I've been experimenting with different animation styles and this one stuck. Flat colors, thin black outlines, fine hatching, paper grain over everything. Closest thing I can compare it to is old Moebius desert illustration style. Workflow was simple: Reference image → Seedance 2.0 → short animated scene Want to see more styles like this? Drop a style you'd like me to try 👇

I created a cool image that turned my picture into a LEGO-style look 🧱👀 Now give it a try! Take any photo and see if you can transform yourself into a LEGO character while keeping the original scene and details. I used Seedream 4.5 for this one. Drop your original + LEGO version below 👇 Let’s see how different models handle the LEGO transformation 🤨

I recorded a real POV video on my iPhone 16 where I walked up to my car, got inside, started it, and pressed the accelerator. Then I created multiple reference images of the same baby blue Lamborghini Revuelto: - exterior - driver side - open door - interior/dashboard I uploaded the POV video and all references into Seedance 2.5 and used the original footage as the master for the exact camera movement, timing, hand movement, body movement, and environment. Seedance only replaced the original car with the Lamborghini while keeping the full POV sequence intact. iPhone 16 POV → Car references → Seedance 2.5 → Final supercar POV A simple way to turn real footage into realistic AI car edits without generating the camera movement from scratch. 🔥

For this test I used our GPT called Behind the scenes to build the miniature city, giant studio set and matched animation prompt. Then I customized the concept by adding Frieza and changing the action so he holds a Death Ball above the city, throws it down, and the impact creates a lava crater that spreads through the miniature streets. The nice part is that you can do the same with any other character — just tell the GPT which character you want to use and describe the action you want them to perform. So basically: Behind the scenes GPT → choose your character → customize the action → final image → matched animation prompt → finished VFX shot. A fun way to use our GPT as the base and then push the scene in your own direction.

This prompt builds a miniature scale model of any city as an editorial poster. No reference image needed, just the name of your city. The only variable in this prompt is the city name. The tile, the camera angle, the lighting and the typography are fixed in the prompt, so every city comes out in the same style. GPT Image 2 fills in the street grid, the architecture and the landmarks. Prompt is pinned in the comments. Replace [CITY] and run it. Run it on your city and drop it below 👇 Let's see how many we can collect.