ganlive-amber

ganlive-amber is a 1536×1024 FastGAN trained on 5,469 of my photographs. Play it in your browser with WebGPU, by hand, with MIDI, or from drums. Nothing to install: it downloads once and runs on your graphics card.

ganlive

ganlive explores a GAN's latent space in real time, with dials derived automatically and played by hand, with MIDI, or from drums. It plays FastGAN and StyleGAN2 models.

Use

Open the player in a browser with WebGPU, such as a current Chrome, Edge, Firefox or Safari, and choose amber. It plays engine/program.json and engine/weights.bin from this repository, which ganlive convert wrote from amber.pt.

To play it in ganlive's desktop app instead, install it as its user guide describes, then:

uv run --no-sync hf download gustavecortal/ganlive-amber --local-dir runs/amber
uv run --no-sync ganlive play --checkpoint runs/amber --console --no-audio --no-midi

Model

• Architecture: FastGAN with a 256-wide latent, output at 1536×1024 (3:2).
• Training: gantrain.
• Speed: on a $350 Intel Arc A770, 150 fps in Edge, and 177 fps in the desktop app.
• Sibling: ganlive-lichen, trained on the same photographs at 3072×2048.

Training data

5,469 of my own photographs of places, objects and textures.

License

MIT