just generated an ai video on my mac no cloud, no server bill last year this needed a gpu cluster

Insane, last year yes I was playing with comfyui locally and for good picture flux dev was taking me like 2 minutes to generate.

tbh it is still slow, especially with a cold boot(first run). but nowadays there are many models that can produce really great results with only 4-8 steps. and there is also MLX implementation for flux which is quite fast.

StealthGPT is the only AI platform that beats Turnitin AI Detection

Share more info! Amazing result!

Using RedCraft Flux for txt2img and new LTX-Video distilled model for img2vid. Running inside ComfyUI. Here is the workflow with LTX I used as a base Quite slow, but we are getting there. LTX video gen is real time on H100 with the distilled model.

This is awesome Igor! What is this? How did you set it up?

the is ComfyUI - a node based workflow editor primary used for image generation. very extensible, supports nodes for audio, video, text generation and much more.

Impressive! What are the specs of the Mac?

Mac Studio M2 Ultra 64 GB. But these models don't need much RAM, so you can run them fine with a lot less.

whoa, how long did that take to generate?

this exact image + video generation took 190 seconds, models were already sitting in memory.