ComfyUI: Configure frame window and motion settings for Infinite Talk
ComfyUI
Demo summary
The user demonstrates how to adjust the frame window size and motion frame overlap to balance video smoothness, lip-sync consistency, and VRAM usage.
Step-by-step
- Connect an upscale image node to resize the input image
- Use a get image node to automatically extract width and height dimensions
- Set the model option to infinite talk
- Adjust the frame window size to balance smoothness and VRAM usage
- Set the motion frame overlap to between 10% and 30% of the window size
Options
- Reduce frame window size to 45 for 1280x720 resolution videos
- Increase motion frame values for steadier transitions
Watch out for
- Larger frame window values require more VRAM
- Higher motion frame values will slow down the generation process
- Leave the model setting on infinite talk even when using multiple voices
Tips
- Use a get image node so you do not have to manually input width and height every time you change image size
- Use larger frame window values to improve smoothness and lip-sync consistency
All demos from “ComfyUI Tutorial Series Ep 60 Infinite Talk (Audio-Driven Talking AI Characters)”
- 0:441:10Generate a talking head video with Infinite Talk in ComfyUIThe creator demonstrates the basic Infinite Talk workflow by uploading a portrait image and an audio file, adjusting the prompt and resolution, and rendering a lip-synced video.ComfyUI· AI Avatar Video Generator
- 5:411:07Configure frame window and motion settings for Infinite TalkCurrentThe user demonstrates how to adjust the frame window size and motion frame overlap to balance video smoothness, lip-sync consistency, and VRAM usage.ComfyUI· AI Lip Sync Generator
- 8:411:11Create a multi-person talking video in ComfyUIThe creator demonstrates the multi-talk version of the workflow, showing how to assign separate audio files to two different characters in a single image using the multimodel setting.ComfyUI· AI Lip Sync Generator
- Watch “ComfyUI Tutorial Series Ep 60 Infinite Talk (Audio-Driven Talking AI Characters)” →
AI Lip Sync Generator
10:070:59Lip-syncing with custom audioThe creator demonstrates importing an audio file into the timeline and using a specific prompt structure to synchronize the character's mouth movements with the audio track.What Dreams Cost
6:310:34Animate talking avatar with LTX 2.3The creator uploads the generated image and audio into the ComfyUI LTX 2.3 workflow and runs the generation to produce a lip-synced video.Prince does AI
14:523:11Lip-syncing audio to images with LTX 2.3The tutorial shows how to upload an MP3 file and a character image to generate a lip-synced video, including a demonstration of the voice isolation subgraph in ComfyUI.AIKnowledge2Go
36:000:39Singing characters with LTX-2 and custom audioThe creator demonstrates how to use LTX-2 to make a character sing by providing a custom audio file and a specific singing prompt.pixaroma
8:381:53Multi-character lip-sync workflow in InfiniteTalkThe tutorial demonstrates the multi-audio workflow, showing how to assign different audio files to specific characters (parallel vs. add types) for a multi-person scene.Atelier Darren
4:460:35Installing and using the Sync custom node in ComfyUISebastian walks through installing the Sync custom node via Git URL in ComfyUI and setting up a workflow with video/audio input nodes and the Sync generate node.Sebastian Kamph- 5:291:09Combining image-to-video models with Sync in ComfyUIThe video demonstrates a complex workflow using Krea.ai to transform a shot into a cinematic style before passing the generated video into Sync for final lip-syncing.Sebastian Kamph
- 5:411:07Configure frame window and motion settings for Infinite TalkCurrentThe user demonstrates how to adjust the frame window size and motion frame overlap to balance video smoothness, lip-sync consistency, and VRAM usage.pixaroma
- 8:411:11Create a multi-person talking video in ComfyUIThe creator demonstrates the multi-talk version of the workflow, showing how to assign separate audio files to two different characters in a single image using the multimodel setting.pixaroma
3:270:37Two-person image to video conversation in ComfyUIThe creator demonstrates how to use the 'add' setting in the ComfyUI node widget to load separate audio files for a back-and-forth conversation between two characters.ComfyUI Studio- 4:041:22Parallel audio lip-sync for two charactersThe video shows how to use the 'PAR' setting in the WAN workflow to merge audio tracks for simultaneous or perfectly timed character interactions, including masking two faces in the node.ComfyUI Studio
4:341:29Video-to-video lip sync with Infinite TalkThe creator shows how to use a vid-to-vid workflow to analyze an audio file and automatically synchronize a character's mouth and expressions in an existing video clip.Yaroflasher