Hi HN, I'm Zura, and I'm building Revline, a fully agentic video editor.
Revline can take your rough draft of a short or long video, watch it, understand the content, and make edits like a real editor would.
It's been super useful for me (not biased, obviously).
I want to post more content, but my output has never matched how frequently I want to post. Editing takes too much time, and every video I don't finish is waste of time(and money).
The problem I've had (and still have) with current "AI video editors" is that they're either single-purpose, closed, or simply don't work well enough.
If I still have to manually interfere with the timeline when using something advertised as an AI video editor, how much editing is the AI actually doing?
For the same reason, I don't want to use three different platforms for clipping, editing, and creating launch videos. I think those should be features of a single video editor, not separate products.
I like to describe Revline as a general-purpose AI video editor. It isn't limited to rough drafts or turning long videos into short clips. You can also create motion-design launch videos.
I specifically did not included hand-customizations ( like customizing VFX, transition, or even subtitle colors ), because once again, if you have to still hand-edit while using platform to reduce friction of editing, you might aswell save money not using it.
AI can already learn your styles, taste, remember throughout editing, that's the repeatable part every creator/non creator has to do while creating content frequently.
Under the hood
Revline loads agent with 100+ tools to edit video, i have defined set of lanes (Which anthropic calls "Progressive disclosure", but i call it "lanes") to load specific set of skills and tools only needed for specific edit you are working with, so each type of edit gets it's own "harness". Footage is analyzed using Deepgram, Gemini agentic-mode, or handing specific frames to agent directly ( depends on situation ), while motion graphics are generated as React/GSAP components ( hyperframes used as previewing and exporting engine). Rendering runs through FFmpeg and distributed workers. The same editing tools are also exposed over MCP, so Claude Code, Codex, and other agents can edit projects directly.
The entire editor is open-source under Apache 2.0. PRs, contributions, and feedback are welcome!
My ultimate goal is to build a video editing experience that existing platforms simply can't offer.
I'd love to hear what HN thinks, especially where you think an agent-first approach to video editing might fall short.
Website: https://tryrevline.com
GitHub: https://github.com/revlinehq/Revline
Comments URL: https://news.ycombinator.com/item?id=50380012
Points: 1
# Comments: 0
