HackerTrans
TopNewTrendsCommentsPastAskShowJobs

srinivasvaddi

no profile record

Submissions

Show HN: Tab Council – turn your AI tabs into a model council

github.com
4 points·by srinivasvaddi·2 ay önce·0 comments

Show HN: Enhanced Copy – copy buttons that include the site's AI prompt

github.com
2 points·by srinivasvaddi·2 ay önce·0 comments

Framecraft – Generate demo videos from a prompt using HTML Canvas/Skill/MCP/LLM

github.com
1 points·by srinivasvaddi·3 ay önce·3 comments

GTabs – AI tab organizer for Chrome that works with any LLM

github.com
1 points·by srinivasvaddi·3 ay önce·0 comments

comments

srinivasvaddi
·3 ay önce·discuss
Thank you! Lmk!Also feel free to drop a PR if you think something should change
srinivasvaddi
·3 ay önce·discuss
framecraft is a Claude Code plugin that generates demo videos from a single sentence. The LLM writes HTML scenes with CSS animations, generates voiceover via Edge TTS, captures frames via Playwright, and composites with FFmpeg — all through MCP servers.

The architecture: a SKILL.md file teaches the LLM story structure, pacing, and visual design. The LLM then orchestrates 3 MCP servers directly. No custom video pipeline needed — the skill alone is sufficient, which we verify by running CI with only the external MCPs (no framecraft wrapper).

Some numbers from CI runs: a 15-second teaser costs ~$1 in tokens, takes ~49 LLM turns, ~1.8M input tokens, and ~14 minutes. Token-heavy but it works.

Built this because I kept postponing the demo video for a Chrome extension I maintain. Screen recording was 45 minutes per update. Now it's one prompt.

Source: github.com/vaddisrinivas/framecraft
srinivasvaddi
·3 ay önce·discuss
[dead]