How Framevo actually works.

Framevo records inside your browser using the standard getDisplayMedia API. You pick the source — a browser tab, a window, or the whole screen — and the browser handles permission.

For tab captures, Framevo additionally captures coarse interaction events to a sidecar JSON file. Every click, scroll, hover, idle stretch, and typing burst is recorded, along with the bounding rectangle of the element you clicked (privacy-safe — only geometry, no DOM text or roles).

External (window or monitor) captures cannot see element rectangles because the Framevo tab and the captured surface are different windows. The capture is still useful — the pipeline degrades to heuristic classification based on cursor signals and CV regions.

The analyze pipeline runs four passes in order:

1. Capture. Interaction sidecar is downloaded from Storage and parsed. Every click is classified into one of five tiers — primary-cta, icon, nav, form, background — by an element geometry + cursor intent heuristic.
2. Understanding. Gemini reads the video, classifies the recording type (SaaS demo, tutorial, talking-head, vertical short, …), and segments it into narrative beats (intro / action / result).
3. Camera. The balancer fuses event-derived moments, CV peaks, and AI gap-fills into a candidate pool. Overlaps collapse, low- signal moments get rejected, and a greedy selector picks the kept set under per-provenance pacing rules.
4. Edit. The finished timeline lands in the editor with focus regions sized per tier, narrative chapters on top, and the attention curve behind every moment.

If a recording loses clicks somewhere in this pipeline, the editor's Analysis Debug panel surfaces exactly which stage dropped them.

The editor is a single page: video preview, timeline below it, presets rail, and an inspector for the selected moment. The AI gave you a first draft — nothing is locked.

• Drag a moment pill to retime it. Hold the edge to resize. The pill renders its tier and effect type inline.
• Open the inspector to tune the focus region, intensity, and effect type without leaving the timeline.
• Pick a preset to swap pacing + cursor styling + click effects in one tap. Presets are reversible.
• Add keyframes for a moving focal point inside a single moment.
• Re-analyze after changing the preset or pacing. Cached analysis is re-used where possible — Gemini is only called when needed.

Export renders the timeline into a video file. The renderer honours your preset, effects settings, and any manual edits you made.

• 1080p on Free; 4K on Creator and above.
• 16:9, 9:16 (vertical), or square output.
• Watermark on Free; watermark-free on paid plans.
• The exact renderer settings are visible in the export modal — what you preview is what ships.