First drop of the challenge. I'm building one Gen Effect node a day, and a week in I already have to correct myself on something I said at the start. I underestimated the range. These nodes go from dead practical to full VFX, and everything in between. On the practical end, I built a Pro Photo Retoucher. One node, drop an image in, cleaner result out, with a compare slider to check it. Boring on paper, but it earns its place every day. On the other end it gets wild. An Inception Fold that folds an image into itself, and the same node works on video, so your footage folds while it plays. A pose transfer where a stick figure drives the pose of a model. Metal emboss, stripe fills, iridescent crystal looks. Two favorites so far: the Inception Fold, because it treats video and stills the same way and that opens a lot of doors, and the stick figure pose transfer, because it's the clearest example of Gen Effect nodes working together instead of alone. And that's the real lesson from week one. These nodes rarely stand on their own. They get powerful when you layer them through a compositor and stack them. That's where it stops being a filter and starts being a workflow. 👇 Which of these would you actually use? And the challenge is still open, if you've got an effect you want built, drop it and I'll make it with my credits.

Quick one on the coffee calls, since we're doing them regularly now. It's a live, casual session. No agenda, no slides. Just me and whoever shows up, talking Weave and AI out loud. Bring a question, a workflow you're stuck on, or just your coffee and listen in. It's the most direct way to get help and to actually meet the people in here instead of only reading the feed. Every call gets added to the community calendar, so that's where you'll always find the next date and time. I'll also drop a reminder here the day before each one. No sign-up, no pressure. Just show up when you can. See you on the next one.

OpenAI released ChatGPT Images 2.5 yesterday. Short rundown of what's in it, because a few things are relevant if you build workflows. What they claim: - Better subject preservation from reference photos, more natural lighting and texture - Precision editing: change one element without the rest drifting - Multi-turn consistency: earlier edits stay intact instead of degrading over a long chain - Up to 50% lower latency vs Images 2.0 - Sketch: draw a rough guide directly in ChatGPT and generate from it - Templates for common formats, plus shareable prompts For API and pipeline work there are two models: GPT-Image-2.5 Flare (the fast default) and GPT-Image-2.5 Sunburst (slower, built for tighter control across edits). I have not run it yet, so treat all of the above as their claims and not my verdict. Anybody played with it?

I got inspired by a post here about creating a pose puppet rig to control your model. I wanted to try and see if it would be possible to create a gen effect that could control the emotion or facial expression. It kind of got there where you can control emotions based on presets or manually adjust sliders. I would say its 60% there but you would kind of need a controllable 3D model instead of 2D pivot points. Anyhow, its fun to test around with the possibilities and see what works and what doesn't. Maybe someone has a suggestion how to fix it!

Got asked if the adjustable pose skeleton from the other day was a native Weave node. It isn't. I built it with Gen Effect, and here's how you make your own. That's the whole point of this challenge. Once you understand Gen Effect, you stop looking for the node that does the thing. You build the node that does the thing. This one gives you a fully adjustable pose skeleton you can drive a model's pose with. Here's the exact prompt: ADJUSTABLE POSE SKELETON — Gen Effect prompt Draws an editable pose rig over a connected image, a draggable dot at each joint. All 14 are individually draggable, each with its own X and Y slider; never collapse the rig to a few control points or make limbs follow other joints. Show Bones links them with openpose lines; off leaves just the dots. Show Head, Show Arms, Show Torso and Show Legs turn each section on or off, so you build only the part that fits your framing (legs off for upper body. Size the figure by dragging the joints. Joints (each with an X and Y, percent of frame): Head, Neck, L Shoulder, R Shoulder, L Elbow, R Elbow, L Wrist, R Wrist, L Hip, R Hip, L Knee, R Knee, L Ankle, R Ankle. When Show Bones is on: Head-Neck, Neck to Shoulders, Shoulder-Elbow-Wrist, Neck to Hips, Hip-Hip, Hip-Knee-Ankle. Y is 0 at top, 100 at bottom; Head near top. Flip Vertical mirrors top-to-bottom. Joint Size sets the dots, Bone Thickness the lines, Colored Limbs the openpose colours else Skeleton Color. Show Input overlays on the photo, else Background Color. Do not analyse the pose. Use toggles, not dropdowns. Build exactly these controls, names, ranges and defaults. Do not add or rename. JSON wins. { "effect":"Adjustable Pose Skeleton","version":7, "input":{"source":"node_input","usage":"Image shown under the skeleton overlay; not analysed."}, "output":{"alpha":true}, "controls":[ {"name":"Show Bones","type":"bool","default":false}, {"name":"Flip Vertical","type":"bool","default":false}, {"name":"Show Head","type":"bool","default":true},