A Claude Code skill that generates complete, deployable avatar presentations from a PDF deck. Voice navigation, real-time Q&A, session memory — delivered as a single HTML file.
Kaltura Q1 2026 Earnings — built with this skill
Install the skill, point it at a PDF, and let it build everything.
Say "go to slide 5", "next", or "back" — the avatar navigates instantly via speech recognition.
Ask questions about any slide. The avatar answers from structured domain data — never hallucinated.
VLM screen capture lets the avatar describe charts, colors, and layout it actually sees on screen.
Slides advance automatically after the avatar finishes speaking. Pauses when you engage.
Returning viewers pick up where they left off. The avatar remembers their interests and progress.
No server, no build tools, no runtime dependencies. Share via a stable Kaltura short link.
SEC safe harbor, non-GAAP handling, formal tone, disclaimer gates, IR contact collection.
CTA-driven, lead collection, fast autoplay, conversational energy, product demos.
Patient pacing, no contact gate, accessibility-aware, comprehension checks.
Source attribution, manual navigation, analytical depth, no autoplay.