Turn any video into another, restyled video

Create new versions of a video using your description of the final video.

AI Video to Video Tool

Advanced

Tip: Pick one close to the aspect ratio of your input video.

Aspect ratio

# of in-between frames that will not be directly diffused

Cadence

The amount of the input video to use. 0 means use the whole duration, after the 'clip start seconds' above.

Tip: Use 2 to test ideas quickly and save credits.

Clip duration seconds

Seek this many seconds into the video before processing.

Trim start seconds

More weight means more match to the edges of the input video, and more stability between frames.

Tip: Double click to add more control points. Pinch to scale y-axis.

Controlnet model #1 (Edge detector) weight schedule

More weight means more match to the poses of any people in the input video, and more stability between frames.

Controlnet model #2 (Human pose detector) weight schedule

Generate an extra side-by-side video

Hide Watermark

Choose video

Seed

More noise means more change from the original image.

Noise multiplier schedule

Sampler steps

How much the previous frame should influence the next frame. At 1.0, there will be almost no change from the original video. Lower values will be more creative, but less smooth between frames. Note: an x-axis value of 100, on the right, stands for the end of the video.

Previous frame strength

Generate privately

Tip: Add as many prompts as you like. The video will interpolate between them. Just one works as well.

Describe your output video (prompts)

Sampler

Stable Diffusion Model

Enable if you want to test ideas on a short snippet of video.

Trim video

Video to Video

Create Video

Results

Answers to: Video to video: mikasa, anime girl, attack on titan, ufo background, mikasa, anime girl, attack on titan, ufo background, mikasa, anime girl, attack on titan, paradise background, gorgeous skies

Results

Request another