#7910 i use cartesia as local api and when i use sonic 3.5 there is no alignment error but when i use sonic 3.6 why is there a

Unable to determine the specific cause based on the knowledge base. The knowledge base does not cover Cartesia as a TTS channel, nor any known discrepancy between Sonic 3.5 and Sonic 3.6 regarding alignment behavior, so I cannot confirm why alignment errors appear only with Sonic 3.6.

What the knowledge base does say about alignment errors in general: alignment issues in pyVideoTrans are not caused by the TTS engine "failing" per se — they are caused by the dubbed audio duration differing from the original subtitle/video duration. The SpeedRate alignment engine (videotrans/task/_rate.py) resolves this using: audio speed-up, video slow-down, silence trimming, and translation shortening. If a channel returns audio that does not respect the requested timing, or returns longer/shorter clips without clean silence boundaries, alignment will fail — and this can differ between model versions of the same service if the newer version changes pacing, adds leading/trailing silence, or returns different audio formats/sample rates.

Troubleshooting steps you can try (general, not Sonic-3.6-specific):

1. Enable Audio Speedup (配音加速) on the main interface, and optionally Slow video (视频慢速) as well — both together usually gives the best result.
2. Go to Menu → Tools/Options → Advanced Options → Dubbing (字幕配音) and enable Remove mute buffer at beginning and end of each subtitle's audio (移除每条字幕配音开头结尾静音缓冲), or Remove all mute buffers (移除每条字幕配音全部静音缓冲). If using the latter, adjust Mute removal level (静音移除力度) between low/default/max.
3. Check whether Sonic 3.6 output contains leading/trailing silence — if so, trimming it is what fixes alignment.
4. Verify the audio returned by Sonic 3.6 is a standard WAV; non-standard formats or wrong sample rate can break alignment in the merge stage.
5. Try selecting 二次识别 (SecondarySTT) — it re-transcribes the generated dubbing and produces tighter subtitle timing, which mitigates minor mismatches.
6. Re-run with Clear Output (清理已生成) checked, so stale cached audio/subtitles from previous runs are not reused.

To diagnose further, please provide:

• Software version (e.g. v4.15) and OS.
• Whether you are using the frozen Windows build or source deployment.
• Whether the Cartesia channel is a built-in channel or a custom API channel.
• The latest log file content (copy the last ~30 lines of error output) from the logs/ folder in the software installation directory.

Please refer to the following documentation:

• Subtitle/Sound/Video Alignment Principles
• Advanced Options
• All Dubbing Channels
• Custom TTS API