AI Storyboarding & Shot Planning

Why Storyboarding Saves Money

Video generation credits are the most expensive part of the AI cinema pipeline. Kling Pro costs roughly $0.10-0.30 per 10-second clip. If you generate blindly, you will burn through $20-50 in failed attempts before landing on the right look. Storyboarding with image generation costs 1/10th as much. A single Midjourney or Flux image costs $0.01-0.03. You can iterate 50 times on a single frame for less than the cost of one video generation attempt. The workflow: generate storyboard frames until every shot looks exactly right, then use those frames as image-to-video inputs. This locks your visual language -- color palette, composition, character design, lighting -- before you ever touch the video pipeline. ``` COST COMPARISON: 8-scene short film Without storyboards: ~40 video generations x $0.20 = $8.00 With storyboards: ~80 images x $0.02 + 12 videos x $0.20 = $4.00 Savings: 50% fewer video credits, 3x faster production ```

Cinematography for AI: What Works

Not all traditional camera techniques translate well to AI video generators. Here is what current models handle reliably versus what they struggle with: **Reliable (use freely):** - Static wide shots and establishing shots - Slow dolly forward/backward - Slow pan left/right - Medium shots of single characters - Atmospheric shots (rain, fog, light shafts) - Rack focus (foreground to background) **Unreliable (use carefully):** - Fast camera movements (whip pan, crash zoom) - Tracking shots following walking characters - Handheld/shaky cam simulation - Complex multi-character blocking - POV shots with hands in frame - Continuous shots longer than 5 seconds **Avoid entirely (for now):** - Steadicam orbits around subjects - Crane shots with vertical movement - Split-screen or multi-angle composites - Shots requiring precise physical interactions (handshakes, fighting) Build your shot list using the reliable column. When you need an unreliable technique, plan for 3-5x more generation attempts and budget accordingly. ``` SHOT LIST FORMAT: Scene | Shot | Type | Movement | Duration | Notes 1 | 1A | Wide | Static | 4s | Establishing, rainy city 1 | 1B | Med | Slow push-in | 6s | Character reveal 2 | 2A | Close | Static | 3s | Hands on photograph 2 | 2B | Med-W | Slow pan R | 5s | Room reveal ```

Building Character Reference Sheets

Character consistency is the hardest challenge in AI cinema. The solution starts here in pre-production with comprehensive character reference sheets. For each character, generate a reference grid: 4-8 images showing the character from different angles, in different lighting, with consistent features. Use a single image generator with a locked seed or style reference. **Midjourney character sheet prompt:** ``` Character reference sheet for a Japanese woman in her 40s, short black hair with grey streaks, dark circles under eyes, angular face, wearing a grey wool sweater and navy trench coat. Four views: front, 3/4 left, 3/4 right, profile. Neutral expression. Studio lighting on grey background. Cinematic photography style. --ar 16:9 --style raw ``` **Flux character sheet prompt:** ``` A professional character reference sheet showing four views of the same person: a 40-year-old Japanese woman with short black hair streaked with grey, angular face with prominent cheekbones, wearing a grey sweater under a navy trench coat. Views: front facing, left three-quarter, right three-quarter, and left profile. Neutral expression, studio lighting, grey backdrop. Photorealistic cinematic style. ``` Save these reference images. They will serve as: 1. **Image-to-video input** for video generation 2. **Style reference** for scene storyboards 3. **Consistency check** for every generated shot

Storyboard Generation Workflow

Follow this sequence for each scene in your script: **Step 1 - Mood frame:** Generate a single image that captures the overall mood, lighting, and color palette of the scene. This is your atmospheric reference. **Step 2 - Composition frame:** Using the mood frame as style reference and your character sheet as character reference, generate the actual composition -- character placement, camera angle, framing. **Step 3 - Action frames:** For scenes with movement, generate 2-3 frames showing the progression. Frame A is the start position, Frame B is mid-action, Frame C is end position. These become your image-to-video keyframes. **Step 4 - Annotate:** Add text overlays noting camera movement, duration, audio cues, and transition type. Tools: Figma (free), Canva, or even Preview on macOS. ``` ANNOTATION FORMAT (add to each storyboard frame): [SCENE 3 | SHOT 3A | MED CLOSE-UP | PUSH IN 2s] Audio: Rain ambience + synth pad Transition: Dissolve from 2B VFX: Subtle lens flare, rain droplets on lens Character: Keiko - reference sheet v2, grey sweater ``` A complete storyboard for an 8-scene short takes 30-60 minutes with AI tools. The same work takes a traditional storyboard artist 2-3 days.

From Storyboard to Production Bible

Your final storyboard package becomes the production bible for video generation: 1. **Master shot list** (spreadsheet or table with all shots) 2. **Character reference sheets** (per character) 3. **Annotated storyboard frames** (per shot) 4. **Color palette document** (hex codes pulled from approved frames) 5. **Camera movement notes** (per shot) 6. **Audio direction** (per scene) Organize this in a single folder with clear naming: ``` project-rain-memory/ characters/ keiko-reference-v2.png keiko-wardrobe-alt.png storyboards/ scene01-shot1a-wide-establishing.png scene01-shot1b-med-character-reveal.png scene02-shot2a-close-hands.png shot-list.md script.md color-palette.png ```