Purz.ai

@PurzBeats • 18,043 subscribers

🍝 Creative & Partner Nodes at @ComfyUI 🧑‍🏫 Creative Exploration with Purz

Shorts

47,542 views

32,228 views

42,549 views

18,033 views

10,730 views

Videos

sweetdream.ai

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Private Show

Join now for exclusive access

Free preview available • Premium content

52,551 views • 1 month ago

276,175 views • 2 years ago

51,316 views • 4 months ago

17,861 views • 1 month ago

42,325 views • 4 months ago

25,078 views • 2 months ago

28,277 views • 7 months ago

14,270 views • 3 months ago

18,573 views • 4 months ago

21,916 views • 6 months ago

10,549 views • 2 months ago

24,578 views • 9 months ago

16,536 views • 6 months ago

49,496 views • 3 years ago

21,007 views • 1 year ago

13,202 views • 7 months ago

17,890 views • 11 months ago

17,139 views • 1 year ago

16,711 views • 1 year ago

14,505 views • 1 year ago

Live Cam

Purz.ai

Shorts

World models aren't the same as regular video models. The benchmarks are different. I'm really stoked to see what visionary creators do with this technology.

1. Reference Video -&gt; Convert to Depth Map 2. Depth Map -&gt; Seedance 2.0 (Reference Video) 3. Add Reference Image 4. Set Duration / Resolution / Aspect 5. Pass Audio from Reference Video -&gt; Final 6. Hit Run

This LTX 2.3 Video Outpainting LoRA by Mohamed Oumoumad is no joke. So cool!!!!

Runway Aleph 2.0 is now in Comfy via Partner Nodes 'The band is playing underwater' I was most impressed by the hair and bubbles!

Image Filter Live does videos now, realtime playback for dialing in fx, and then it renders on next run.

Videos

Watch Anya Live

Astra + Blender

With LivePortrait you can just perform what you want the character on the screen to do. This clip is using cc0 footage from Pexels, but you can do this with all the AnimateDiff / Gen3 / Luma / Pika etc clips you've already made as well!

Ran SCAIL-2 through a bunch of tests, it's extremely prompt-adherent, the motion capture is extremely impressive. Definitely recommend playing with this if you're looking to do motion transfer into a reference image.

I've opened Unreal Engine about seven times in my life and got scared six of them. This episode I wire three MCP servers — Unreal Engine, Comfy and Blender — into a single coding agent with Astra and let it build a game while I just describe what I want.

Spent $20 on a tiny 1.5" ESP32-S3 TFT LCD board, plugged it in, and Claude basically knew the whole chip immediately. Now it's running a falling-sand sim with gusty wind, curated color palettes, and a different personality every boot — at 91 fps. 🏜️

Flux 3 is pretty great.

Claude → CRT. Built a procedural sci-fi interface generator in Three.js + TypeScript that outputs animated HUD compositions from modular elements—radar sweeps, waveforms, data cascades, ring gauges—arranged via BSP-subdivided layouts.

People forget Pika were one of the first model companies to get reference images and ingredients working in a simplistic interface. This is an oldie but a goodie.

This is crazy performant wow

Don't mind the weird cuts, still working on the workflow... this is nuts!

be me have access to every sota model around make 80s slop nonstop

How to Actually Control Style with LTX-2 LoRAs

A bunch of LTX 2.3 Video Outpainting tests. It's not perfect, but this still feels like illegal technology.

It's been a crazy couple of days, trying all sorts of things with ComfyUI + AnimateDiff-Evolved and various ControlNets. I'm absolutely blown away by the consistency and clarity of the animations using this method. Workflow below 👇

My Sora cameo has been public for about 24 hours now. These aren't even the really weird ones. Also, I see all your drafts. 🤣🤣🤣

Claude got busy simulating the SNES PPU at the hardware level then corrupting it just to give lil' old me some nostalgic SNES glitches for my CRT display.

🏃‍♂️‍Wan ATI with ComfyUI using a new Path Editor by Machine Delusions that can make spline animations with a user friendly interface. 🧑‍🏫 Check out his tutorial, linked below!

This was state of the art 3 years ago. In September 2022, I was running Deforum non-stop on any free compute I could find. It really felt like piloting a little spaceship through latent space, plumbing the depths of the models.

InfiniteTalk: Audio-driven Video Generation for Sparse-Frame Video Dubbing

Wan2.2-Animate feels like something very very fresh. Context windows allow for extended runs well past the normal 81 frame threshold....

1. Reference Video -> Convert to Depth Map 2. Depth Map -> Seedance 2.0 (Reference Video) 3. Add Reference Image 4. Set Duration / Resolution / Aspect 5. Pass Audio from Reference Video -> Final 6. Hit Run