Category: provider

Model Studio Qwen LiveTranslate

Validation

mkdir -p output/aliyun-qwen-livetranslate
python -m py_compile skills/ai/audio/aliyun-qwen-livetranslate/scripts/prepare_livetranslate_request.py && echo "py_compile_ok" > output/aliyun-qwen-livetranslate/validate.txt

Pass criteria: command exits 0 and output/aliyun-qwen-livetranslate/validate.txt is generated.

Output And Evidence

• Save translation session payloads and response summaries under output/aliyun-qwen-livetranslate/.

Critical model names

Use one of these exact model strings:

• qwen3-livetranslate-flash
• qwen3-livetranslate-flash-realtime

Typical use

• Chinese/English meeting interpretation
• Live subtitles in another language
• Call-center agent assist with translated captions

Normalized interface (audio.livetranslate)

Request

• model (string, optional): default qwen3-livetranslate-flash
• source_language (string, required)
• target_language (string, required)
• audio_format (string, optional): e.g. pcm
• sample_rate (int, optional): e.g. 16000

Response

• translated_text (string)
• source_text (string, optional)
• audio_url or audio_chunk (optional, model dependent)

Quick start

python skills/ai/audio/aliyun-qwen-livetranslate/scripts/prepare_livetranslate_request.py \
  --source-language zh \
  --target-language en \
  --output output/aliyun-qwen-livetranslate/request.json

Notes

• Prefer the realtime model for continuous streaming sessions.
• Prefer the non-realtime flash model for simpler integration and lower client complexity.

References

• references/sources.md