magnific

Magnific is a generative-media API (image, video, audio, and stock content) served at https://api.magnific.com/v1. Almost every generation endpoint works the same way: you submit a job, get back a task_id, then either poll a GET endpoint or let a webhook deliver the result. Learn that one rhythm and the other ~60 endpoints are just different request bodies.

This skill is the operating manual. The exhaustive, per-endpoint parameter/request/response detail for every model lives in references/ — reach for it when you need an exact enum value or field name.

When to use

• Building anything against api.magnific.com — generating, editing, upscaling images; making video/music/SFX; pulling stock assets.
• Choosing the right Magnific model for a task ("which model upscales while inventing detail?", "fastest text-to-image?", "image-to-video from a still?").
• Wiring up the async lifecycle: submitting jobs, polling task_id, or receiving + verifying webhooks.
• Debugging a 401/400, an empty generated array, a job stuck IN_PROGRESS, or a webhook signature mismatch.

Do NOT use for

• Non-Magnific image/video tools (Replicate, fal, raw Stability/RunwayML APIs), or the generic OpenAI/Anthropic SDKs — those have their own skills.
• The separate Freepik stock-photo API. Magnific's stock endpoints (/v1/resources, /v1/icons, /v1/videos) are covered here; the standalone Freepik API is not.

The one mental model: submit → poll or webhook