objectremover-video-object-ops

# ObjectRemover Video Object Ops ## Use This First - **Single install for users:** this one skill is enough from upload to final result, including troubleshooting. - Best for common ObjectRemover jobs: remove watermarks/distracting objects, or extract a subject for reuse in CapCut/Premiere/DaVinci. - Supports both **chat-like natural language selection** and API-style automated processing. ## Authentication Model This skill supports multiple authentication modes depending on deployment policy: - Browser / guest flow: - No manual credential setup in most hosted deployments. - API automation flow: - May use bearer-style authentication if enabled by host configuration. - This bundle is instruction-only and does not require local install-time secret files. - Any credential prompt and storage are controlled by the host platform. ## Service Contract - Backend: `https://apiobjectremover.tokenlens.ai` ## Processing Flow 1. Prepare asset (upload file, URL-imported asset, or existing asset id). 2. `POST /api/processing/calculate-cost` (optional). 3. `POST /api/processing/generate-mask` 4. `POST /api/processing/start-task` 5. `GET /api/processing/task/:taskId` until terminal state. 6. Read `outputUrl` / UI when completed. ## Required Request Rules - If using guest identity, include the guest session header on processing endpoints. - If using authenticated API mode, use the credential mechanism provided by the host platform. - Route processing calls to the backend domain. ## Fallback Rules - No credits: low-trial when allowed, else checkout / top-up. - 401 on guest flow: issue guest session first, retry with `x-guest-session-id`. - 401 on authenticated API mode: verify credential validity and backend routing. ## When Something Breaks - Use **[reference.md](reference.md)** troubleshooting for auth mismatch, polling stalls, and output retrieval issues. ## Additional Resources - Full endpoint contracts and troubleshooting: [reference.md](reference.md)