banana-youtube

name	banana-youtube
description	Generate bold, high-contrast YouTube thumbnail images with dramatic framing, exaggerated expressions, and click-optimized compositions via Gemini. Triggers on: youtube thumbnail, thumbnail, youtube, video thumbnail, yt thumbnail, video cover image, click-worthy thumbnail.
allowed-tools	Bash(uv:*), Bash(ls:*), Read

YouTube Thumbnail Maker

Input: $ARGUMENTS (optional — video topic or thumbnail description)

If $ARGUMENTS is provided, use it as the video topic and skip to prompt building. If empty, ask what the video is about.

Requirements

• uv installed
• GEMINI_API_KEY environment variable set (get one at https://aistudio.google.com/apikey)

Workflow

1. Understand — Parse the user's request. Identify the video topic, target emotion, and channel style. If the subject or intent is unclear, ask ONE clarifying question.
2. Build prompt — Construct a detailed Gemini prompt using the Prompt Formula below. Prioritize dramatic facial expressions, bold colors, and high visual contrast.
3. Configure — ALWAYS use 16:9, 2K, and the Pro model. These are non-negotiable for thumbnails.
4. Generate — Run the script.
5. Deliver — Report the saved file path. Do NOT read the image file back. Offer to iterate: adjust expression, background color, composition, or prompt.

Prompt Formula

Build every thumbnail prompt using this structure:

[Subject with Expression] + [Dramatic Reaction/Gesture] + [Bold Colored Background] + [Close-Up Rule-of-Thirds] + [Hyper-Saturated Photography with Rim Lighting]

Rules

• Positive framing: Describe what IS in the image, never what is absent.
• Strong verb opener: Start with Generate, Create, Capture, Render, Design, Compose.
• Faces sell clicks: Always include a face with an exaggerated expression — wide eyes, open mouth, raised eyebrows, shocked grin.
• Bold backgrounds: Use solid, saturated colors (electric blue, hot pink, neon green, deep orange) or dramatic gradients.
• Rim lighting: Add a strong backlight or rim light to separate the subject from the background and create a cinematic pop.
• Text space: Reserve approximately one-third of the frame for text overlay — typically the opposite side from the subject.
• No celebrity likenesses: Never reference real public figures. Describe features generically.

Defaults

Parameter	Default
Aspect ratio	16:9
Resolution	2K
Model	gemini-3-pro-image-preview

ALWAYS use these defaults. Thumbnails require widescreen, high resolution, and the Pro model for maximum quality. Do not downgrade.

Thumbnail Psychology

• Faces with exaggerated expressions drive clicks. A shocked, excited, or curious face is the single strongest thumbnail element. Always make the expression bigger than real life.
• Place the subject on the left or right third. Use rule-of-thirds positioning to leave space for title text on the opposite side.
• High saturation, deep contrast. Thumbnails are viewed at small sizes on crowded feeds. Muted tones disappear. Push saturation and contrast to make the image pop at 120px tall.
• Shallow depth of field. Blur the background aggressively to make the subject the undeniable focal point.
• Reserve ~1/3 of the frame for text overlay. The creator will add bold text in post — leave clean, uncluttered space for it.

Common Video Genres

Genre	Subject Guidance	Background Style	Expression / Gesture
Tech review	Person holding the device, angled toward camera	Electric blue or dark gradient	Wide-eyed amazement, mouth slightly open
Cooking	Chef holding the finished dish at eye level	Warm orange or kitchen bokeh	Proud smile, steam rising from plate
Travel	Traveler with arms wide open, scenic landmark behind	Vivid sky or sunset gradient	Joyful awe, looking into the distance
Tutorial	Person pointing at a floating code block or diagram	Deep purple or dark teal	Confident nod, direct eye contact
Fitness	Athlete mid-motion, muscles defined, sweat visible	Bold red or neon green	Intense determination, clenched jaw
Unboxing	Hands lifting the product out of the box	Bright yellow or white spotlight	Surprised delight, eyebrows raised
Storytime	Person leaning toward camera, conspiratorial angle	Moody dark background with spotlight	Whispered secret expression, wide eyes
Reaction	Person recoiling or leaning back in shock	Hot pink or chaotic gradient	Jaw dropped, hands on cheeks
Educational	Presenter beside a whiteboard or floating infographic	Clean white or soft blue	Thoughtful curiosity, finger raised

Generation

uv run "${CLAUDE_PLUGIN_ROOT}/scripts/banana.py" \
  --prompt "<detailed prompt following the formula>" \
  --output "<YYYY-MM-DD-HH-MM-SS-descriptive-name>.png" \
  --resolution 2K \
  --aspect-ratio 16:9 \
  --model gemini-3-pro-image-preview

Editing (with input images)

When the user provides an existing image to modify (e.g., swap a background, change expression, adjust colors):

uv run "${CLAUDE_PLUGIN_ROOT}/scripts/banana.py" \
  --prompt "<edit instruction: what to change AND what to keep>" \
  --output "<YYYY-MM-DD-HH-MM-SS-descriptive-name>.png" \
  --input-image "/path/to/source.png" \
  --resolution 2K \
  --aspect-ratio 16:9

For edits, describe both the change and what must stay identical:

"Change the background to a bold neon green gradient. Keep the subject's face, expression, clothing, and pose exactly the same."

Up to 14 input images can be passed (repeat --input-image for each).

References

See references/prompts.md for ready-to-use example prompts covering common YouTube thumbnail scenarios across multiple video genres.