MultiTalk/InfiniteTalk broken with ComfyUI 0.19.x - shape mismatch in multitalk.py

Open
  • Is something wrong?
#1,998 3 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Assessment

Difficulty
4/5
Estimated time
3-5 days
Newbie friendliness
52/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Quiet
Tech stack
python
Domain
backend

Research direction

Start with multitalk.py at line 247, then trace how the frame count and audio cross-attention tensor dimensions are calculated. Reproduce the reported workflow with ComfyUI 0.19.x and compare the comfy and default rope settings. Done means MultiTalk/InfiniteTalk runs without either shape error for the reported 480-frame input.

Written by the indexing model from the issue text.

Description

Environment

  • ComfyUI: v0.19.1
  • ComfyUI-WanVideoWrapper: main branch (latest)
  • GPU: NVIDIA RTX 4500 Ada (24GB VRAM)
  • OS: Windows 11

Problem

MultiTalk/InfiniteTalk workflow stopped working after updating ComfyUI to 0.19.x.

Two errors occur depending on rope_function setting:

With rope_function = comfy:
RuntimeError: The size of tensor a (30450) must match the size of tensor b (29925) at non-singleton dimension 1

With rope_function = default:
RuntimeError: shape '[20, 1400, 5120]' is invalid for input of size 146048000
at multitalk.py line 247: x = x.view(B * N_t, S, self.dim)

Details

  • Audio: 20s WAV, 16kHz, cropped via AudioCrop node
  • Frames: 480 (log shows "Audio duration (480 frames) is shorter than requested (481 frames). Using 480 frames.")
  • Resolution: 800x456
  • Model: InfiniteTalk single speaker

The off-by-one (480 vs 481 frames) seems to cause the tensor mismatch in the audio cross-attention reshape.

Workflow was working fine on ComfyUI 0.18.3.

Dominant language
Python
Stars
6.8k
Forks
712
No merged PRs in 30d

Getting set up

This project ships no dev container, Dockerfile or contributing guide, so setting up is up to you: start from its README, and see our first-contribution guide for the general steps.

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from kijai/ComfyUI-WanVideoWrapper

All issues in kijai/ComfyUI-WanVideoWrapper

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.