Explore a moment in motion.
One movement.
Every moment, visible.
Bring in a video to extract frames,
inspect subjects and build your sequence.
Every pose. One artwork.
A movement, spread across the picture.
Select at least three isolated moments below, then build your layout.
Video moments → complete AI artwork
Bring the action to life
Static poster accents inspired by the action. Switch off to see the original.
Build an artwork to try the effects.
AI artwork finish
Improve detail and blend the complete current artwork in one GPU pass. No separate pose or background rebuilding is needed.
Build an artwork first, then check the model connection.
Finish options
One GPU pass rebuilds detail in the complete image. Inspect poses, clothing and equipment against the preserved previous artwork. Your isolated layers are retained.
Shape the sequence
Overflow lets subjects cross their divider. Edge fade softly blends their outer edges; rebuild the layout to apply it.
Vertical at 0° · diagonal at other angles. Your selected moments supply the poses.
Connect Nano Banana · ComfyUI credits
Use a ComfyUI account API key from ComfyUI Platform → API Keys. This is separate from the server login and uses your existing Comfy credits.
Checking the local connection…
The key is saved in a protected local settings file. It stays outside experiment exports. Saving does not generate an image or spend credits.
Finish the real action
Original silhouette preservedEnhance the visible cutout detail with Nano Banana, or choose local cleanup. Enlarge the figures, match their finish to the scenery and add soft ground shadows. Strips and typography stay precise.
Cutout references are sent to Comfy and Google. Existing transparency is retained. AI detail can alter numbers, logos and other fine details; transparent holes and missing equipment are not repaired.
Only the scenery is sent to the configured ComfyUI server. Inspect generated field markings and perspective.
Cutout cleanup preserves the real image and runs on the lab server. Enlargement cannot recover detail absent from the video.
Build a layout guide to calculate the paid image requests.
Build a layout guide first.
Generate full artwork · Nano Banana / FLUX
Generate the complete artwork
Whole image recreatedAI redraws the athletes, equipment, scenery and strips together, using your video moments and layout as references. It can rebuild cropped bodies and blend the action into one scene with shared lighting, perspective and ground contact.
For several interacting subjects, start with 3–4 clear moments. AI can change poses, uniforms and equipment; compare the result with your layout guide. Nano Banana sends the layout and source reference frames to Comfy and Google and spends account credits when you click Generate. The FLUX option uses your configured GPU server.
Build your layout to generate the complete artwork.
Earlier experiments · background treatment & gentle enhancement
Give it an art treatment
Original poses keptAI reimagines the scenery from the video. The real athlete, equipment and layout are placed over it unchanged. This sends the scenery and layout guide to your configured GPU server.
Checking art engine…
Sharpen poses & scenery
Faithful enhancementRefine each displayed pose and the background at a higher resolution. The sequence, silhouettes and layout stay in place; a shared athlete reference helps keep clothing and equipment consistent.
Lower strength stays closer to the source. Higher strength recreates more detail. AI can change faces, hands or equipment details; compare the result before choosing it. This sends the pose crops and scenery to your configured ComfyUI server.
Build a layout to enhance it.
Selected moments —
Process monitor
ReadyWaiting for a video
- 00:00Every processing step is recorded here.
Inspect & export
R&DInspect the estimated background first, then check the masks for missing details and unwanted objects.
Action strips can combine original poses with a generated background.
Settings changed. Create a new composite to apply them.
Download experiment ZIPZIP includes available outputs, source frames, masks, transparent layers and settings.