OpenScreen 2
Record your demo and let your AI agent do the edits
212 followers
Record your demo and let your AI agent do the edits
212 followers
Record your screen and the demo looks great on the first take: zooms planned from your clicks, smooth cursor, 3D camera moves, cursor themes, backgrounds and webcam layouts. Then let your AI agent do the edit over MCP, or cut the video yourself by deleting words in the transcript. Windows, macOS and Linux. Free, MIT, no account, no watermark.
Free
Launch Team / Built With
CloudflareYour next app belongs on Cloudflare Workers. Deploy for free
Promoted
Hi Product Hunt, I'm Etienne and I maintain OpenScreen.
It started as Siddharth Vaddem's side project, a free alternative to Screen Studio. He archived it in June and I took it over with his approval. 2.0 is four months of work since then, and 137 people have contributed to the project so far.
What's new:
a local MCP server, so Claude Code or another agent can do the edit for you
zooms planned from your clicks, as shots, not one per click
cut the video by deleting words in the transcript, captions transcribed locally
webcam beside or above the recording, with background blur or removal
a native GPU renderer, exports about 15x faster than before on our reference laptop
a 3D camera that moves with your zooms, and five cursor themes with sculpted 3D versions
Windows, macOS and Linux. Free and MIT, no account, no watermark. Thanks to Siddharth for starting it, and to everyone who contributes.
What do you make your demos with today, and what would OpenScreen need for you to switch?
@artk That's a good question. It mainly depends on the harness you use (Claude Code, Codex) and the model you choose. Opus 5.5 and GPT-6.1 Sol are very good at that kind of task.
@etienne_lescot1 Good to know! thanks
@soufi Thanks for giving it a try, Mostafa! Let me know how it goes.
Congrats Etienne, and good on you for picking it up instead of letting it sit archived.
I went through the OpenScreen source a while back. The comment in zoomSpring.ts about the eased target causing a jerk at zoom seams changed how I think about camera motion. Small detail, but you can feel it in the result.
Curious about "zooms as shots": how do you decide which clicks go in one shot? Time gap, distance on screen, or both?
@dixitram Thanks! For shots it's both: a click joins the current zoom if it comes within 3 s of the last one and the zoom can still frame them all. Otherwise it starts a new shot.
free, MIT, no watermark on a tool this polished is rare, respect for keeping it that way through the rewrite. question on the transcript-based cutting: local transcription is never perfect, so what happens when the agent deletes a word that got misheard and the cut lands mid-sentence instead of in the pause? does it show you the edit list before committing, or is undo the main safety net there?
@etienne_lescot1 makes sense that undo is the safety net for now. the thing I'd flag is that it's weakest exactly where it's needed most - English and French get the forced aligner so the cuts land cleanly, but the looser Whisper timing on other languages is where a misheard word is more likely in the first place. if the review-before-apply mode ever gets built, seems like it should ship for the non-aligned languages first rather than everywhere at once.
@galdayan True. Whisper already gives a confidence score for each word, we just don't pass it to the agent yet. If it did, it could ask before cutting a word it's not sure about. Adding it to the list.