v2mixer.livev2.comDIGITIZED as NEW VERSION
- (English)
- v2mixer (standard/profession edition) voice to subtitle, use model from OpenAI whisper -> whisper.cpp -> whisper.cpp.j
- privacy reason, v2mixer this funcntion design for offline mode, it never post the audio data to server, so have to download the model run by your machine.
- you can trainyour model, or download public model
- for GPU mode(is the default), the model will load in video memory, so:
- video memory <= 2G, ggml-small.binis recommend.
- video memory <= 4G, ggml-medium-*.binare supported
- video memory > 4G, ggml-large-*.binare supported
- for CPU mode, doesn't need video memory, support all model
- you can convert from OpenAI whisper model, see README
- the small version small.bin here
- (Chinese)
- v2mixer （标准版和专业版）语音转字幕功能，使用模型来自 OpenAI whisper -> whisper.cpp -> whisper.cpp.j
- 隐私原因，v2mixer 此功能为离线设计，即不会发送任何语音数据到服务器，所以需要单独下载模型以使用你的设备进行识别。
- 你可以训练自己的模型，或者下载公共版本的模型
- 对于显卡模式（默认），模型将在显存加载，因此大小由显存决定：
- 显存<=2G，推荐 ggml-small.bin
- 显存<=4G，额外支持 ggml-medium-*.bin
- 显存>4G，额外支持 ggml-large-*.bin
- 公版模型推荐 small 模型，尤其对非英语识别，错映污染要小的多
- 对于CPU模式，模型大小由设备内存决定，只要内存足够就支持任意模型
- 你也可以转换 OpenAI whisper 的模型，详情点 README