Skip to content

An example of using the whisper_encode / whisper_decode part of the API? #3271

Open
@artlav

Description

@artlav

Whisper API seem to expose a lot of the functions which seem to be meant to let you do what whisper_full does.

whisper_pcm_to_mel, whisper_encode, whisper_decode, as well as various token and logit handling stuff.

However, there are zero examples of it being used, either in the codebase, or on the searchable internet.
And looking through the whisper_full_with_state code doesn't help, since it's using internal calls and context structures.
So far i was not able to get it to produce any legible output.

Which brings us to this suggestion:

Could someone with a good understanding of how the code works please put together an example of how to use these?

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions