Open source

We publish the tools, models and datasets we build for our own work and for our courses. Each one is free to use under its licence.

Why it matters to us

Artists keep control when they can run, read and change their tools. Open models and ComfyUI make that possible, so we build on them and give back what we make.

Our projects

ainvfx-fluid

IC-LoRA for LTX 2.5

Turns painted keyframes into fire, smoke, explosions, steam, liquids and dust with LTX 2.5. We trained it only on synthetic clips that we generated and publish.

Active

Its training set holds 176 synthetic scenes, free to reuse under CC BY 4.0.

LTX-2.x Community License, the licence of its base model.

SeedVR2 for ComfyUI

ComfyUI custom node

Upscales and restores video and images with the SeedVR2 model, on GPUs with little memory and with alpha channels. We co-developed it with NumZ before ComfyUI added native support.

No longer maintained

ComfyUI now ships its own SeedVR2 nodes, so we no longer maintain this one. Our model files stay online.

Apache 2.0

ainvfx-runpod

Runpod toolkit

Starts a Runpod GPU machine with ComfyUI and the models a course needs, in one command. It also trains LoRAs. We use it in our courses.

Active

Apache 2.0

Support our open source work

Building, testing and documenting these projects takes time. If they help your work, a donation helps us keep them free and up to date.

Make a donation