ComfyUI

ComfyUI just added OpenRouter support. Instead of being locked into a single LLM, you can now access 20+ models directly inside ComfyUI. More flexibility, less friction, same workflow. Links to the workflows 👇 Cloud: https://lnkd.in/gPK7AXpN Local: https://lnkd.in/gwEzQ_mg

Create transformation videos where a cube, sphere, capsule, or custom object transforms into a person or item with a button press using Seedance 2.0 inside ComfyUI. The workflow automatically generates matching first and last frames while keeping the environment and style consistent throughout the effect. Try the workflow, link below. https://lnkd.in/dSzNxmaG

The open source community has been delivering on LTX 2.3 LoRAs. Fine tuning LTX 2.3 unlocks control that you can't get from closed models.  Here are 7 LoRAs that can save your footage.  There are too many incredible LoRAs to cover, so comment below your favorite that needs a showcase! Links to all the workflows below 👇 Lipdub (by Lightricks) Input a text prompt dialogue to dub/translate a characters speech. https://lnkd.in/g3Q_VURk Obscura Remova (by WepeNerd) Used as occlusion removal, haze cleanup, foreground-object removal, and temporal scene reconstruction. https://lnkd.in/gVgM5JkQ Head Swap (by Alissonerdx) High-fidelity video head swap with natural tone blending and consistent lighting. https://lnkd.in/gV8ARH3E Outpaint (by oumoumad) Extend the canvas of an input video by generating new content. https://lnkd.in/gwe3HuYu Refocus (by oumoumad) Turn out-of-focus footage back into sharp, in-focus video. https://lnkd.in/g-RWNGHc Dearchive (by oumoumad) Input archive video and generates the footage as if shot more recently. https://lnkd.in/g8sTamRj Remove Subtitles (by JoyFox) Removes subtitles, overlays, and text occlusions from videos. https://lnkd.in/gJCQgsqm

Krea 2 Image is now a Partner Node in ComfyUI Krea's first foundation image model — trained from scratch — with tunable creativity, style references, and moodboard conditioning. → Creativity dial (Raw → High) to trade prompt adherence for aesthetic depth → Style reference image input → Moodboard conditioning for consistent art direction across a body of work Krea 2 Large + Medium, both live now.

Every change ComfyUI ships to 2.5 million monthly creators passes through a system Hunter built in his first week here. He joined as a contractor before our cloud public beta. His trial week was a preview environment system that provisions full-stack staging on production hardware. Every database migration, every new endpoint, every change to our job dispatch system and inference engine gets tested end-to-end before it ships. We've provisioned thousands of those environments since, and every feature has passed through one. Hunter had quit Coinbase and spent almost a year between roles before he applied. He flew out for the launch sprint and stayed. A few months in, he shipped our Free Tier and grew daily active users nearly 10x. He led the migration from A100s to RTX Pro 6000s with autoscaling that brings hundreds of GPU VMs online and back offline every day. Hunter shipped first and let the spec catch up. If that's how you work, we're hiring: comfy.org/careers

One engineer's GitHub repo with 331 stars is the only bridge between ComfyUI and Nuke, the VFX compositor every major studio uses. The handle is vinavfx, and the repo is ComfyUI-for-Nuke. It does what the name says: it lets you call ComfyUI workflows from inside Nuke's node graph. Pixel buffers go in, the workflow runs on a ComfyUI server, and the output comes back as a Nuke node ready to composite into a shot. That bridge used to be a multi-week engineering project. Studios would build it once, keep it proprietary, and rebuild it every time either tool updated. Now it's an open-source plugin maintained on GitHub. A VFX house running ComfyUI through Nuke is no longer waiting on either company to ship the integration. The community already did. What I keep noticing in this corner of the platform: the bridges between open-source AI and the tools real studios use are getting built by one person at a time, because they need them.

Comfy will be at AI on the Lot. Our CEO Yoland Y. is joining a panel of AI video leaders for a progress report on the state of video models - where they excel, where they fall short, and what's coming next. AI On The Lot → Culver City, Ca The largest AI media conference bringing together artists, innovators, filmmakers, media execs, startups, creators, and developers. Wednesday May 27th | 3:00 – 3:45 PM | Theater 1 → AI Video Models Progress Report: AI Leaders on the Next Big Hurdles Alongside Mahmoud Ellithy, Charles Migos, Erik Barmack, Morgan Suo If you're attending, Stop by.

Groove Jones needed to deliver the impossible: giant NFL-licensed Crocs parachuting into Dick's Sporting Goods parking lots; hyper-realistic, multi-location, vertical 9:16 on a holiday deadline. A traditional CG pipeline wasn't an option. So they built a new one inside ComfyUI. Their Senior Creative Technologist Doug Hogan architected a single-graph workflow that orchestrated Flux, Gemini Flash, Veo 3.1, and Moonvalley's Marey. With custom LoRAs trained on Crocs NFL product and DSG storefronts ensuring brand accuracy across every shot. Client iterations that once took weeks happened same-day. Final compositing in Nuke brought it to 2K for Reels, TikTok, and Shorts. The campaign delivered on time. It also won the Aaron Award for Best AI Workflow for Production. "VFX projects used to operate at razor thin margins. Comfy solved that for us." - Dale Carman, Co-founder @ Groove Jones This is what a modern production pipeline looks like. Learn more about the project: https://lnkd.in/gh2fATbF