Bug: Faster-Whisper fails on no-speech audio due to language detection error

[le-luchina]

Description
The transcribe method fails with a ValueError (max() arg is an empty sequence) when processing audio files with no speech, especially after VAD filtering. This issue occurs during language detection when no segments remain after filtering.

Steps to Reproduce

1. Use an audio file with no discernible speech or where VAD filters remove all speech.
2. Call transcribe with vad_filter=True.
3. Observe the following error:

   ValueError: max() arg is an empty sequence
   
   

Expected Behavior
The method should return a transcription object with empty or None fields instead of raising an error.

Environment
Faster-Whisper Version: 1.0.3
Python Version: 3.10
OS: Ubuntu 22.04
GPU

Suggested Fix
Add a check after VAD filtering to handle empty audio gracefully and return a formatted but empty result.

Error Traceback

File "/opt/src/audio/transcript_extractor.py", line 280, in _transcribe
transcription_segments, transcription_info = model.transcribe(
File "/usr/local/lib/python3.10/dist-packages/faster_whisper/transcribe.py", line 419, in transcribe
language = max(…
ValueError: max() arg is an empty sequence

[MahmoudAshraf97]

This was solved in the latest version, please update and test again