Drift-Proof 3-Panel Character Sheets and Frame Maps
Builds 3-panel face-lock character sheets and Frame Map shot prompts with degree-based FOV to stop face drift and camera drift in AI video.
Before & after
Same prompt, same agent — the only difference is the skill.
I can help improve your character sheets and reference materials. A 3-panel layout might work better for focusing on faces and facial expressions. You could arrange it with front view, rear view, and a detailed close-up. That approach would give you much more pixel space for facial details and expression clarity. What overall style are you going for with your characters? Should I adjust the template based on your existing sheets or start fresh? How many different characters do you need to update with this new template system?
3-PANEL CHARACTER SHEET v3.0 SPECIFICATION: PANEL 1 — FULL BODY FRONT (headless): 80% of canvas width. Standing neutral pose. Full garment detail, seams, fabric texture visible. Neck cut choice required: clean edge for transparent fabrics, ghost mannequin style for opaque clothing to show body silhouette underneath. Shoes and base positioning shown clearly. PANEL 2 — FULL BODY REAR + HEAD INSET: 80% canvas width rear silhouette. Head profile positioned top-right corner (20% of panel). Shows back garment detail, seams, hem line. Hair flow direction marked. Profile nose, ear, jawline visible. Back posture and spinal curve clear. PANEL 3 — CHEST-UP FACE LOCK: 40% canvas width dedicated to face detail. Face takes 70% of panel space. Expression: neutral (not smiling, not grimacing). 3/4 angle showing both eyes clearly. Used for identity anchoring in downstream video generation tools. Eyebrow arch, pupil direction, facial proportions precise. FOV CONTROL: Frame Map specifies camera distance in degrees before rendering. Example: 50mm lens equivalent = 47 degrees horizontal FOV. Locking depth value first prevents drift when sheet feeds into video generators. Supports 24mm to 200mm equivalent focal lengths. NAMED ELEMENT TAGS: Use semantic tags @panel1_front, @panel3_face, @garment_neck, @fabric_detail instead of numbered references. Named tags survive template edits and refinements without breaking downstream tool integrations.
About this skill
Your character's face changes between cuts. The camera drifts or pushes in when you asked for a locked shot. A "35mm" lens comes out looking like any other lens. Usually the cause is a crowded character sheet where the face gets a few hundred pixels, plus a shot prompt that describes the person before it pins down the frame. This skill rebuilds both.
What it does
- Rebuilds crowded 6-panel sheets as 3 panels: a headless full-body front, a full-body rear view with the head, and a large chest-up face-lock panel that gets most of the pixels
- Chooses the neck cut for each garment: a ghost mannequin for hoods, collars and high necklines, a clean neck cut for open necklines
- Writes shot prompts in Frame Map order: screen position and depth for each element first, then identity, behavior, camera and light
- Gives field of view in degrees, using a millimeter-to-degree conversion table, and states the camera behavior so the camera doesn't drift
- Turns vague direction ("moody fog", "walks confidently") into numbers a model can render: speeds, distances, density percentages
- Replaces "the woman in image 2" with stable named tags such as @sol_ref, so the prompts survive when references get reordered
What you get
A ready-to-paste prompt for the character sheet, plus one or more Frame Map shot prompts with a tag map. If you paste in a prompt that keeps drifting, you get a diagnosis of the likely cause and a corrected version.
Best for
AI filmmakers and creators making multi-shot videos with recurring characters in Seedance, Higgsfield, Kling or similar tools, who keep losing face consistency or camera control between shots.
SKILL.md preview
---
name: drift-proof-character-sheets-frame-maps
… (sign up to view the full skill)More creator skills
View all Creator skills →Explainer B-Roll Prompt Pack for AI Video
Turns a VO transcript into a style-locked AI video prompt pack: coverage map, P1-P3 clips timed to the VO, exact data routed to code.
Seedance Scene Prompts: Action, Dialogue and Blockouts
Seedance prompts for scenes with people: weighty action, living performances, timed dialogue, multi-clip continuity, and blockout v2v.
Faceless 3D Animated Shorts Pipeline
Concept, timed narration, style bible, and per-scene image, motion, VO, and SFX prompts for faceless 3D shorts that stay consistent.
Seedance 2.0 Video Prompt Writer
Turns an idea into a timecoded multi-shot Seedance 2.0 prompt with camera moves, continuity, reference roles, and synced sound cues.