Vixel Community
GPT Image 2 prompts for story creators
A useful story-image prompt is a small production brief: it separates persistent identity from the shot-specific change, fixes the frame's narrative job, and names what must not drift.
Method
- Lock the identity block: Define the traits that must survive every shot: silhouette, age range, hair, wardrobe anchors, signature prop, and stylization level.
- Write the shot delta: Describe only what changes in this frame: action, emotion, camera distance, angle, location, lighting, and the story beat the image must communicate.
- Constrain composition and output: Specify aspect ratio, subject placement, readable focal hierarchy, and exclusions such as logos, accidental text, duplicate limbs, or generated UI.
- Compare proof, then revise one variable: Judge identity, scene readability, and continuity separately. Change one block at a time instead of rewriting the whole prompt after every miss.
Common failure modes
The character changes between shots
Keep one canonical identity block and move pose, action, camera, and emotion into a separate shot block.
The frame is beautiful but unusable
State the beat's narrative function and reserve clean space for the action or choice rather than asking only for a style.
The prompt becomes a keyword pile
Use ordered production language—identity, shot, composition, light, exclusions—and remove conflicting style adjectives.
Related generated recipe proofs
- Timed cast-and-shot handoff - Pair a clean cast identity strip with a timed shot strip: the cast strip locks face, silhouette, wardrobe, and prop ownership; the shot strip locks start pose, camera lane, action…
- Chronological motion filmstrip - Design a strict chronological filmstrip: nine panels progress setup, player pressure, action, reveal, and consequence without jumping camera side or changing prop state ambiguousl…
- Numbered combat storyboard replay board - Build a numbered combat replay board: each panel owns one start/contact/reaction/recovery state, same silhouette, same prop target, stable camera side, and a clean player-action p…
- Key-pose anime action plate - Build a cel-shaded key-pose plate with one readable start pose, contact pose, reaction pose, and consequence pose, preserving silhouette, prop target, light direction, and route a…
- Mid-poly motion preview board - Use a mid-poly stylized first frame with simple geometry, tactile material blocks, clear route prop, and a camera lane that video models can continue without fighting fine facial…
- Character-sheet motion bible - Build a character bible first: locked silhouette, face geometry, wardrobe tokens, expression ladder, and prop ownership, then generate branch frames from that bible.
- Monochrome martial phase sheet - Use a monochrome rough storyboard as a motion phase sheet: one performer, one pose phase per panel, strong silhouette, and strict screen direction before handing the board to vide…
- Motivated foreground handoff board - Build a storyboard where each camera move is motivated by an onscreen object or character action, then keep the Seedance prompt focused on focus handoff, timing, and transition lo…
Frequently asked questions
What should a GPT Image 2 character prompt include?
Use a persistent identity block, a shot-specific action and emotion block, camera and lighting instructions, aspect ratio, and a short exclusion list. Keep stable traits separate from changing shot details.
Do Vixel recipe pages expose every full prompt?
Generated proofs, technique notes, sources, and prompt previews are public. Five allowlisted recipes are fully free; the complete library requires a verified Creator Standard or Pro entitlement.
Compare Creator plans