ConsistI2V

ConsistI2V: Enhancing Visual Consistency for Image-to-Video Generation [TMLR 2024]

★ 26114 forksPythonMITTIGER-AI-Lab

Official code for ConsistI2V (TMLR 2024), a diffusion method that turns a still image plus text prompt into a consistent video using first-frame spatiotemporal attention and low-frequency noise initialization. Includes inference and training scripts, a Gradio demo, and Hugging Face weights.

Use Cases

Animate a single still image into a short video clip from a text promptGenerate visually consistent image-to-video results for creative and film previsualizationRun local Gradio demo for I2V generationDeploy I2V generation on Replicate or Hugging Face SpacesFine-tune the diffusion I2V model on a custom video-caption datasetBenchmark I2V consistency with the I2V-Bench datasetResearch spatiotemporal attention and noise initialization techniques

Built With

Language: Python
Frameworks: PyTorch · Hugging Face Diffusers · AnimateDiff · FreeInit · Gradio · Cog · Weights & Biases

ConsistI2V

Use Cases

Built With

Tags