Skip to content

New package @remotion/whisper-wasm #5023

@JonnyBurger

Description

@JonnyBurger

This is a bounty issue which is preassigned to @hunxjunedo. Please do not attempt this!

We currently have the @remotion/install-whisper-cpp module which can only be used on the server. Oftentimes it would be useful to bring this functionality to the frontend, for ease of use and simplicity.

This would require 2 functions:

  • downloadWhisperModel() - This should download one of the models that are available and store it in IndexedDB (same as in [1] from [2])
  • transcribe() - This should take some 16Hkz WAV AudioData and return the raw transcript of the Whisper.cpp output

Criteria

  • The 2 APIs should resemble the ones from @remotion/install-whisper-cpp, but changes which seem appropriate should be made.
  • There should be 2 new documentation pages for the APIs, following the same structure as the @remotion/install-whisper-cpp documentation.
  • There should be an index page, like there is for @remotion/install-whisper-cpp, with a table of contents.
  • transcribe() should have an onProgress() callback.

[1]

Image

[2] https://whisper.ggerganov.com/

💎 This issue has a bounty on it!

Read our contributing guidelines:

/bounty 600

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions