Learn

How to run the workflows behind image and video generation, model by model — plus what's moving in the wider open-source model landscape. Grounded in real work, not theory.

Animating Stills in Practice: Wan 2.2 Image-to-Video

A field guide to turning a single still image into a short, coherent video clip with Wan 2.2 I2V — what each setting does, and where it breaks.

Local Lip-Sync on a 12GB GPU: InfiniteTalk with GGUF

You don't need a cloud API to make a character talk. Here's how we run InfiniteTalk lip-sync locally on consumer VRAM with GGUF quantization — at zero cost per shot.

MiniMax-H3: What the Model Everyone's Running Actually Requires

MiniMax-H3 generates synchronized audio and video natively, and it's the open video model most of the community is standing on right now. Here's the honest VRAM math, and why FlixML isn't shipping it yet.

Multi-User ComfyUI: How to Keep Agents From Seeing Each Other's Work

A login protects your generation server but still shows every caller everything inside it. What it takes to give each agent and person their own view: keys that scope what gets read, not just what gets made.

What Is Quantization? How Big AI Models Run on a Consumer PC

Quantization shrinks image and video models so they fit in your computer's memory. What it does, what it costs, and how to pick a version for your graphics card and RAM.

Why We Built FlixML: Agent-Driven ComfyUI

ComfyUI can run almost anything and makes you wire every graph by hand. FlixML puts an agent-native API in front of it so you describe what you want and your agent builds the graph.