替换了llm基座,修改了调用接口,后续ASR采用vllm部署qwen3omni-30B-A3B可以做更细腻的音频识别(生成的是上文本描述),使用vllm部署qwen-asr-1.7B以满足实时场景(生成的是一对一字幕),TTS采用UV部署GPT-SoVITS轻量级TTS引擎可复制音色且300M的参数满足实时对话要求

This commit is contained in:
minorQAQ committed 2026-05-16 00:08:50 +08:00
1 parent 110fe78016
commit 64cabe63d7
5 files changed
+38 -9

No files matched your search

+3 -1
View File
@@ -8,7 +8,9 @@ REDIS_URI=redis://localhost:6379/0
# LLM_API_KEY='semi-ai.tech'
# LLM_BASE_URL="http://192.168.137.96:8000/v1"
# LLM_MODEL = "Qwen/Qwen3.6-35B-A3B-FP8"
ASR_TTS_BASE_URL = "http://192.168.137.96:8100"
# ASR_TTS_BASE_URL = "http://192.168.137.96:8100"
TTS_BASE_URL = "http://192.168.137.96:8902"
ASR_BASE_URL = "http://192.168.137.96:8901/v1"
#6000pro:
LLM_API_KEY='semi-ai.tech'