Instructions to use MoYoYoTech/VoiceDialogue with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use MoYoYoTech/VoiceDialogue with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("text-to-speech", model="MoYoYoTech/VoiceDialogue")

# pip install -U transformers accelerate
# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("MoYoYoTech/VoiceDialogue", device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use MoYoYoTech/VoiceDialogue with llama.cpp:

Install (macOS, Linux)

curl -LsSf https://llama.app/install.sh | sh
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf MoYoYoTech/VoiceDialogue:Q6_K
# Run inference directly in the terminal:
llama cli -hf MoYoYoTech/VoiceDialogue:Q6_K

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama serve -hf MoYoYoTech/VoiceDialogue:Q6_K
# Run inference directly in the terminal:
llama cli -hf MoYoYoTech/VoiceDialogue:Q6_K

Build from source code

git clone https://github.com/ggml-org/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-app
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama serve -hf MoYoYoTech/VoiceDialogue:Q6_K
# Run inference directly in the terminal:
./build/bin/llama cli -hf MoYoYoTech/VoiceDialogue:Q6_K

Use Docker

docker model run hf.co/MoYoYoTech/VoiceDialogue:Q6_K

Llama new
LM Studio
Jan
Ollama
How to use MoYoYoTech/VoiceDialogue with Ollama:
```
ollama run hf.co/MoYoYoTech/VoiceDialogue:Q6_K
```
Unsloth Desktop

How to use MoYoYoTech/VoiceDialogue with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf MoYoYoTech/VoiceDialogue:Q6_K

Configure the model in Pi

# Install Pi:
npm install -g @earendil-works/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "MoYoYoTech/VoiceDialogue:Q6_K"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Docker Model Runner
How to use MoYoYoTech/VoiceDialogue with Docker Model Runner:
```
docker model run hf.co/MoYoYoTech/VoiceDialogue:Q6_K
```

Lemonade

How to use MoYoYoTech/VoiceDialogue with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull MoYoYoTech/VoiceDialogue:Q6_K

Run and chat with the model

lemonade run user.VoiceDialogue-Q6_K

List all available models

lemonade list

Hermes Agent

How to use MoYoYoTech/VoiceDialogue with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf MoYoYoTech/VoiceDialogue:Q6_K

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default MoYoYoTech/VoiceDialogue:Q6_K

Run Hermes

hermes

Atomic Chat

OpenClaw

How to use MoYoYoTech/VoiceDialogue with OpenClaw:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama serve -hf MoYoYoTech/VoiceDialogue:Q6_K

Configure OpenClaw

# Install OpenClaw:
npm install -g openclaw@latest
# Register the local server and set it as the default model:
openclaw onboard --non-interactive --mode local \
  --auth-choice custom-api-key \
  --custom-base-url http://127.0.0.1:8080/v1 \
  --custom-model-id "MoYoYoTech/VoiceDialogue:Q6_K" \
  --custom-provider-id llama-cpp \
  --custom-compatibility openai \
  --custom-text-input \
  --accept-risk \
  --skip-health

Run OpenClaw

openclaw agent --local --agent main --message "Hello from Hugging Face"

hzeng412 Claude Fable 5 commited on Jun 11

Commit

21dbad4

1 Parent(s): 3c70498

ASR: use auto language detection to reject non-speech (fixes hallucination)

Browse files

Forcing a language makes Qwen3-ASR output transcription text
unconditionally, so silence/noise segments that slip through VAD
produce hallucinated words (FunASR/Whisper do this on every such
segment). In auto mode the model returns an empty string for
non-speech, which the pipeline already drops.

Verified: silence, low and medium noise -> '' ; zh/en speech unchanged.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Files changed (1) hide show

src/voice_dialogue/asr/models/qwen.py +4 -10

src/voice_dialogue/asr/models/qwen.py CHANGED Viewed

@@ -10,12 +10,6 @@ from voice_dialogue.asr.models.base import ASRInterface
 from voice_dialogue.asr.utils import ensure_minimum_audio_duration
 from voice_dialogue.utils.logger import logger
-# Qwen3-ASR 的 language 参数使用语言全名
-LANGUAGE_NAME_MAPPING = {
-    'zh': 'Chinese',
-    'en': 'English',
-}
 DEFAULT_MODEL = os.environ.get('QWEN_ASR_MODEL', 'Qwen/Qwen3-ASR-1.7B')
 TARGET_SAMPLE_RATE = 16000
@@ -60,11 +54,11 @@ class QwenASRClient(ASRInterface):
     def transcribe(self, audio_array: np.ndarray, language: str = None) -> str:
         audio_array = ensure_minimum_audio_duration(audio_array)
-        # 未指定语言时交给模型自动检测（Qwen3-ASR 自带语种识别）
-        qwen_language = LANGUAGE_NAME_MAPPING.get(language)
         results = self.model.transcribe(
             audio=(audio_array, TARGET_SAMPLE_RATE),
-            language=qwen_language,
         )
         return ' '.join(result.text for result in results).strip()

 from voice_dialogue.asr.utils import ensure_minimum_audio_duration
 from voice_dialogue.utils.logger import logger
 DEFAULT_MODEL = os.environ.get('QWEN_ASR_MODEL', 'Qwen/Qwen3-ASR-1.7B')
 TARGET_SAMPLE_RATE = 16000
     def transcribe(self, audio_array: np.ndarray, language: str = None) -> str:
         audio_array = ensure_minimum_audio_duration(audio_array)
+        # 始终使用自动语种检测：指定语言会强制模型"只输出转写文本"，
+        # 静音/噪声段会被迫编出幻听文字；自动模式下非语音段返回空串，
+        # 由上游丢弃，从根上消除幻听。
         results = self.model.transcribe(
             audio=(audio_array, TARGET_SAMPLE_RATE),
+            language=None,
         )
         return ' '.join(result.text for result in results).strip()