Character Cut
This capability is available through partnership. This page introduces the capability only and does not publish API parameters. Contact business@ai-mcn.tv to confirm the integration scope; the internal API reference and integration support are provided after that.
Character Cut finds editable intervals for a selected person across a complete video. It first builds a shot timeline, then uses visual references to identify the target across different shots and appearances, while retaining useful reactions, exchange shots, and scene-establishing context. The result is a structured timeline for the editor to review and refine.
What it does
- Locks the identity with a reference image and recognizes the same person in frontal views, profiles, back views, occlusion, and changing outfits;
- Returns the target person's core intervals by physical shot, with explainable evidence frames and source timecodes;
- Marks reaction shots, exchange shots, and establishing context separately so the editorial relationship remains visible;
- Marks unclear identity or uncertain relationships explicitly instead of turning weak evidence into a confident match;
- Returns a structured timeline that the CLI or editor can merge, trim, and render locally.
Where it fits
Long video / feature footage → prepare a character reference → Character Cut → review and merge character intervals → character recap, asset organization, or final edit
Character Cut answers “where does this person appear?” It does not render a new video. Its structured result can feed a local editing project, character asset organization, and later B-roll or narration workflows.
Good fits
- Collecting a character's appearances before making a film or drama commentary;
- Building character recaps, relationship edits, or actor-focused compilations;
- Extracting a selected person's material from a long video for an editor to refine;
- Organizing footage where the character's appearance and necessary context both matter.
Data boundary and billing
The analysis uses a character reference, the source shot timeline, and the required analysis proxy. The exact local preprocessing and evidence-retention boundary are agreed as part of the partnership. The service returns structured intervals and recognition evidence; it does not replace editorial review or promise frame-perfect isolation of the person.
Billing, quotas, and integration scope are defined in the partnership plan. The confirmed commercial terms are the source of truth.
How to integrate
This capability is currently available through partnership. Share your footage scale, number of target characters, existing editing pipeline, and preferred deliverable. We will confirm the CLI or custom integration path and provide the internal API reference and integration support.
Contact business@ai-mcn.tv.