I’m also trying to add the language parameter when transcribing. Because sometimes faster-whisper (using distil-large-v3) incorrectly determines the language as english instead of german, and then the whole transcription is trash. Gemini and Claude are suggesting adding a ‘body’ field of type string, name language and value de. BUT it doesn’t support type string but form-data only, and when choosing form-data, faster-whisper just doesn’t read it.
Anyone successfully added language parameter?
Hey @latoqumu Did you try sending language in the body like this:
@latoqumu move model out of query params and into the body alongside language — faster-whisper’s OpenAI-compatible endpoint expects everything as multipart form fields, not query strings, so it’s probably ignoring your query param too.
swap de for whatever language code you need, the rest should just work.
hi @latoqumu 
my workaround would be to import a working cURL request into the HTTP Request node and compare the raw multipart payload, because that usually reveals whether the local Whisper server is actually OpenAI-compatible for language or expects a different field name/body format.