chrissewell/whisperX
1
0
Fork 0
mirror of https://github.com/m-bain/whisperX.git synced 2025-07-01 18:17:27 -04:00
Code Issues Packages Projects Releases Wiki Activity

Improvement to transcription starting point with VAD

Browse Source
This commit is contained in:
Antoine Dufour
2023-02-18 11:12:23 -05:00
parent 4cb167a225
commit a1d2229416
1 changed files with 5 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
whisperx/transcribe.py
View File

@ -262,7 +262,6 @@ def merge_chunks(segments, chunk_size=CHUNK_LENGTH):
TODO: Make sure VAD segment isn't too long, otherwise it will cause OOM when input to alignment model
TODO: Or sliding window alignment model over long segment.
"""
curr_start = 0
curr_end = 0
merged_segments = []
seg_idxs = []
@ -275,7 +274,11 @@ def merge_chunks(segments, chunk_size=CHUNK_LENGTH):
for speech_turn in segments.get_timeline():
segments_list.append(Segment(speech_turn.start, speech_turn.end, "UNKNOWN"))
for sdx, seg in enumerate(segments_list):
assert segments_list, "segments_list is empty."
# Make sur the starting point is the start of the segment.
curr_start = segments_list[0].start
for seg in segments_list:
if seg.end - curr_start > chunk_size and curr_end-curr_start > 0:
merged_segments.append({
"start": curr_start,