Introduction
This project specifically adds the functionality of start and end frame guided video generation.
The nodes support Wan2.1 models in both 720P and 480P versions. It is recommended to generate videos with a frame count of 25 or higher, as a lower frame count may affect the consistency of character identity.
Currently, the start and end frame video generation approach is in its early stages. It primarily implements the start and end frame video generation functionality at the code level and does not yet involve model or LoRA fine-tuning, which is planned for future work. Additionally, incorporating end frame guidance in Image-to-Video (I2V) seems to degrade video generation quality, which is another area for future improvement.
Note: Video generation should ideally be accompanied by positive prompts. Currently, the absence of positive prompts can result in severe video distortion.
https://github.com/raindrop313/ComfyUI-WanVideoStartEndFrames
Recommended machine:Ultra-PRO
Workflow Overview
Step 1: Upload the picture and Input prompt word
1.Upload the first frame of the picture.
2.Upload the second frame of the picture.
3.Input prompt word.
Step 2: Improve video quality and reduce the time required to generate videos
1.TeaCache could be considered to be sort of an automated step skipper
The relative l1 threshold -value determines how aggressive this is, higher values are faster but quality suffers more. Very first steps should NEVER be skipped with this model or it kills the motion. When using the pre-calculated coefficients, the treshold value should be much higher than with the default coefficients.
2.Enhance-a-video can increase the fidelity of the results, too high values lead to noisy results.
Step 3: Adjust the video resolution and modify the video length
1.When I was testing, the effect of generating a two-dimensional video was very good when step=30, and the real person would have a bad face; when step=50, the real person's facial texture gradually became clear, and there was a small probability of drawing a card.
2.Adjust the "generation_width" and "generation_height" parameters to change the video resolution.
3.The original code had automatic resolution adjustment based on input image total pixels and aspect ratio. If you want to set it manually, disable the adjust_resolution.
4.You can change the video length by setting frame_rate or num_frames (in WanVideo Empty Embeds). Video length = num_frames/frame_rate.