Lepton whisperx API translates audio instead of transcribing it

TL;DR

I send a base64 audio file with the language parameter set, but sometimes it still translates instead of transcribing.

Steps

1. The Lepton whisperx API translates my audio to another language automatically. I send a base64 audio file with the language parameter set, but sometimes it still translates instead of transcribing. How do I make it just transcribe?
2. This is a known quirk in the whisperx template: if the language is not specified, the model tries to auto-detect it, and depending on what it detects you can get a translation instead of a transcription. There was also a bug where the default language was hard-coded to en, which broke autodetection. The reliable fix is to pass the language explicitly in your request, for example language: zh for Chinese audio, and it will transcribe rather than translate. Source: https://github.com/leptonai/leptonai/issues/325

When to use

You are seeing this: The Lepton whisperx API translates my audio to another language automatically. Use this skill when you run into "Lepton whisperx API translates audio instead of transcribing it".

When not to use

If your error message or symptom does not match what is described above, this is probably not your fix. Search for your exact error text instead of forcing this one to fit.

Versions

No specific versions are mentioned in the source material, so treat the fix as generally applicable and check the examples against whatever you have installed.

Why this happens

The original report does not dig into a root cause. It documents the symptom and the fix that resolved it.

Sources

• https://github.com/leptonai/leptonai/issues/325

Question: The Lepton whisperx API translates my audio to another language automatically. I send a base64 audio file with the language parameter set, but sometimes it still translates instead of transcribing. How do I make it just transcribe?

Verified resolution: This is a known quirk in the whisperx template: if the language is not specified, the model tries to auto-detect it, and depending on what it detects you can get a translation instead of a transcription. There was also a bug where the default language was hard-coded to en, which broke autodetection. The reliable fix is to pass the language explicitly in your request, for example language: zh for Chinese audio, and it will transcribe rather than translate. Source: https://github.com/leptonai/leptonai/issues/325

Thread: https://vectle.com/threads/thr_7XcsFX4cw7BN5oZdRQPhZw