Se o seu personagem muda o rosto, o cabelo ou detalhes da roupa a cada reroll, o problema não é “criatividade”. É falta de invariants.
Este post é um pacote de gpt image 2 character sheet prompts para você copiar e gerar turnarounds, expression sheets e pose sheets. A meta é ter um personagem reutilizável em thumbnails, storyboards e mockups sem drift. Em produção, tente padronizar um arquivo único: um gpt image 2 character sheet prompt versionado, que todo mundo usa como baseline.
- See consistent sets for reference: /showcases
- Use layout + constraints patterns: /blog/gpt-image-2-prompt-patterns
- Store baselines in a prompt library: /blog/gpt-image-2-prompt-workspace
Keywords this post intentionally covers (for search intent)
- gpt image 2 character sheet prompts (principal)
- gpt image 2 character sheet prompt (baseline única)
- gpt image 2 character consistency prompts (anti drift)
- gpt image 2 character consistency (termo curto)
- gpt image 2 turnaround sheet prompt (multi-view)
- gpt image 2 expression sheet prompt (expressões)
- gpt image 2 pose sheet prompt (poses)
- gpt image 2 model sheet prompts (naming de produção)
- gpt-image-2 character sheet prompts (variante com hífen)
TL;DR: the character sheet workflow in 90 seconds
Para fazer gpt image 2 character consistency prompts funcionar:
- Escreva um identity block com invariants (geometria do rosto, cabelo, roupa, paleta).
- Adicione um layout spec fixo (grid, margens, labels).
- Gere uma baseline sheet primeiro.
- Itere com uma variant ladder (mude uma coisa por vez: expressão ou pose).
- Salve a baseline como gpt image 2 character sheet prompt reutilizável.
What a character sheet is (and why prompts drift)
Uma character sheet é uma spec, não “uma imagem”:
- turnaround sheet: múltiplas views (front/side/back ou 3–4 ângulos)
- expression sheet: mesmo framing, expressões diferentes
- pose sheet: mesma identidade, poses diferentes
Drift em gpt image 2 character sheet prompts geralmente vem de:
- identity sem invariants
- layout sem spec (escala e framing mudam)
- variáveis demais (você muda tudo e não consegue debugar)
The character bible (lock identity invariants)
Antes de escrever um gpt image 2 turnaround sheet prompt, escreva o identity block. Esse bloco é o núcleo de gpt image 2 model sheet prompts: você cola igual em todas as sheets.
Identity block (must stay constant):
- Name: [character name / short ID]
- Age range + vibe: [e.g., early 20s, energetic, friendly]
- Face geometry: [jawline, cheekbones, nose shape, eye shape]
- Hair: [style, length, color, bangs]
- Skin tone: [descriptor]
- Outfit: [top/bottom/shoes], colors, signature accessories
- Proportions: [head-to-body ratio], height vibe, body type
- Palette: [3–5 colors], avoid new colors
- Style family: [anime / semi-realistic / Pixar-like], keep consistent
Invariants:
- Keep facial identity consistent across all panels
- Keep outfit and accessories identical across all panels
- Same lighting direction and intensitySe você quer consistência, trate esse bloco como a base das suas gpt image 2 character consistency prompts.
Turnaround sheet prompt (3–4 views)
Use este gpt image 2 turnaround sheet prompt para obter multi-view estável:
Goal:
- Create a 4-panel character turnaround sheet (front, 3/4, side, back)
[PASTE identity block here]
Layout (do not change):
- 2x2 grid, equal panels, consistent margins
- Full-body framing in every panel, consistent scale
- Neutral background, no props
- Label each panel (front / 3-4 / side / back), readable text
Invariants (do not change):
- Same outfit, same hair, same accessories
- Same camera distance and angle
- Same lighting direction and intensity
- Keep facial identity consistent across all panels
Output spec:
- Aspect ratio: 4:5
- Quantity: 1Expression sheet prompt (8–12 expressions)
Um gpt image 2 expression sheet prompt deve mudar expressão, não identidade:
Goal:
- Create a 12-panel expression sheet for the same character
[PASTE identity block here]
Layout (do not change):
- 3x4 grid, equal panels, consistent margins
- Head-and-shoulders framing in every panel
- Label each panel with the expression name (readable)
Invariants (do not change):
- Same outfit, same hair, same accessories
- Same camera distance and angle
- Same lighting and background
- Keep facial identity consistent across panels
Variables (change only facial expression):
- Expressions (12): neutral, happy, sad, angry, surprised, confused, smug, scared, determined, tired, laughing, thinking
Output spec:
- Aspect ratio: 4:5
- Quantity: 1Pose sheet prompt (6–10 poses)
Use este gpt image 2 pose sheet prompt para poses controladas:
Goal:
- Create an 8-panel pose sheet for the same character
[PASTE identity block here]
Layout (do not change):
- 2x4 grid, equal panels
- Full-body framing in every panel
- Consistent margins and scale
Invariants (do not change):
- Same outfit, same palette, same accessories
- Same camera angle and distance
- Same lighting and background
- Keep facial identity consistent across panels
Variables (change only pose):
- Poses (8): standing neutral, walking, running, sitting, pointing, waving, jumping, arms crossed
Output spec:
- Aspect ratio: 4:5
- Quantity: 1QA checklist: fix identity drift
Checklist rápida para drift em gpt image 2 character sheet prompts:
- Identity block foi colado sem mudança?
- Outfit + acessórios estão marcados como “must stay identical”?
- Câmera fixa (sem “dynamic angle”)?
- Lighting fixo?
- Você muda uma variável por vez?
- Você está reutilizando baseline (ou começando do zero)?
Se o rosto driftar, encurte o prompt e mova detalhes para o identity block. Esse é o fix mais confiável para gpt image 2 model sheet prompts.
Quick terminology map (to keep your library consistent)
- gpt image 2 character sheet prompts
- gpt image 2 character sheet prompt
- gpt image 2 turnaround sheet prompt
- gpt image 2 expression sheet prompt
- gpt image 2 pose sheet prompt
- gpt image 2 character consistency prompts
- gpt image 2 character consistency
- gpt image 2 model sheet prompts
Next steps
- Browse consistent sets: /showcases
- Build a prompt library: /blog/gpt-image-2-prompt-workspace
- Learn layout/constraints patterns: /blog/gpt-image-2-prompt-patterns

