Replicate goo logo generator

This Cog model generates vibrant, animated or still "goo" backgrounds and overlays a logo in the center. It can output a single image (PNG, JPEG, TIFF) or a looping MP4 video, with a variety of customization options for the goo effect.

output.mp4

See also: https://github.com/replicate/goo

What does it do?

• Generates a colorful, animated "goo" effect using a ModernGL shader.
• Overlays a logo (provided in the model) at the center of the output.
• Outputs either a static image or a video, depending on your choice.

Inputs

You can control the output with the following options:

Name	Type	Default	Description
seed	int	-1	Seed for the random number generator. Use -1 for a random seed.
width	int	512	Width of the output image/video (1–4096).
height	int	512	Height of the output image/video (1–4096).
scale	int	1	Scale factor for the goo effect. Higher values zoom in/out on the goo pattern.
depth	int	3	Number of iterations for the goo effect. Higher values create more complex patterns.
format	str	"png"	Output format: "png", "jpeg", "tiff", or "mp4" (for video).
speed	float	2.0	Speed of the goo animation (0.0–10.0). Higher is faster.
num_frames	int	600	Number of frames for video output (only used when format is "mp4").
fps	int	60	Frames per second for video output (only used when format is "mp4").

Example usage

Generate a PNG image

cog predict -i format=png -i width=1024 -i height=1024 -i seed=42

Generate a JPEG image with a different goo pattern

cog predict -i format=jpeg -i scale=2 -i depth=5 -i seed=1234

Generate a 5-second MP4 video at 30 FPS

cog predict -i format=mp4 -i num_frames=150 -i fps=30 -i speed=5.0

The output will be written to a file, e.g. output.png or output.mp4.

Output

• Image formats: PNG, JPEG, TIFF
• Video format: MP4 (H.264, browser-compatible)

All outputs have the goo effect as the background, with the logo centered and sized to 50% of the smallest image dimension.

How it works

• Uses a ModernGL shader to generate the goo effect.
• Overlays a logo (decoded from base64) using Pillow.
• For videos, frames are streamed to FFmpeg for efficient MP4 encoding.

Requirements

• Cog installed (see below).
• Docker (Cog uses Docker to run the model).

Installation

Install Cog by following the official instructions:

brew install cog
# or
sh <(curl -fsSL https://cog.run/install.sh)

Running the model

After installing Cog and Docker, run predictions as shown above. You can also serve the model as an API:

cog serve -p 8080

Then POST to http://localhost:8080/predictions with your desired input options.

Customization

• Logo: The logo is embedded in the model as a base64 PNG. To use a different logo, you’ll need to modify the model code.
• Goo effect: Tweak scale, depth, speed, and seed for different visual results.

Troubleshooting

• If you get errors about OpenGL or EGL, ensure your Docker environment supports headless OpenGL rendering.
• For video output, FFmpeg must be available in the container (the provided cog.yaml should handle this).

License

MIT (or specify your license here).

---

Questions?
Join the Replicate Discord or open an issue!

---

Let me know if you want to include example images, API schema, or further details!