config-example.toml
1 [app] 2 segment_duration = 5 # 音频切分处理间隔,单位:分钟,建议值:5-10,如果视频中话语较少可以适当提高 3 transcribe_parallel_num = 1 # 并发进行转录的数量上限,建议值:1-3,如果使用了本地模型,最好调成1 4 translate_parallel_num = 3 # 并发进行翻译的数量上限,建议值:3,倍于转录的并发量,如果使用TPM限制严格的API,可以适当调低 5 transcribe_max_attempts = 3 # 转录最大尝试次数,建议值:3 6 translate_max_attempts = 5 # 翻译最大尝试次数,建议值:5,如果模型参数量较少或翻译失败率较高可以适当调高 7 max_sentence_length = 70 # 每句最大字符数,超过这个长度的句子会被拆分,建议值:50-70 8 proxy = "" # 网络代理地址,格式如http://127.0.0.1:7890,可不填 9 10 [server] 11 host = "127.0.0.1" 12 port = 8888 13 14 # 下方的配置不是都要填,请结合文档说明进行配置 15 16 [llm] #支持openai,deepseek,通义千问等所有兼容openai请求格式的模型服务 17 base_url = "" # 自定义base url,可配合转发站密钥使用,留空为openai官方api 18 api_key = "" # API密钥 19 model = "" # 指定模型名,可通过此字段结合base_url使用外部任何与OpenAI API兼容的大模型服务,留空默认为gpt-4o-mini 20 json = false # 所使用的llm接口是否支持json格式,如果支持请设置为true,若不知道这是什么,请保持为false 21 22 [transcribe] # 视频转文本支持多种方案,配置时先填provider,再填对应的配置 23 provider = "openai" #语音识别,当前可选值:openai,fasterwhisper,whisperkit,whisper.cpp,aliyun。(fasterwhisper不支持macOS,whisperkit只支持M芯片) 24 enable_gpu_acceleration = false # 给fasterwhisper进行GPU加速选项,50系显卡请务必开启,否则无法正常运行 25 [transcribe.openai] 26 base_url = "" 27 api_key = "" 28 model = "whisper-1" 29 [transcribe.fasterwhisper] 30 model = "medium" # fasterwhisper的本地模型可选值:tiny,medium,large-v2。建议medium及以上 31 [transcribe.whisperkit] 32 model = "large-v2" # whisperkit的本地模型可选值:large-v2 33 [transcribe.whispercpp] 34 model = "large-v2" # whispercpp的本地模型可选值:large-v2 35 [transcribe.aliyun] # provider选aliyun这块就都要填 36 [transcribe.aliyun.oss] 37 access_key_id = "" 38 access_key_secret = "" 39 bucket = "" 40 [transcribe.aliyun.speech] 41 access_key_id = "" 42 access_key_secret = "" 43 app_key= "" 44 45 [tts] 46 provider = "aliyun" # 可选值:openai,aliyun,edge-tts 47 [tts.openai] 48 base_url = "" 49 api_key = "" 50 model = "" # gpt-4o-mini-tts, tts-1, tts-1-hd 51 [tts.aliyun] # provider选aliyun这块就都要填 52 [tts.aliyun.oss] 53 access_key_id = "" 54 access_key_secret = "" 55 bucket = "" 56 [tts.aliyun.speech] 57 access_key_id = "" 58 access_key_secret = "" 59 app_key= ""