3.17时间轴错误，字幕过密

[188198]

00:06:03,225 --> 00:06:03,725
这件事

682
00:06:03,726 --> 00:06:04,226
我可不会告诉任何人哦

683
00:06:04,227 --> 00:06:04,727
知道了？

684
00:06:04,728 --> 00:06:05,228
那就

685
00:06:05,229 --> 00:06:05,729
明天还要早起呢

686
00:06:05,730 --> 00:06:06,230
早点睡吧

687
00:06:06,231 --> 00:06:06,731
…

688
02:49:52,667 --> 02:49:54,667
字幕由 Transub 生成 [ www.transub.cc ]

近3小时的视频，字幕时间轴压缩在6分钟了。每条字幕最多1秒钟。cpu跑的，旧版本无问题

[dlsandy]

使用的模型也是和旧版本一样的吗？方便的话可以将引擎日志发一下。以便我判断该问题是什么原因引起的。

[188198]

使用的模型也是和旧版本一样的吗？方便的话可以将引擎日志发一下。以便我判断该问题是什么原因引起的。

模型和参数一样的，下面部分引擎日志
===== Transub Engine log 2026-09-06T23:59:46.782Z =====
[engine] 启动 O:\Transub-3.1.7-win\transub-engine\runtime\python.exe -m transub_engine serve --host 127.0.0.1 --port 8765
[transub-engine] ffmpeg: O:\Transub-3.1.7-win_internal\bin\ffmpeg.exe
[transub-engine] gpu: status=no_gpu name=- cublas12=False ct2Cuda=False ortGpuCuda=False pathInjected=1
INFO: Started server process [1812]
INFO: Waiting for application startup.
INFO: Application startup complete.
INFO: Uvicorn running on http://127.0.0.1:8765 (Press CTRL+C to quit)
INFO: 127.0.0.1:50694 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50696 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50695 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:50696 - "GET /v1/runtime/gpu HTTP/1.1" 200 OK
INFO: 127.0.0.1:50695 - "POST /v1/models/recommend HTTP/1.1" 200 OK
INFO: 127.0.0.1:50696 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:50757 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50758 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50756 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:50798 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50800 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50799 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:50808 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50810 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50809 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:50921 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50923 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50922 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:50929 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50934 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50928 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50935 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50936 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50937 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50930 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:50928 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:50931 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:50933 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50932 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50929 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50928 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:50929 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:50930 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:51275 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:51274 - "GET /v1/models HTTP/1.1" 200 OK
INFO: 127.0.0.1:51273 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
INFO: 127.0.0.1:51274 - "GET /v1/runtime/gpu HTTP/1.1" 200 OK
INFO: 127.0.0.1:51274 - "POST /v1/models/recommend HTTP/1.1" 200 OK
INFO: 127.0.0.1:51275 - "GET /v1/runtime/audio-separate HTTP/1.1" 200 OK
[engine] 开始批次 · 1 个文件 · ASR=qwen3-asr-1.7b-ja-anime-galgame · 档位=quality
[engine] 外部翻译适配器 · Sakura · http://127.0.0.1:51276/translate
[engine] #1/1 启动 NHDTC-248.mp4 · 准备引擎任务…
[engine] #1/1 翻译后端 · Sakura（sakura-galtransl-7b-q6k）
[engine] #1/1 加载识别模型 NHDTC-248.mp4 · 创建引擎任务（qwen3-asr-1.7b-ja-anime-galgame）…
[engine] #1/1 加载识别模型 NHDTC-248.mp4 · 加载识别模型（qwen3-asr-1.7b-ja-anime-galgame）…
INFO: 127.0.0.1:51272 - "POST /v1/jobs HTTP/1.1" 200 OK
[engine] #1/1 启动 NHDTC-248.mp4 · 任务已启动
INFO transub_engine.asr.qwen3_asr: Qwen3 island batch plan: model=qwen3-asr-1.7b-ja-anime-galgame mode=auto batch=2 device=cpu vram_free=None vram_total=None
[engine] #1/1 加载识别模型 NHDTC-248.mp4 · 加载 Qwen3-ASR（VAD 帧时间戳）…
