fix(qwen-asr): enable timestamp output when forced_aligner is configured#10013
Merged
Conversation
There was a problem hiding this comment.
Pull request overview
Note
Copilot was unable to run its full agentic suite in this review.
Adds support for returning and parsing word/segment timestamps when the Qwen ASR forced aligner is available.
Changes:
- Detects presence of a forced aligner and requests timestamps from
model.transcribe. - Adds parsing for forced-aligner timestamp objects (
start_time,end_time,text) in addition to tuple/list timestamps.
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+155
to
+158
| results = self.model.transcribe( | ||
| audio=audio_path, language=language, context=context, | ||
| return_time_stamps=has_aligner, | ||
| ) |
Comment on lines
+171
to
+176
| if hasattr(ts, 'start_time') and hasattr(ts, 'end_time') and hasattr(ts, 'text'): | ||
| # ForcedAlignItem dataclass (from qwen_asr forced aligner) | ||
| start_ms = int(ts.start_time * 1000) if ts.start_time is not None else 0 | ||
| end_ms = int(ts.end_time * 1000) if ts.end_time is not None else 0 | ||
| seg_text = ts.text or "" | ||
| elif isinstance(ts, (list, tuple)) and len(ts) >= 3: |
Two bugs prevented timestamps from working in the qwen-asr backend: 1. transcribe() was called without return_time_stamps=True, so the forced aligner was loaded but never invoked. Now we pass return_time_stamps=True when a forced_aligner is present. 2. The timestamp parsing code expected (list, tuple) items, but the qwen_asr library returns ForcedAlignItem dataclass instances with .text, .start_time, .end_time attributes. Added hasattr() check to handle this correctly, falling back to tuple parsing for backward compatibility.
- Wrap return_time_stamps kwarg in try/except TypeError for safety - Add defensive float() normalization for timestamp times - Use str() for text extraction to ensure string type
fqscfqj
force-pushed
the
fix/qwen-asr-timestamps
branch
from
4f283ba to
346c5d2
Compare
The Go server reads TranscriptSegment.start/end via time.Duration, which is in nanoseconds. Previously the backend sent milliseconds (* 1000), causing timestamps to be 1000x too small (e.g. 8e-8 instead of 0.08). Convert seconds → nanoseconds (* 1e9) instead. Also applies to the legacy tuple path for consistency.
Contributor
Author
Additional fix: seconds → nanosecondsWhile testing this PR against a real deployment, I discovered a third issue beyond the two bugs described above: Bug 3: Timestamp unit mismatch between Python backend and Go serverThe Go server reads (int64) and wraps them in : // core/backend/transcript.go
segments = append(segments, &schema.TranscriptSegment{
Start: time.Duration(s.Start),
End: time.Duration(s.End),
})Go's Fix: Convert seconds → nanoseconds ( Verified outputAfter applying all three fixes, {"segments": [{"id": 0, "start": 0.08, "end": 0.24, "text": "今"}, ...]}Pushed the additional commit to the PR branch. |
added 4 commits
May 26, 2026 10:05
Read request.timestamp_granularities from the gRPC request. - 'word': return one segment per aligned item (character / word) - 'segment' (default): merge consecutive items at sentence boundaries Sentence boundaries detected via CJK punctuation (。！？；…) and Latin endings (. ! ? ;). This matches the OpenAI Whisper API contract where omitting the parameter defaults to segment-level.
Unicode curly quotes (U+2018/2019) were being interpreted as Python string delimiters, causing SyntaxError. Use explicit unicode escapes.
The forced aligner strips punctuation from its output, so text-based sentence detection doesn't work. Instead, detect segment boundaries by measuring time gaps between consecutive aligned items. Threshold = max(median_gap * 4, 0.3s). This cleanly separates intra-sentence gaps (< 0.24s) from inter-sentence gaps (> 0.3s) across Chinese, English, and other languages.
The forced aligner strips whitespace from tokenized text, so English words like ['hello', 'world'] were joined as 'helloworld'. Add _smart_join() that inserts spaces between non-CJK tokens while keeping CJK characters and punctuation unspaced. Works for Chinese, English, Korean, Japanese, and mixed-language text.
mudler
approved these changes
May 26, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
The qwen-asr backend loads the
forced_alignermodel correctly but never actually produces timestamps. All segments returnstart=0, end=0.Two bugs cause this:
Bug 1:
return_time_stampsnot passed totranscribe()Qwen3ASRModel.transcribe()defaultsreturn_time_stamps=False. The backend never passesTrue, so the forced aligner is loaded but silently skipped during inference.Bug 2: Timestamp item format mismatch
The parsing code checks
isinstance(ts, (list, tuple)), butqwen_asrreturnsForcedAlignItemdataclass instances with.text,.start_time,.end_timeattributes — not tuples. The check always fails, so timestamps are zeroed out even if Bug 1 were fixed.Fix
return_time_stamps=Truetotranscribe()when aforced_aligneris loaded.hasattr()check forForcedAlignItemdataclass before falling back to tuple parsing.Testing
Verified against
qwen3-asr-0.6bwithQwen/Qwen3-ForcedAligner-0.6B— timestamps now return correctly inverbose_json,srt, andvttformats.