TokiToki Internal
Atlas + S3
All runs
CompletedComfyUI

4761c1b8-2d4e-477b-bc5c-30c6eeaabe39

Queued
Jul 17, 2026, 02:25 PM
Completed
Jul 17, 2026, 02:26 PM
Execution time
9.3s
Outputs
1 · 1.2 MB
Inputs
2
Cached nodes

Outputs

1

Click an image to view it full size.

prj_demo_sh001_t01_seg2_00001_.png

node 3 · 1.2 MB

Text prompts

2
Promptnode 1 · PrimitiveStringMultiline · valueCharacter Description (your input)

Arjun turns to face Meera, hesitates, then speaks. Arjun begins to speak, facing Meera on a rooftop at night, mouth slightly open.

Systemnode 2 · GeminiImageDirect · system_promptCharacter Image Generator (Nano Banana)

Output ONE photoreal cinematic frame. No collage, no text, no watermark, no border, no split panels. The reference images define identity and setting: reproduce the face, hair, wardrobe and location EXACTLY as given. Identity is not open to interpretation - treat any drift from the references as a failure. Frame it as a still from a film: motivated directional lighting, natural skin texture, shallow depth of field, subtle grain. This is the CLOSING frame of the shot. One reference is the PREVIOUS approved frame: maintain its wardrobe, lighting and grade. This frame may also be handed to the next shot as its starting point, so leave the character cleanly framed and readable rather than mid-blur or mid-occlusion.

Input media

2

Source images and video this run consumed.

6b7fa434f92a8b80aab02d9bf1a12e49.png

node 4 · 70 B

0f7468edc1de42a555833845ac83e420.png

node 5 · 1.2 MB