fal Assets, KREA 2, Recraft V4.1 & Generative Media Surge
Welcome back. A lot happened in the last 48 hours across generative media — new model drops, infrastructure updates, and a few things that genuinely change how creators work. Let's get into it.
Starting with fal — they launched something called fal Assets on May 29th. It's a unified library that stores everything you generate: images, video, audio, 3D assets — all in one place, with semantic search and instant reuse across their Sandbox, Playground, and Workflows. This is the kind of unglamorous infrastructure work that actually matters. No more hunting down that image you generated three sessions ago. It just lives there, searchable, ready to drop back into any workflow.
Also on fal — FLUX Virtual Try-On landed May 28th, straight from Black Forest Labs. You give it a person reference and a clothing reference — and it puts the outfit on the person while preserving their face, pose, and identity. For fashion and e-commerce pipelines, that's a big deal.
Now, KREA 2 — this one's worth paying attention to. It dropped May 28th and it's their first model trained from scratch. Not a fine-tune, not a wrapper — built ground-up to balance aesthetic quality with creative control. It's specifically designed for moodboard and brand pipelines, where you need consistency and style fidelity across a whole project. Same day it launched, it was already live on Replicate in medium and large variants — and as a ComfyUI partner node, where you can tune creativity, feed in style references, and use moodboard conditioning directly. Oh — and it's already integrated into Hermes Agent, which is a workflow-level move worth noting.
Speaking of image generation — Recraft dropped V4.1 on May 29th. The focus here is fluid, luminous gradients — the kind that work for abstract and dreamy scenes — plus improved realistic product mockups across packaging, signage, and apparel. They're also claiming the number one spot among independent image generation labs on public benchmarks, just two weeks after launch. Bold claim. But they're putting numbers behind it.
Moving over to video — the Runway API got a serious update on May 29th. They pushed in Seedance 2.0, GPT Image 2, HappyHorse 1.0, Nano Banana Pro, and Magnific Precision Upscaler V2, among others. That's a lot of models landing in one API update. If you're building on Runway's infrastructure, you've got a lot of new toys to play with right now.
And while we're on Seedance 2.0 — ComfyUI dropped new transformation video templates using it. We're talking cube-to-person and sphere-to-person transitions with consistent style held throughout. Visually, those demos are striking. ComfyUI also added OpenRouter support on May 29th, which means 20-plus LLMs are now accessible directly inside your ComfyUI workflows. That's a meaningful bridge between visual generation pipelines and language model capabilities.
Project Luxo released two AI-generated short films over May 28th and 29th — "Last Night" and "The Rogue." Both single-person productions. The trend of solo creators shipping finished short-form narrative content with AI tooling keeps accelerating.
Two more quick ones — Magnific launched 3D Scenes on Marble on May 28th. Single image in, navigable 3D environment out. It's early days for this kind of thing but the direction is clear.
And Replicate added P-Video-Animate on May 28th — it takes a reference image plus a reference video and generates new video that preserves the motion and style of the reference. It's running at a promotional rate of $0.009 per second right now if you want to experiment.
That's your AI digest for 30 May 2026.