mcpField
M
muapi-multi-angle-reshoot

muapi-multi-angle-reshoot

github.com/samuraigptCreative & media25665 installsUpdated 4w ago
Repository

About muapi-multi-angle-reshoot

An LLM-orchestrated muapi skill for creating a subject and rendering it from six camera angles. It is for users who want a set of visually consistent perspectives, with an optional reference image to anchor the subject.

Key features

  1. 01Generate a base subject image
  2. 02Create six camera-angle variations
  3. 03Use an optional reference image
  4. 04Present a multi-angle gallery

Use cases

  1. 01Explore alternate viewpoints for a scene
  2. 02Create angle variations of a product or person

Install

Copy the command and run it in a terminal.

$npx skills add https://github.com/samuraigpt/generative-media-skills --skill muapi-multi-angle-reshoot

After installing, type/muapi-multi-angle-reshootin Claude Code to confirm it loaded.

Details

From the project README

Multi-Angle Reshoot

Re-render a subject or scene from multiple dramatic camera angles, such as fish-eye, bird's-eye, low-angle, and macro, while maintaining consistent identity and detail.

Inputs

NameTypeRequiredDefaultDescription
subject_descriptiontextyes—Description of the subject or scene (e.g. "a professional woman in a black suit holding a laptop").
reference_imageimage_urlno—Optional reference image to maintain consistency across angles.

Steps

Phase A — Subject Establishment

If {{reference_image}} is not provided, submit the plan with ONE step to create the base subject:

  1. Subject Generation — muapi image generate (model=nano-banana-pro):
    • Prompt: A high-quality, professional photograph of {{subject_description}}. Clean studio lighting, sharp focus, realistic textures, cinematic quality.
    • Aspect ratio: 3:4 or 1:1

Present the base image to the user for approval.

Phase B — Multi-Angle Reshoot

Once the subject is established, submit the plan using muapi image edit (model=nano-banana-2-edit) to generate the requested angles. Use the approved image from Phase A as the reference for ALL steps.

  1. Bird's Eye View
    • Reference: Base Image
    • Prompt: A dramatic bird's eye view (from directly above) of the same subject. High-angle perspective, wide shot, looking down. Maintain exact clothing, face, and setting consistency.
  2. Fish Eye Lens
    • Reference: Base Image
    • Prompt: A high-distortion fish-eye lens photograph of the same subject. Ultra-wide angle, curved edges, dramatic perspective, immersive feel. Maintain exact clothing and face consistency.
  3. Low Angle (Hero Shot)
    • Reference: Base Image
    • Prompt: A low-angle "hero shot" looking up at the subject. Makes the subject appear powerful and dominant. Maintain exact clothing and face consistency.
  4. Dutch Angle
    • Reference: Base Image
    • Prompt: A stylized Dutch angle (tilted horizon) photograph of the same subject. Cinematic, slightly uneasy or dramatic vibe. Maintain exact clothing and face consistency.
  5. Macro Close-up
    • Reference: Base Image
    • Prompt: An extreme macro close-up focused on a specific detail of the subject (e.g., eyes, texture of the suit, or a feature of the laptop). Shallow depth of field, incredible detail. Maintain exact colors and textures.
  6. Worm's Eye View
    • Reference: Base Image
    • Prompt: A worm's eye view from ground level looking straight up at the subject. Dramatic vertical perspective. Maintain exact clothing and face consistency.

After generating the variations, present the multi-angle gallery to the user. You can also offer to generate specific individual angles based on their feedback.

Trigger Keywords

multi angle, photo reshoot, change camera angle, fish eye view, birds eye view, low angle shot, macro photography, camera perspective


Notes for the Executing Agent

  • This recipe is LLM-orchestrated: read each phase, gather any missing inputs from the user, then call muapi CLI commands. Use muapi auth configure first if MUAPI_API_KEY is unset.
  • For model IDs without a CLI alias yet, fall back to the raw endpoint via curl -X POST https://api.muapi.ai/api/v1/<endpoint> -H "x-api-key: $MUAPI_API_KEY" -H 'content-type: application/json' -d '{...}' and poll with muapi predict wait <request_id>.
  • Substitute {{input_name}} placeholders with the user's actual inputs before issuing each call.

Comments

Markdown supported