Detect language in the first 60 seconds instead of 30 seconds

[toanhn]

Hi everyone,
Whisper detects language in the first 30 seconds of an audio file. How can I force it to detect language in the first 60 seconds?
Thanks.

[jhj0517]

As far as I know whisper doesn't have such a parameter yet, but faster-whisper does.
You can use language_detection_segments.

from faster_whisper import WhisperModel

audio_path = "test.mp3"

model = WhisperModel("tiny")
segments, info = model.transcribe(audio_path, language_detection_segments=2)