The following layers were not sharded: thinker.model.layers..self_attn.v_proj.weight, thinker.audio_tower.layers..self_attn.out_proj.bias, thinker.audio_tower.layers..self_attn_layer_norm.bias, thinker.model.layers..self_attn.o_proj.weight, thinker.audio_tower.proj*.bias, thinker.audio_tower.layers..final_layer_norm.bias, thinker.model.layers..input_layernorm.weight, thinker.audio_tower.layers..self_attn.k_proj.weight, thinker.audio_tower.layers..self_attn.out_proj.weight, thinker.audio_tower.layers..self_attn.q_proj.weight, thinker.audio_tower.ln_post.weight, thinker.audio_tower.layers..fc*.weight, thinker.model.layers..self_attn.k_norm.weight, thinker.model.layers..post_attention_layernorm.weight, thinker.model.embed_tokens.weight, thinker.model.norm.weight, thinker.audio_tower.convd.bias, thinker.audio_tower.convd.weight, thinker.audio_tower.layers..self_attn.v_proj.weight, thinker.audio_tower.layers..self_attn.q_proj.bias, thinker.model.layers..mlp.gate_proj.weight, thinker.audio_tower.proj.weight, thinker.audio_tower.layers..final_layer_norm.weight, thinker.model.layers..self_attn.q_proj.weight, thinker.audio_tower.layers..self_attn.v_proj.bias, thinker.lm_head.weight, thinker.model.layers..self_attn.k_proj.weight, thinker.model.layers..self_attn.q_norm.weight, thinker.audio_tower.layers..fc*.bias, thinker.audio_tower.layers..self_attn_layer_norm.weight, thinker.model.layers..mlp.down_proj.weight, thinker.audio_tower.ln_post.bias, thinker.audio_tower.layers..self_attn.k_proj.bias, thinker.model.layers..mlp.up_proj.weight, thinker.audio_tower.conv_out.weight
The following generation flags are not valid and may be ignored: ['temperature']. Set TRANSFORMERS_VERBOSITY=info for more details.
INFO transub_engine.vad.ten_vad_seg: TEN VAD collect start: NHDTC-248.mp4 | threshold=0.45 | merge_gap=0.50s | min_speech=0.18s | max_media=10195s
INFO transub_engine.vad.ten_vad_seg: TEN VAD decode: ffmpeg | NHDTC-248.mp4 | max 10195s
INFO transub_engine.vad.ten_vad_seg: TEN VAD decode ok: ffmpeg | NHDTC-248.mp4 | 10194.7s @ 16000 Hz
INFO transub_engine.vad.ten_vad_seg: TEN VAD collect done: NHDTC-248.mp4 | audio=10194.7s | raw=13625 | merged=1917 | spoken=3389.4s (33%) | first=4.9-7.6s | last=10133.4-10134.0s
INFO transub_engine.vad.ten_vad_seg: TEN VAD PCM ready for island ASR | samples=163114653 | ~622.2 MB | reuse=yes
INFO transub_engine.asr.qwen3_asr: Qwen TEN frames: NHDTC-248.mp4 | islands=1917 -> frames=1925 | pcm=1925/1925 | batch=2 (auto) | first=4.9-7.6s
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 人声帧 1925 段 · 批 2 (auto) | Qwen3-ASR
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 1–2/1925 · 0:04
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 7–8/1925 · 0:32
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 13–14/1925 · 0:58
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 21–22/1925 · 1:33
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 29–30/1925 · 2:09
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 39–40/1925 · 2:37
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 47–48/1925 · 3:11
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 57–58/1925 · 3:59
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 65–66/1925 · 4:26
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 73–74/1925 · 4:51
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 77–78/1925 · 5:14
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 85–86/1925 · 6:44
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 97–98/1925 · 8:29
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 115–116/1925 · 10:35
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 127–128/1925 · 11:26
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 139–140/1925 · 12:18
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 147–148/1925 · 12:42
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 155–156/1925 · 13:10
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 161–162/1925 · 13:36
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 167–168/1925 · 14:00
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 175–176/1925 · 14:32
[engine] #1/1 原文提取 NHDTC-248.mp4 · TEN 183–184/1925 · 15:09

[dlsandy]

你可以尝试恢复默认设置（入口在设置 > 更多 > 重置为默认设置），再重新进行任务。

如果问题依旧未能解决，给你的临时建议是：暂时使用3.1.6版本（翻译质量跟3.1.7版本是一致的）
近期会更新3.2.0版本，在测试阶段我会加强时间轴方面的测试。

[188198]

你可以尝试恢复默认设置（入口在设置 > 更多 > 重置为默认设置），再重新进行任务。

如果问题依旧未能解决，给你的临时建议是：暂时使用3.1.6版本（翻译质量跟3.1.7版本是一致的） 近期会更新3.2.0版本，在测试阶段我会加强时间轴方面的测试。

好的，现在在使用旧版本。