GhostCut AI Video Localizer

Problem

For short-drama, AI-drama, and video localization workflows, an agent may need to erase subtitles, burn subtitles, extract OCR or ASR subtitles, translate speech, re-dub audio, and translate images. Calling the GhostCut gateway directly is error-prone because endpoint paths, field nesting, language codes, media URLs, task status, and localization project IDs are scattered across different documents. Guessing from memory often breaks request payloads, signing, or result lookup.

How It Works

This skill organizes the GhostCut API guide into an agent-readable document map:

• Start with references/api-guide/01-quickstart.md and an example payload to create a video text-removal task, then check task status.
• Use AppKey / AppSecret / AppSign for authentication; upload local files per 10-file-upload.md before passing a URL or material ID.
• Single-video tasks cover subtitle erasure, subtitle burning, OCR and ASR subtitle extraction, basic editing, background music separation, and video speech translation.
• The localization module targets short-drama, AI-drama, idSeries, and idMaterialVideo scenarios, including subtitle extraction, dubbing, burning, audio separation, glossaries, multi-episode merge, and character management.
• Bundled scripts can reduce repetitive signing, upload, status, and balance queries, while real credentials should be injected via environment variables.

Boundaries

This is an API integration guide, not a final media generator, and it does not replace business content review. Generic video translation, dubbing, or subtitle requests should default to single-video APIs; use the localization module only when short-drama or AI-drama export, project-level materials, or batch episode handling is explicit. Success is determined by code == 1000, not HTTP 200; task/list.body[].id is a task or project ID, while work IDs come from body.content[].id in /work/status.