LlamaGen Logo
立即訂閱

Image prompt collection · 獨立創作教學

Nano Banana 漫畫與角色提示詞範例

使用固定參考圖製作漫畫分格與封面，並為可編輯對白保留空間。

Create comic panels Fix character drift

獨立教學內容。GPT-6 與 Astra 為 OpenAI 產品，Claude 屬於 Anthropic，Gemini 與 Nano Banana 為 Google 產品；本頁並非官方合作或背書。 2026 年 10 月 15 日更新

Prompt-to-comic artwork example with illustrated panels that demonstrate comic prompt composition
圖片來源: LlamaGen prompt-to-comic gallery image for composition reference. Not a verified Nano Banana model sample. 查看來源 ↗

此頁目前僅部分翻譯；詳細實作案例、提示詞及常見問答仍為英文。完成全文翻譯及審核前，不會開放搜尋引擎索引。

重點解答

使用固定參考圖製作漫畫分格與封面，並為可編輯對白保留空間。

An effective image prompt gives the tool constraints that a comic artist would recognize: stable character identity, camera, negative space for text and a clear panel function. Start with original characters rather than existing franchise lookalikes.

模型能力重要說明 Nano Banana is Google's image-model family. Different versions and surfaces support different edit and reference features; see Google's current documentation.

資料實際支持的結論 Google image docs explain model support and reference constraints. Recipes are illustrative prompt-writing advice, not a reported test. Verify the source ↗

Worked example · editable production brief

Nano Banana comic prompt recipe

Build each comic prompt from controllable parts instead of stuffing the entire comic script into one image description.

Prompt ingredientIdentity
Example valueC01: copper robot, teal eyes, yellow apron
Why it mattersStable cast across scenes
Prompt ingredientFraming
Example valueMedium shot, waist-up, left third of panel
Why it mattersLeave space for dialogue
Prompt ingredientAction
Example valueTips watering can toward old boot
Why it mattersOne clear narrative beat
Prompt ingredientStyle & palette
Example valueClean bold inks, muted teal/gold cel shading
Why it mattersVisual continuity
Prompt ingredientConstraints
Example valueNo lettering, no logos, keep reference appearance
Why it mattersReduce fix-up and IP confusion

Use the same reference sheet whenever supported; add balloon text in an editor after the visual pass.

操作教學

Follow the sequence using an original or authorized project. The workshop above is an editorial plan, not output falsely attributed to a model.

  1. Step 1

    Fix your image aspect ratio

    Use a layout appropriate for portrait panels, a square strip or a tall webtoon rather than improvising the crop after generation.

  2. Step 2

    Supply original reference artwork

    Provide the same character sheet in every relevant generation where the tool supports it.

  3. Step 3

    Describe the shot and emotional beat

    Specify wide, medium or close-up, direction of gaze, scene action and contrast. Keep the story beat singular.

  4. Step 4

    Reserve lettering zones

    Prefer clean negative space for dialogue balloons; add exact text in the layout editor.

  5. Step 5

    Evaluate the sequence, not one image

    Lay out panels at the final intended reading size and inspect face consistency and visual pacing.

可複製的提示詞

Replace placeholder details with your own original story and reference assets. Paste into a compatible model or workflow; available tools and results differ.

1. Four-panel slice-of-life strip

Create four distinct visual beats of an original comic with two recurring characters: a teal-haired bike mechanic in an orange jacket and a fluffy grey dog with a red bandana. Panel beats: flat tire, surprise pawprint, improvised repair, playful reunion. Keep cast, clothes, bicycle and dog markings consistent. No words or logos in the artwork.

2. Manga cover composition

Original sci-fi manga cover, portrait composition, confident mechanic hero centered slightly left, futuristic station background, controlled red/teal palette, dramatic rim lighting, large clear area at top for manually added title. No generated lettering, trademarks or recognizable existing characters.

3. Webtoon vertical panel

A single tall webtoon panel of an original detective in a rainy neon alley. Low camera angle; significant vertical breathing room for scroll pacing; blue rain reflections, cinematic shadows, reserved blank space for dialogue. Maintain the same coat, haircut and badge from the attached reference.

Failure → Cause → Fix

漫畫創作常見問題與修正方法

Specific failure patterns matter more than a generic 'try again'. These fixes target the output you can inspect.

1. Text spelling is inconsistent

可能原因 Generated pixels are hard to retype.

處理方式 Request blank balloon space; letter manually.

2. Three different art styles appear

可能原因 Style varies from panel to panel.

處理方式 Carry a fixed palette and ink/lighting description.

3. Image crops off heads and balloons

可能原因 Aspect ratio and composition were underspecified.

處理方式 Lock output shape and blank-space requirements.

發佈前檢查項目

Review progress: 0 of 4 checks complete

A personal QA worksheet, not an automated pass/fail evaluation of your image.

Create comic panels

常見問題

Is Nano Banana the same as Gemini?

Nano Banana names Google's Gemini image-generation models; exact model versions and supported features can differ.

What is the best prompt for several comic panels?

A consistent character sheet and separate prompts for each beat are often easier to edit than a single sprawling page prompt.

Can I put lettering inside the prompt?

You can request text, but inspect accuracy closely; manual lettering is usually easier to correct and translate.

參考來源

Primary documentation and public creator discussions; not a substitute for hands-on results.