Korean | | `language` | string | "Auto" | Language (see supported languages below) |
| `instructions` | string | "" | Voice style/emotion instructions | | `instructions` | string | "" | Voice style/emotion instructions |
| `max_new_tokens` | int | 2048 | Maximum tokens to generate | | `max_new_tokens` | int | 2048 | Maximum tokens to generate |
**Supported languages:** Auto, Chinese, English, Japanese, Korean, German, French, Russian, Portuguese, Spanish, Italian
### Voice Clone Parameters (Base task) ### Voice Clone Parameters (Base task)
| Parameter | Type | Required | Description | | Parameter | Type | Default | Description |
|-----------|------|----------|-------------| |-----------|------|----------|-------------|
| `ref_audio` | string | Yes* | Reference audio (file path, URL, or base64) | | `ref_audio` | string | null | Reference audio (URL or base64 data URL) |
| `ref_text` | string | No | Transcript of reference audio (for ICL mode) | | `ref_text` | string | null | Transcript of reference audio |
| `x_vector_only_mode` | bool | false | Use speaker embedding only (no ICL) | | `x_vector_only_mode` | bool | null | Use speaker embedding only (no ICL) |
## Python Usage ## Python Usage
### Using OpenAI SDK
```python
from openai import OpenAI
client = OpenAI(base_url="http://localhost:8091/v1", api_key="none")
response = client.audio.speech.create(
model="Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice",
voice="vivian",
input="Hello, how are you?",
)
response.stream_to_file("output.wav")
```
### Using httpx
```python ```python
import httpx import httpx
# Simple request
response = httpx.post( response = httpx.post(
"http://localhost:8000/v1/audio/speech", "http://localhost:8091/v1/audio/speech",
json={ json={
"model": "Qwen/Qwen3-TTS-12Hz-1.7B-CustomVoice", "input": "Hello, how are you?",
"input": "Hello world", "voice": "vivian",
"voice": "Vivian", "language": "English",
}, },
timeout=300.0, timeout=300.0,
) )
@@ -169,7 +212,8 @@ with open("output.wav", "wb") as f:
## Troubleshooting ## Troubleshooting
1. **Connection refused**: Make sure the server is running on the correct port 1. **"TTS model did not produce audio output"**: Ensure you're using the correct model variant for your task type (CustomVoice task → CustomVoice model, etc.)
2. **Out of memory**: Reduce `--gpu-memory-utilization` in run_server.sh 2. **Connection refused**: Make sure the server is running on the correct port
3. **Unsupported speaker**: Check supported speakers via model documentation 3. **Out of memory**: Use smaller model variant (`Qwen/Qwen3-TTS-12Hz-0.6B-CustomVoice`) or reduce `--gpu-memory-utilization`
4. **Voice clone fails**: Ensure you're using the Base model variant for voice cloning 4. **Unsupported speaker**: Use `/v1/audio/voices` to list available voices for the loaded model
5. **Voice clone fails**: Ensure you're using the Base model variant for voice cloning
@@ -5,10 +5,10 @@ to generate audio from text using Qwen3-TTS models.
Examples: Examples:
# CustomVoice task (predefined speaker) # CustomVoice task (predefined speaker)
python openai_speech_client.py --text "Hello, how are you?" --voice Vivian python openai_speech_client.py --text "Hello, how are you?" --voice vivian
# CustomVoice with emotion instruction # CustomVoice with emotion instruction
python openai_speech_client.py --text "I'm so happy!" --voice Vivian \ python openai_speech_client.py --text "I'm so happy!" --voice vivian \
--instructions "Speak with excitement" --instructions "Speak with excitement"
# VoiceDesign task (voice from description) # VoiceDesign task (voice from description)
@@ -30,7 +30,7 @@ import os
import httpx import httpx
# Default server configuration # Default server configuration
DEFAULT_API_BASE = "http://localhost:8000" DEFAULT_API_BASE = "http://localhost:8091"
DEFAULT_API_KEY = "EMPTY" DEFAULT_API_KEY = "EMPTY"
@@ -178,8 +178,8 @@ def parse_args():
parser.add_argument( parser.add_argument(
"--voice", "--voice",
type=str, type=str,
default="Vivian", default="vivian",
help="Speaker/voice name (default: Vivian). Options: Vivian, Ryan, etc.", help="Speaker/voice name (default: vivian). Options: vivian, ryan, aiden, etc.",
) )
parser.add_argument( parser.add_argument(
"--language", "--language",
@@ -33,7 +33,7 @@ echo "Starting Qwen3-TTS server with model: $MODEL"
vllm-omni serve "$MODEL" \ vllm-omni serve "$MODEL" \
--stage-configs-path vllm_omni/model_executor/stage_configs/qwen3_tts.yaml \ --stage-configs-path vllm_omni/model_executor/stage_configs/qwen3_tts.yaml \
--host 0.0.0.0 \ --host 0.0.0.0 \
--port 8000 \ --port 8091 \
--gpu-memory-utilization 0.9 \ --gpu-memory-utilization 0.9 \
--trust-remote-code \ --trust-remote-code \
--enforce-eager \ --enforce-eager \