thewh1teagle / vibe

Transcribe on your own!
https://thewh1teagle.github.io/vibe/
MIT License
430 stars 28 forks source link

[Feature Request]: Input prompts, translation #129

Closed Kompootor closed 6 days ago

Kompootor commented 2 weeks ago

Describe the feature

Other local-run Whisper apps like Aiko (iOS only) support using prompts to cue transcription, which in my experience achieves dramatically better results when the prompt cues the number of speakers, topic of conversation (especially if specialized, like science), and/or multiple languages or accents.

Another Aiko feature, for which I am not sure the implementation, is to translate and transcribe on the same run (has the same processing time, so using the same model), As one might expect, in combination with prompting, this again gives much better results than transcribing and translating in separate runs on whatever models.

Vibe is especially nice in that it allows one to choose small or large models, or custom models, per choice. But prompting in particular I think would be a significant improvement.

thewh1teagle commented 2 weeks ago

Describe the feature

Other local-run Whisper apps like Aiko (iOS only) support using prompts to cue transcription, which in my experience achieves dramatically better results when the prompt cues the number of speakers, topic of conversation (especially if specialized, like science), and/or multiple languages or accents.

Another Aiko feature, for which I am not sure the implementation, is to translate and transcribe on the same run (has the same processing time, so using the same model), As one might expect, in combination with prompting, this again gives much better results than transcribing and translating in separate runs on whatever models.

Vibe is especially nice in that it allows one to choose small or large models, or custom models, per choice. But prompting in particular I think would be a significant improvement.

Hey, thanks for reporting! Vibe already supports prompting and it's indeed help improve the transcription. In the main window, just click the down arrow to open the advanced options before transcribing, and fill the prompt section. Let me know if it worked and improved your transcription :)