Skip to content

Latest commit

 

History

History
21 lines (17 loc) · 859 Bytes

README.md

File metadata and controls

21 lines (17 loc) · 859 Bytes

jwst-img-to-vid

Python scripts to download JWST observations and their associated metadata. For novelty research purposes, we create free animations or videos from deep neural networks.

Stable Video Diffusion XL Parameters

The following parameters produced 5 second videos on an A100 GPU after pre-editing the photos for a 576 H x 1024 W pixel image.

  • Frames: 50
  • Frames per second: 10
  • Number of Steps: 60
  • Seed: Random
  • Number of frames decoded at a time: 1
  • Motion bucket id: 200
  • Condition augmentation factor: 0.02

The current notebook supports frames that are in multiples of 25. Length in time is Frames / Frames per second (5 seconds = 50 frames / 10 fps). Increase in Condition augmentation factor increases how much can change in the output video from changing it for the same input as Motion bucket id is the rate of change.