← Back

It’s fine.

25 s
It’s fine. — representative frame

Production

Images
ChatGPT
Video & audio
Seedance 2.0
Edit
FFmpeg + OpenCV
Reference
@olima_omega

Scenes

The plate.

The plate.

Laughter in the foreground. An empty plate in the background. Keep both reactions readable in the same frame.

Settings
Model
Seedance 2.0
Duration
8 s
Ratio
9:16
Source
480p
Audio
On
Seed
161016202

Connect the first frame only. Preserve camera distance, eye direction and the original room layout.

Image prompts
s02-as02-a
Edit the supplied photographic storyboard frame: replace BOTH adults with East Asian adults, keep exact camera position/vertical framing, near white-T-shirt laughing woman in front-right and black-T-shirt man behind-left with same empty white plate held horizontally at waist. Woman: East Asian late twenties, shoulder-length dark brown hair, laughing widely, bending forward. Man: East Asian Korean/Taiwanese appearance, age32, rather lean slim face/body, short straight black hair with small natural fringe, mildly hooded double eyelids, sparse brows, natural imperfect skin. Subtly aggrieved blank distant unfocused look, holding in irritation, lips closed, no crying, no angry grimace, no smile. Looks slightly above and beyond camera to distant off-screen right, small band of lower sclera visible beneath pupils. Empty plate clearly readable in his two hands. Keep same black plain T-shirt and domestic room/warm soft phone lighting, depth-layer composition. Do not glamorize, no facial distortion caricature. Important: preserve photographic realism and direct phone-shot visual tone, no text.
Video prompt
One continuous live-action East Asian reaction scene in the supplied image. Keep the SAME slim East Asian man with short black fringe, imperfect skin, black plain T-shirt, standing behind-left holding the SAME EMPTY white plate horizontally with both hands. The SAME East Asian woman in white T-shirt is near foreground-right, laughing freely, bending forward. The joke is the contrast: she laughs while the man silently realizes nobody can eat cake now. Preserve this deep foreground/background composition, room, soft warm phone lighting and natural focus; do not rack focus or move closer, no new cake or added people. At 0.50s both the laughter and empty plate must be clearly readable. From 0.00s to 1.50s the foreground woman laughs with small shoulder bounces and shifts slightly forward-right, while the man keeps his lips closed and empty plate clearly visible. From 1.50s to 3.00s the woman continues laughing; by 3.00s the man settles his eyes beyond the camera toward distant off-screen right, slightly above lens level, detached and unfocused rather than staring into lens. From 3.00s to 5.50s he makes one small quiet exhale and tiny natural hand/breath movement, holding in irritation; the plate remains completely empty. From 5.50s to 8.00s he keeps the same restrained aggrieved blank distant gaze and closed mouth with live breathing, while her chuckles ease a little. Warm real room laughter only, no spoken words, no music, no narrator. The man never laughs, cries, grimaces, shouts, raises plate or turns away. No slow motion, no frozen picture, no subtitles.
ComfyUI

This work uses Comfy Cloud partner nodes.

Load the source image with LoadImage. Connect it to ByteDance2FirstLastFrameNode; add the last frame for the cake shot.

Choose Seedance 2.0, paste the shot prompt and set duration, ratio, resolution and seed. Enable audio, then connect SaveVideo.

Generation is 480p. The edited master is 1080 × 1920. A cleaned, verified workflow download is not available yet.

Editing

Hard-cut to a fixed close crop for the last replay. Slow it from 0.75× toward 0.45×, lowering the laugh’s pitch with it. Soften focus before the small agi.geek label appears. Keep the first laugh at normal speed.

By agi.geek · Updated 2026-10-19.
Prompts are the originals used in production.
Workshops & production enquiries · All works