Need help fixing audio.
When I try to have audio enabled, I get this error: RuntimeError: The size of tensor a (11776) must match the size of tensor b (1539584) at non-singleton dimension 2 . When I skip all audio steps, I'm able to fully generate the video, obviously without audio. I have no idea how to fix this, any help would be great. I'm using the 10Eros_10SNodes_TripleSample_I2V.json workflow if that gives you any clue as to why audio won't work.
Across all generation types and seeds with different size image inputs? I'd delete the workflow and redownload and try a generation without touching anything and make sure it's a checkpoint model in the checkpoint loaders and all installed in the right spots. Try another different workflow too and see if it works at all in another one. Could be a pytorch version issue too if you haven't updated anything in a really long time.
Tried a new workflow and model, still not working. Also everything should be up to date. Are there any specific things you do to make audio work? Like low-res picture or specific loras?
Nevermind, fixed it by deleting the ComfyUI_smZNodes folder and restarting Comfy. Its in custom_nodes in case anyone has this problem.