heimoshuiyu/faster-whisper
1
0
Fork 0
Code Issues Pull Requests 1 Packages Projects Releases Wiki Activity

Fix window end heuristic for hallucination_silence_threshold (#706)

Browse Source 
Removes the wishful heuristic causing more issues than it's fixing.

Same as https://github.com/openai/whisper/pull/2043

Example of the issue: https://github.com/openai/whisper/pull/1838#issuecomment-1960041500
This commit is contained in:
Purfview
2024-02-29 16:59:32 +00:00
committed by GitHub
parent 09cd57e7f3
commit 5090cc9d0d
1 changed files with 0 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
faster_whisper/transcribe.py
View File

@@ -661,14 +661,6 @@ class WhisperModel:
# skip silence before possible hallucinations
if options.hallucination_silence_threshold is not None:
threshold = options.hallucination_silence_threshold
if not single_timestamp_ending:
last_word_end = get_end(current_segments)
if last_word_end is not None and last_word_end > time_offset:
remaining_duration = window_end_time - last_word_end
if remaining_duration > threshold:
seek = round(last_word_end * self.frames_per_second)
else:
seek = previous_seek + segment_size
# if first segment might be a hallucination, skip leading silence
first_segment = next_words_segment(current_segments)