A probable pigeon.
24 s
Production
- Images
- ChatGPT
- Upscaling
- LapSRN
- Motion
- Wan 2.2 + Lightx2v
- Edit
- Remotion + FFmpeg
Scenes
Sixty pixels.
Start with a 6 × 10 input. Use the same source image throughout the real upscaling sequence.
Settings
- Input
- 6 × 10
- Pixels
- 60
The neutral source image contains no raised middle finger.
ComfyUI
The reveal uses Wan 2.2 image-to-video with Lightx2v 4-step LoRA in Comfy Cloud.
Use the neutral full-body image as the start frame and the raised-hand pose as the end frame. Set 33 frames at 16 fps, 4 steps, CFG 1 and seed 151515.
Keep the real LapSRN outputs separate from the generated reveal. A cleaned, verified workflow download is not available yet.
Editing
Add slight blur, grain and lower-resolution texture to the generated ending. Use a 0.6-second bridge to reduce the jump in sharpness.
By agi.geek · Updated 2026-10-19.
Prompts are the originals used in production.
Workshops & production enquiries · All works