Understand what happens on screen
Search actions, objects, locations, compositions, visual styles, and emotions with natural language or a reference image.
AI video analysis
AI video analysis and understanding for editors
Jumper gives editors and AI agents a practical understanding of unfamiliar footage: what happens on screen, what people say, who appears, and which source ranges deserve a closer look.
Definition
What AI video understanding means for an editor
AI video analysis in Jumper combines several views of the same source media. Visual models understand frames and scenes, speech models create time-aligned transcripts, face analysis groups recurring people, and Summary Analysis creates layered descriptions tied back to source time ranges. The result is an editorial map of the footage—not a replacement for reviewing and verifying the source.
Capabilities
Understand picture, sound, people, and structure
Understand what was said
Create searchable, time-aligned transcripts for dialogue, interviews, speeches, podcasts, sermons, and archive recordings.
Follow people and relationships
Group and name recurring faces, then combine a person with a visual description to find specific appearances, interactions, or reactions.
Navigate long-form context
Summary Analysis creates layered, task-focused descriptions so an editor or agent can survey long footage, drill into relevant sections, and return to exact source ranges.
How it works
From raw footage to an editorially useful index
- 1
Analyze media locally
Choose the visual, speech, face, or summary analysis needed for the editorial question. Source footage stays on the machine.
- 2
Explore the footage
Search directly in Jumper or let an AI agent inspect summaries before it runs targeted visual, transcript, or people searches.
- 3
Verify and use the source
Review promising results against frames, transcripts, and source media, then prepare selects or continue in the editing workflow.
Technical documentation
Frequently asked questions
What can Jumper analyze in a video?
Jumper can analyze visual content, spoken language, recurring faces, technical file metadata, and—with Summary Analysis—the broader subjects, events, characters, and structure of long-form footage.
Can Jumper analyze long videos and large footage collections?
Yes. Jumper is designed for professional footage libraries. Summary Analysis can provide layered overviews of individual files and organize related summaries into collections before more targeted searches.
Can an AI agent understand and search the analyzed footage?
Yes. Jumper’s local MCP server exposes supported summary, visual, transcript, people, clip, export, and timeline workflows to compatible AI agents such as Claude and Codex.
Is Jumper a cloud video-analysis API?
No. Jumper is local AI video-search and agentic-editing software for professional editors. It provides a local API and MCP server, but its media analysis runs on the editor’s own macOS or Windows machine.
Does AI video analysis replace source review?
No. Analysis and summaries help editors and agents find where to look. Exact quotes, identities, edit points, and factual claims should still be verified against the transcript, frames, or source media.