ComfyUI v0.29.0 Dropped & It's Absolutely Unhinged

Bestie, if you blinked, you missed it. ComfyUI just dropped v0.29.0 — and no, this isn't your grandpa's "oh look, another bugfix release." This is the kind of update that makes you cancel your plans, grab an energy drink, and start wiring nodes like a caffeinated wizard.

Let's break down the chaos, shall we?

Video Workflows Just Got a MASSIVE Buff

Remember when ComfyUI would yeet your entire video into RAM and your poor GPU fans sounded like a jet engine taking off? Yeah, that's over. The headline fix in v0.29.0 streams video transcodes instead of buffering every single frame in memory. Translation? You can now work with longer videos without watching your RAM bar scream for mercy. This is the kind of quality-of-life improvement that makes you want to kiss the devs on the forehead.

In internet terms: we went from "loading... please don't crash... I'm begging you..." to "we ball."

Partner Nodes Went Absolutely Ham

This release is basically the Avengers: Endgame of partner node integrations. Here's what got added:

• OpenAI GPT-5.6 — OpenAI's latest brain child is now cookin' in your ComfyUI workflows. Generate, analyze, vibe. Whatever.
• Gemini 3.5 Flash LLM — Google's speed demon is here and it's READY.
• Claude Opus 5 — Anthropic said "hold my beer" and dropped their most powerful model into the node graph.
• OpenRouter new models — Because one API to rule them all wasn't enough, now there's MORE.
• HeyGen nodes — AVATARS. TALKING PHOTOS. VIDEO TRANSLATE. You can now deepfake your face into a presentation without leaving ComfyUI. (Please use this power responsibly. Or don't. We're not your mom.)
• ByteDance audio model — New multilingual audio model for all your sound needs.
• Recraft V4.1 — With V4 image editing nodes. Yes, you can now edit images inside ComfyUI like it's nothing.

Honestly, the Partner Nodes program is starting to look like the Marvel post-credits scene that never ends. Every week there's a new cameo. And we're HERE for it.

Gemma4 & MageFlow Walk Into a Bar

ComfyUI v0.29.0 also adds support for Gemma4 12B — Google's open-weight model that's been turning heads in the LLM world. If you've been wanting to run a capable language model inside your image generation pipeline, baby, now you can.

Oh, and MageFlow support? Yeah, that's in there too. Because apparently the devs looked at the feature request board and said "let's just do ALL of it."

FreSca 5D+ Got Fixed (No More Weird Artifacts)

The FreSca 5D+ (formerly Anima) fix is model-agnostic now. If you're into advanced latent manipulation — and let's be real, if you're reading this you probably are — this one's for you. Fewer weird artifacts, smoother iterations, more "hell yeah" per second.

The Verdict?

ComfyUI v0.29.0 is not just an update. It's a statement. The Comfy team is moving faster than ever, partner integrations are popping off like fireworks, and the core improvements (hello, video streaming!) show they actually listen to what the community is screaming about on GitHub.

If you haven't updated yet: what are you even doing? Go run git pull and update your manager. Your workflow is about to get a whole lot spicier.

Now if you'll excuse me, I have some talking photos to generate and absolutely no adult supervision.