Image Lab

infographic / diagram

AI-generated · not a benchmark

Zoro & Robin storyboard

One brief. Three models. Eight ways to see a Zoro x Robin storyboard.

8 candidates · 3 models · $0.5547 total

The run

What Image Lab planned, and what it actually cost.

Intent

infographic / diagram

Models

3 — GPT Image 2, Nano Banana Pro, Nano Banana 2

Candidates

8 generated

Requested size

1536×1024

Actual cost

$0.5547 (est. ~$0.1782)

Generated

2026-09-06

The prompt

The exact brief this run used. Copy it, or adapt it.

Goal: Create an information-dense anime storyboard / image-generation sheet for a 30-second One Piece-style scene titled ワンピース 30秒用 絵コンテ「ゾロとロビンの静かな会話」, showing a quiet emotional conversation between Zoro and Robin. Canvas: Landscape 16:9 white storyboard page with thin black borders, clean grid lines, and Japanese typography. Add a small top-right note reading "尺:30秒 解像度:16:9". Use the look of a production storyboard sheet: crisp black frame boxes, white caption bands under each frame, and anime screenshots inside the boxes. Layout: exactly 10 numbered storyboard panels plus 2 text note boxes. Panels 1-4 across the top row, panels 5-8 across the middle row, panel 9 at bottom left, panel 10 wide at bottom center, and two note boxes on the bottom right. Each panel: a circled number, a time range, a cinematic anime still, and a Japanese caption below it. Panels: 1. "① 0:00-0:03" Wide night shot, blue tones: Zoro and Robin sit apart among ancient stone ruins and broken columns. Caption: ロビン「…また迷ってるの?」 2. "② 0:03-0:06" Close-up of Zoro facing forward, green hair, stern expression, one eye closed, earrings, dark green robe. Caption: ゾロ「…別に」 3. "③ 0:06-0:10" Close-up of Robin resting her cheek on her hand, long black hair, gentle knowing smile, purple outfit, stone wall. Caption: ロビン「強いからって、全部ひとりで背負わなくていいのよ」 4. "④ 0:10-0:13" Side profile of Zoro in warm low light against stone ruins, serious. Caption: ゾロ「背負うつもりはねぇ」 5. "⑤ 0:13-0:16" Close-up of Robin smiling softly, wind in her hair, cloudy sky. Caption: ロビン「じゃあ…信じてるのね、仲間を」 6. "⑥ 0:16-0:19" Side close-up of Zoro looking down with a small smile, leaves blowing, daylight. Caption: ゾロ「ああ。てめェがいるからな」 7. "⑦ 0:19-0:23" Medium two-shot, bright daylight: Robin and Zoro by a stone balcony railing, blue sky, drifting leaves. Caption: ロビン「フフ…心強いわ」 8. "⑧ 0:23-0:26" Close-up of Zoro grinning confidently, one eye narrowed, gold earrings, green hair, wind-blown leaves. Caption: ゾロ「当たり前だ」 9. "⑨ 0:26-0:28" Close-up of Robin, eyes closed, serene smile, pink flower petals drifting. Caption: ロビン「ありがとう、ゾロ」 10. "⑩ 0:28-0:30" Wide sunset shot from behind: Robin and Zoro at a balcony overlooking the ocean, purple-orange sky. Caption: (無言で、同じ未来を見つめる二人) Note boxes (bottom right, exactly 2): Box 1 title 演出メモ, bullets: 全体のトーン:静かで信頼感のある会話劇 / BGM:しっとりとしたピアノメインの感動系 / SE:風の音、波の音を控えめに / セリフは間を大切に、表情の「間」で魅せる, final line: ラストカットは余韻を大切に. Box 2 title キャラクター設定: ゾロ:クールだが仲間想い、言葉は少ないが真っ直ぐ / ロビン:落ち着いていて優しい、仲間を信じて支える強さ. Visual style: high-quality hand-drawn Japanese TV anime aesthetic, clean cel shading, expressive faces, cinematic lighting, blue night ruins to bright sky to warm sunset. Zoro: short green hair, three gold earrings, a scarred closed eye, muscular build, dark green clothing. Robin: long black hair, elegant features, calm eyes, purple patterned top. Constraints: all visible text in Japanese as specified, precise readable grid, exactly 10 panels, exactly 2 note boxes, no watermarks, no logos, no photorealism.

→ use Image Lab

  1. Copy the prompt above.
  2. Paste it into any agent that has the Image Lab skill — or install it below.
  3. Run it as-is, or swap the subject, style, or details to make it yours.

How this was made

Image Lab is not a benchmark — different models and seeds make these an unfair comparison. It turns one brief into strong choices. Everything below did real, billed work in this run.

  • CogFoundry model gatewayRan every generation in this run through POST /api/v1/tasks/generations and reported the real per-image cost. Run total: $0.5547.
  • GPT Image 23 alternatives (A, B, C).
  • Nano Banana Pro3 alternatives (D, E, F).
  • Nano Banana 22 alternatives (G, H).
  • Image LabA loomloom community example — the skill that classified the brief, chose the model allocation, gated the cost, and rendered this page.

From image exploration to AI work.

loomloom will bring agentic workflows to Image Lab — turning a single prompt into an iterative process of generation, evaluation, selection, and refinement.

Bring your own prompt

Install Image Lab, give it any image prompt, and run cost-gated exploration across the best-fit models.

npx skills add cogfoundry-labs/loomloom --skill image-lab -a claude-code -g -y
See Image Lab on GitHub