For any given cue, no audio is produced in exactly two situations, and neither one stops the whole task — which is why only some cues are affected:
- The cue’s text is empty. A blank subtitle line produces no dubbing. After AI translation, blank lines commonly appear because the AI merged two adjacent short subtitles into one sentence and left the next line empty (this is intentional, to keep the timeline aligned with the original video). Blank lines are never dubbed.
- The TTS request for that single line failed and was given up on. Individual line failures are logged and skipped; the task only aborts if all lines fail. Edge-TTS 403 / rate-limit / proxy errors are the most common cause of sporadic missing lines.
Steps to confirm and fix:
Open the target-language SRT in the output folder (_video_out/) and check whether the cues with no audio are blank lines.
- If yes: go to
Advanced Options - Subtitle Translation and set AI Translation Batch Size (Lines) to 1, or uncheck SendSRT to use line-by-line translation, so the AI cannot merge lines. Alternatively fill those blank lines manually before dubbing.
Open the latest file in the logs/ folder and search for Dubbing failed for subtitle # or edge-tts error.
- If found: go to
Advanced Options - Dubbing Settings, set EdgeTTS Max Concurrency to 1 and Post-Dubbing Delay (s) to 5–10, then retry with Clear Output checked.
- Make sure the spoken/target language and voice are correct for those lines; a voice that does not support the target language fails only on some lines.
- Confirm the failing cues are not lines consisting only of punctuation/symbols or pure numbers — those can be stripped during recognition cleanup or text normalization and end up empty.
- To verify per-line output, enable
Advanced Options - Dubbing - Keep Segment Audio Files, re-run, and inspect the individual audio files.
If none of the above matches (the SRT lines are non-empty and the log shows no dubbing error), the cause cannot be determined from the information given — please provide the last ~30 lines of the newest logs/YYYYMMDD.log around the dubbing stage, plus the software version and whether it is the packaged Windows build or a source deployment.
Please refer to the related documentation: