开始着手重构，要快的话还是直接调模型，现在处理8K单眼，在我电脑上处理1分钟时间从原来的20分钟减到10分钟了，还有继续压的余地
zelefans 发表于 2026-5-30 19:13
开始着手重构，要快的话还是直接调模型，现在处理8K单眼，在我电脑上处理1分钟时间从原来的20分钟减到10分 ...
加油,有没有办法提高下lada的质量，跟Jasna差得有点远
kf018@10050 发表于 2026-5-30 20:37
加油,有没有办法提高下lada的质量，跟Jasna差得有点远
先只考虑效率优化，质量再说。需要的质量的还是选jasna模型。
正在做前置马赛克识别，抽取有效片段喂给lada/jasna。这样一键模式效率可以大幅提高
Error: the JSON object must be str, bytes or bytearray, not NoneType
Process completed. Total time elapsed: 0 hours, 0 minutes, 0 seconds.
字幕工具无法用啊。哪一步了出错了。
shui 发表于 2026-5-31 06:48
Error: the JSON object must be str, bytes or bytearray, not NoneType
Process completed. Tot ...
字幕工具哪个功能项报这个错。
看起来有点像全局配置文件有问题。看一下程序exe同目录里的vr_toolbox_config.json有没有，config目录里的subtitle_trans_config.json。项目只有这两个json文件
zelefans 发表于 2026-5-31 10:11
字幕工具哪个功能项报这个错。
看起来有点像全局配置文件有问题。看一下程序exe同目录里的vr_toolbox_con ...
这二个目录下文件都有。
为啥批量翻译出来的是音频不是字幕文件
yxryyy 发表于 2026-6-1 22:43
为啥批量翻译出来的是音频不是字幕文件
会先提取音频文件然后再处理 翻译
zelefans 发表于 2026-6-1 22:54
会先提取音频文件然后再处理 翻译
提取音频后就一直卡着不动了
本帖最后由 zelefans 于 2026-6-3 08:19 编辑
yxryyy 发表于 2026-6-3 07:39
提取音频后就一直卡着不动了
应该是你CUDA环境有问题导致卡住。刚跑了一下，正常的日志应该如下：
开始批量生成字幕...
Initializing ASR model...
Loaded ASR model on CUDA (workers: 4)
Processing: 4k2.me@atvr00068_1_8k.mp4
Converting video to audio: 4k2.me@atvr00068_1_8k.mp4
Audio DSP preset: mild
Executing: ffmpeg -i G:\Downloads\atvr-068\4k2.me@atvr00068_1_8k.mp4 -hide_banner -loglevel warning -y -vn -ar 16000 -ac 1 -b:a 128k -f wav -af afftdn=nr=6:nf=-35:tn=1:ad=0.35:gs=6 G:\Downloads\atvr-068\4k2.me@atvr00068_1_8k.asr.wav
Transcription started...
Running ASR profile 'stable' (chunk=28.0s, scene_split=whisperseg, model=large-v3, scene_internal_vad=False)
Loaded WhisperSeg ONNX with providers: ['CPUExecutionProvider']
WhisperSeg VAD processing block 1/58 (1.7%)
WhisperSeg VAD processing block 5/58 (8.6%)
WhisperSeg VAD processing block 10/58 (17.2%)
WhisperSeg VAD processing block 15/58 (25.9%)
WhisperSeg VAD processing block 20/58 (34.5%)
WhisperSeg VAD processing block 25/58 (43.1%)
WhisperSeg VAD processing block 30/58 (51.7%)
WhisperSeg VAD processing block 35/58 (60.3%)
WhisperSeg VAD processing block 40/58 (69.0%)
WhisperSeg VAD processing block 45/58 (77.6%)
WhisperSeg VAD processing block 50/58 (86.2%)
WhisperSeg VAD processing block 55/58 (94.8%)
WhisperSeg VAD processing block 58/58 (100.0%)
WhisperSeg split into 113 speech regions, coalesced to 64 ASR chunks, kept 316.04s / 1710.19s (threshold=0.5, rms_gate_removed=23)
Transcribing chunk 1/64 at 24.38s, duration 2.86s
Transcribing chunk 2/64 at 29.64s, duration 2.90s
Transcribing chunk 3/64 at 36.28s, duration 5.76s
Transcribing chunk 4/64 at 47.54s, duration 14.68s
Transcribing chunk 5/64 at 70.08s, duration 7.94s
Transcribing chunk 6/64 at 82.84s, duration 2.64s
Transcribing chunk 7/64 at 92.66s, duration 3.90s
Transcribing chunk 8/64 at 100.32s, duration 5.02s
Transcribing chunk 9/64 at 111.78s, duration 2.52s
Transcribing chunk 10/64 at 119.98s, duration 2.52s
Transcribing chunk 11/64 at 142.70s, duration 7.32s
Transcribing chunk 12/64 at 160.72s, duration 2.44s
Transcribing chunk 13/64 at 170.72s, duration 6.24s
Transcribing chunk 14/64 at 179.62s, duration 5.38s
Transcribing chunk 15/64 at 187.64s, duration 2.62s
Transcribing chunk 16/64 at 204.44s, duration 3.56s
Transcribing chunk 17/64 at 221.58s, duration 14.14s
Transcribing chunk 18/64 at 285.14s, duration 5.90s
Transcribing chunk 19/64 at 329.98s, duration 4.10s
Transcribing chunk 20/64 at 381.62s, duration 3.50s
Transcribing chunk 21/64 at 430.56s, duration 3.74s
Transcribing chunk 22/64 at 457.98s, duration 4.16s
Transcribing chunk 23/64 at 522.64s, duration 5.46s
Transcribing chunk 24/64 at 542.02s, duration 5.50s
Transcribing chunk 25/64 at 582.62s, duration 4.58s
Transcribing chunk 26/64 at 596.70s, duration 5.60s
Transcribing chunk 27/64 at 605.74s, duration 2.00s
Transcribing chunk 28/64 at 655.60s, duration 3.18s
Transcribing chunk 29/64 at 699.22s, duration 3.76s
Transcribing chunk 30/64 at 779.60s, duration 3.88s
Transcribing chunk 31/64 at 795.80s, duration 3.80s
Transcribing chunk 32/64 at 806.16s, duration 8.56s
Transcribing chunk 33/64 at 851.88s, duration 2.96s
Transcribing chunk 34/64 at 865.54s, duration 4.46s
Transcribing chunk 35/64 at 914.74s, duration 2.34s
Transcribing chunk 36/64 at 923.84s, duration 6.18s
Transcribing chunk 37/64 at 952.16s, duration 2.70s
Transcribing chunk 38/64 at 960.84s, duration 5.66s
Transcribing chunk 39/64 at 1004.58s, duration 5.66s
Transcribing chunk 40/64 at 1015.48s, duration 6.46s
Transcribing chunk 41/64 at 1033.58s, duration 4.30s
Transcribing chunk 42/64 at 1062.64s, duration 10.08s
Transcribing chunk 43/64 at 1075.90s, duration 3.18s
Transcribing chunk 44/64 at 1118.84s, duration 3.10s
Transcribing chunk 45/64 at 1139.98s, duration 7.88s
Transcribing chunk 46/64 at 1156.64s, duration 2.06s
Transcribing chunk 47/64 at 1188.94s, duration 8.42s
Transcribing chunk 48/64 at 1199.98s, duration 6.84s
Transcribing chunk 49/64 at 1237.96s, duration 2.50s
Transcribing chunk 50/64 at 1242.80s, duration 5.24s
Transcribing chunk 51/64 at 1251.40s, duration 2.54s
Transcribing chunk 52/64 at 1264.36s, duration 5.90s
Transcribing chunk 53/64 at 1304.06s, duration 2.80s
Transcribing chunk 54/64 at 1317.92s, duration 5.22s
Transcribing chunk 55/64 at 1338.30s, duration 5.70s
Transcribing chunk 56/64 at 1413.76s, duration 7.76s
Transcribing chunk 57/64 at 1435.30s, duration 2.54s
Transcribing chunk 58/64 at 1464.90s, duration 6.50s
Transcribing chunk 59/64 at 1561.26s, duration 2.78s
Transcribing chunk 60/64 at 1568.84s, duration 4.04s
Transcribing chunk 61/64 at 1663.18s, duration 6.40s
Transcribing chunk 62/64 at 1673.26s, duration 3.26s
Transcribing chunk 63/64 at 1678.88s, duration 5.74s
Transcribing chunk 64/64 at 1696.94s, duration 4.68s
Postprocess removed 0 duplicates, 1 noisy/repeated lines, 0 known hallucinations, 0 low-confidence silence lines
Saved SRT: G:\Downloads\atvr-068\4k2.me@atvr00068_1_8k.jp.srt | Time elapsed: 129.87s
Transcription complete.
Processing: 4k2.me@atvr00068_2_8k.mp4
Converting video to audio: 4k2.me@atvr00068_2_8k.mp4
Audio DSP preset: mild
Executing: ffmpeg -i G:\Downloads\atvr-068\4k2.me@atvr00068_2_8k.mp4 -hide_banner -loglevel warning -y -vn -ar 16000 -ac 1 -b:a 128k -f wav -af afftdn=nr=6:nf=-35:tn=1:ad=0.35:gs=6 G:\Downloads\atvr-068\4k2.me@atvr00068_2_8k.asr.wav
Transcription started...
Running ASR profile 'stable' (chunk=28.0s, scene_split=whisperseg, model=large-v3, scene_internal_vad=False)
WhisperSeg VAD processing block 1/58 (1.7%)
WhisperSeg VAD processing block 5/58 (8.6%)
WhisperSeg VAD processing block 10/58 (17.2%)
WhisperSeg VAD processing block 15/58 (25.9%)
WhisperSeg VAD processing block 20/58 (34.5%)
WhisperSeg VAD processing block 25/58 (43.1%)
。。。。。
你可以去下载我在论坛里最新发布的CUDA专版 https://4k2.me/forum.php?mod=viewthread&tid=46001，那里整合了CUDA完整环境，看看行不行。
