无言以对 发表于 2026-4-24 22:16
显卡太差的一次不要生成太长，重新运行生成短点的
我的显存16G，运存32G的，我只用了10秒的音频。
songtso8 发表于 2026-4-25 06:58
我的显存16G，运存32G的，我只用了10秒的音频。
复制控制台信息最后几行贴出来
失败了，报错
* To create a public link, set `share=True` in `launch()`.
开始生成视频
图片已上传: 96v.jpg
音频已上传: 录音.m4a
视频帧数分配: 节点1=81, 节点2=0, 节点3=0
got prompt
任务已提交: da4f12d1-ef19-453e-b0af-bd15532dd2bf
目标输出节点: 177 (帧数: 81)
Using pytorch attention in VAE
Using pytorch attention in VAE
VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
unexpected audio encoder: ['project_hid.bias', 'project_hid.weight', 'project_q.bias', 'project_q.weight', 'quantizer.codevectors', 'quantizer.weight_proj.bias', 'quantizer.weight_proj.weight']
Requested to load Wav2Vec2Model
loaded completely; 13654.80 MB usable, 189.00 MB loaded, full load: True
Detected service crash
Shutting down all services...
All services stopped
Shutting down all services...
All services stopped
Press any key to continue . . .
* To create a public link, set `share=True` in `launch()`.
开始生成视频
图片已上传: 96v.jpg
音频已上传: 6月4日.mp3
视频帧数分配: 节点1=81, 节点2=0, 节点3=0
got prompt
任务已提交: ce55a5c8-c7be-4302-9746-1bce6c61e77b
目标输出节点: 177 (帧数: 81)
Using pytorch attention in VAE
Using pytorch attention in VAE
VAE load device: cuda:0, offload device: cpu, dtype: torch.bfloat16
unexpected audio encoder: ['project_hid.bias', 'project_hid.weight', 'project_q.bias', 'project_q.weight', 'quantizer.codevectors', 'quantizer.weight_proj.bias', 'quantizer.weight_proj.weight']
Requested to load Wav2Vec2Model
loaded completely; 13654.80 MB usable, 189.00 MB loaded, full load: True
Detected service crash
Shutting down all services...
All services stopped
Shutting down all services...
All services stopped
Press any key to continue . . .
支持20系显卡吗？
不太行，几台电脑全部报一样的错
8188界面也没看到工作流啊，7860界面运行就报错，我4090才传了5秒音频
12G 显存 + 32G 内存起步，这门槛够喝一壶，50系虽然支持，但风扇估计要起飞。自动对齐帧数这个倒是真香，终于不用手动数帧数了。但 UC 网盘只给单次生成，夸克和百度给批量，UC 用户是犯了天条吗？批量都不配。
测试一下看看，每个都说的很完美
想试试看:lol