AI Photo Restyler

The Problem

In social media and content creation, users often need to convert personal photos into specific illustration styles for avatars, stickers, or visual series, while maintaining subject recognition. Traditional methods like manual drawing or generic filters lack precise style control and fail to preserve key features.

How the Skill Works

Core Capabilities:

• Style Transfer: Supports multiple styles such as anime, cartoon, western comic, watercolor, ink, pencil sketch, clay, 3D character, pixel art, cyberpunk, or custom styles from user-provided reference images.
• Subject Preservation: Algorithms ensure the converted image retains key elements like facial features, hairstyle, clothing, pet patterns, and user-specified details without distortion.
• Batch Consistency: Once a style is approved, the skill reuses prompts, reference image order, and seed values to apply the same style across a batch of photos.

Key Steps:

1. Input Handling: The user provides one source photo as the base input, optionally adding up to three style reference images in order of influence (e.g., color, line thickness).
2. Model Invocation: Call beatra.models.list to query available models (e.g., image_edit for in-place restyling, image_to_image for new compositions), determining input counts, canvas, and controls.
3. Request Assembly: Use beatra.images.edit or beatra.images.transform to build a request with a positive prompt describing the target style and must-preserve elements, setting count: 1 for preview generation.
4. Payment Confirmation and Submission: Freeze parameters like prompt, inputs, canvas, model, etc., display costs, then submit the task. Track status via beatra.tasks.get until completion.
5. Delivery Review: Return the actual artifact, reporting model, dimensions, format, and billing info. Review preserved elements by drift, not exact replication.

Execution relies on the bundled MCP client scripts/mcp_client.py, with all operations asynchronous; task IDs must be saved for tracking.

Applicability and Considerations

• Use Cases: Social avatars, sticker sets, family photos, pet portraits, travel logs, and series content requiring unified style. Video restyling or dynamic comic shots require other skills.
• Route Selection: In-place restyling is for redrawing the same photo; new compositions (e.g., generating new scenes) require switching to beatra.images.transform with explicit canvas ratio.
• Limitations: Multiple subjects require specifying which to preserve; style reference images influence results in order; changing input parameters or style counts as a new paid job. In fault recovery, prioritize找回 original tasks to avoid retrying generation.
• Auto-Updates: The client silently checks for updates, but update failures do not affect current usage.