Need an API to get all the available languages based on a model

[rpatnayakuni22]

Based on the model we download, I'm guessing the number of languages it supports will vary. Is there an API to detect what all languages a model supports? Or Is there a way to do it myself? Thanks in advance.

[ZachNagengast]

We dont have an API to detect it but its a good question. The main difference is that some whisper models are english only, indicated by the en in the naming scheme

WhisperKit/Sources/WhisperKit/Core/Models.swift

Lines 54 to 64 in 9c673e3

	case tiny
	case tinyEn
	case base
	case baseEn
	case small
	case smallEn
	case medium
	case mediumEn
	case large
	case largev2
	case largev3

As well as large-v3 which contains 1 extra language cantonese openai/whisper#1762

[rpatnayakuni22]

Thank you @ZachNagengast for your response.
I'm currently using "openai_whisper-small", so is it just English only? Are there any ways using the model folder structure or something else, we can get language information?

[ZachNagengast]

openai_whisper-small is multilingual because it doesnt have an .en in the name like openai_whisper-small.en

The other caveat is that whisper large v3 turbo (which we call openai_whisper-large-v3-v20240930 ) is technically multilingual but was not fine tuned on multiligual data, so it will still run in the same way as a multilingual model but not provide good outputs for anything beside english.