Runway weren't kidding about the game company thing..

🚨OPEN SOURCED WEIGHTS NOW AVAILABLE FOR FLUX KONTEXT [DEV] Code already implemented for ComfyUI - expect it to be up and running within the next couple hours here. LINK TO WEIGHTS BELOW 👇

"Change the brown jacket to blue." (OmniGen2)

I can’t explain but I’m testing something new.

Just tested Hunyuan3D-V3.0 and my impression is that it's very good. Took a character we created in Midjourney → generated 4 T-pose angles in nano-banana → fed into Hunyuan3D-V3.0 → got a 3D model ready for rigging in OBJ, FBX & GLB in a few minutes. This will be great for

Luma has launched Ray3. I am particularly excited about this: Studio-grade HDR: Generate video in 10, 12, and 16-bit High Dynamic Range with accurate details in shadows and vivid color and bring your generations as EXR frames into post-production, editing and color-grading

Interesting workflow test: took a storyboard illustration and converted it to photorealistic previsualization using a multi-step process. Process: Nano Banana converts illustration to photorealistic frame → Flux LoRA/Ideogram/Seedream fixes character identity in frame → VEED

Why can't I get any of my Wan 2.5 generations to come out with an American accent?

I took this interview shot and used a 3D pose reference to generate the lying position while maintaining the same character and environment. Used Nano Banana with the prompt: "Match this man exactly to the pose in the reference image. Maintain the original couch and room that the

I fed a random Nano Banana-generated storyboard page into Sora 2 with just "Cinematic Hollywood action sequence" as the prompt. The results are amazing. This is only 12 seconds. What happens when we get longer generations, seed control, and multiple runs? Is this what Veo 3.1 is

I entered this: A stylish commercial selling a high end dreadnought acoustic guitar made by Fowler guitars. The head of the guitar features the name Fowler. And I got this video. It's not perfect. The strings are wonky in one shot, the music needs to be improved, and I don't

It’s staggering how fast AI video is moving. Here’s a split-screen of two versions of the same scene I've been testing. Bottom Video (From the beginning of the year): As excited as I was about this, it was a constant battle. I couldn't get the exact car I wanted, the hands kept