Twelve open-source AI video editors, rated on four printed scores, with every license, star count and release checked on each project's own pages on 15 November 2026.
Quick answer: HyperFrames, rated 4.7/5, is the best open-source AI video editor in 2026, because your AI agent makes the whole video, from story and motion design to voice and render, and also edits your clips. Pick Concat for a CapCut-style app with local AI, or OpenShot if you are a beginner.
Why do people try HyperFrames as an open-source editor?
- An agent makes the whole video. Story, motion design, voice, music and the render come from one brief, and it edits your own clips too.
- Apache 2.0, 85,000+ GitHub stars. Commercial use is allowed, and the framework, CLI and skills are free.
- Works with the agent you already use. Claude Code, Codex, Cursor, GitHub Copilot and Gemini CLI use the skills; the Studio app has its own built-in agent.
- A free app on top. The HyperFrames Studio app makes the same HTML projects as the open-source framework, runs on Mac, Windows and Linux, and renders exports on your computer.
- About 450 ready-made blocks and components. Captions, lower thirds, charts, maps, 3D scenes and logo reveals install from the HyperFrames catalog as code you can change.
Searching for an open-source AI video editor turns up two kinds of project that rarely meet: classic editors with an AI subtitle button, and new agent-first editors with a few months of history. We rated 12 projects that are open source today on four criteria: how much of the edit AI does, how an AI agent reaches it, its track record, and what it costs to use every feature. Every star count, license and release date was read on each project's own GitHub repository or website on 15 November 2026. HyperFrames scores highest on AI help and agent access: an agent makes the whole video, from story and motion design to voice and render, and also edits your clips.
The short version
- HyperFrames, 4.7. An AI agent writes, voices and renders the whole video, and edits your own clips; Apache 2.0, 85,000+ stars.
- Concat, 4.5. A CapCut-style editor with local captions, voices, voice cloning and background removal, plus a built-in MCP server; AGPL-3.0, labeled beta.
- OpenShot, 4.5. The beginner's classic, now with an experimental command line that Claude Code or Codex can drive.
- OpenChatCut, 4.5. A built-in agent and outside agents edit one multitrack timeline, with transcript cuts and generation.
- WeftCut, 4.4. A desktop editor whose every tool is open to your agent over MCP, with a timeline diff of each change; a one-person project with 337 stars.
- Kdenlive, 4.4. The classic editor with the most AI tools here: Whisper or VOSK subtitles, translation and a SAM2 object mask.
- auto-editor, 4.4. Cuts silence and dead space from the command line, with agent skills and a Whisper command.
- Shotcut, 4.3. A light classic editor with speech-to-text and text-to-speech.
- OpenReel Video, 4.3. An editor in the browser with an AI Editor panel that runs on your own model key; alpha.
- video-use, 4.3. A Claude Code skill that turns a folder of raw takes into an edited video; needs an ElevenLabs key.
- LosslessCut, 4.1. Trims without re-encoding, so no quality is lost; no AI.
- LTX Desktop, 4.1. Generates AI clips on your own graphics card; beta, and it needs a big one.
Open-source AI video editors compared
Why we judge open-source editors this way
An open-source AI video editor is one whose code is public under a license approved by the Open Source Initiative, such as Apache 2.0, MIT, GPL or AGPL, and that uses AI for part of the work: captions, cuts, masks, generated clips, or an AI agent making the edits. Remotion's code is public, but its license is not an open-source license, so it has its own section below.
The question we asked of each project is the one a reader has: how much of a finished video will AI do for me here, and can the agent I already use do it? Stars measure attention, not fit, so they only break ties.
How we rated
Every project gets four sub-scores out of 5, and the overall rating is their average, rounded half up. Ties are broken by the unrounded average, then by GitHub stars. Each score comes from the project's own pages on 15 November 2026.
What decided this ranking
Agent access split the field. Kdenlive and Shotcut are mature, free and capable, but an outside agent can reach them only through small community MCP servers, so they score 3.5 there. The new agent-first editors score 4.6 to 4.7 and lose points on track record instead.
"Free" hides a key or a GPU. Every project is free to download, but video-use needs a paid ElevenLabs key for transcription, OpenChatCut's generation needs provider keys you pay for, and LTX Desktop asks for an NVIDIA card with 24 GB of video memory or an Apple Silicon Mac with 48 GB for local generation.
The 12 best open-source AI video editors, rated
#1 HyperFrames: Best for whole videos made and edited by an AI agent
Our rating: 4.7/5
- AI help: 4.9/5
- Agent access: 4.9/5
- Track record: 4.4/5
- Cost to use fully: 4.6/5
License and pricing: Apache 2.0, free for any company. The framework, CLI, skills and browser editor are open source; the HyperFrames Studio app is free to use with a free HeyGen account but is not open source, and its Pro plan costs $25 a month.
Standout feature: Your coding agent writes the video as HTML, CSS and animation code, HyperFrames renders it frame by frame, and every scene stays editable.
It covers the whole video. Its skills have the agent plan scenes, add voiceover and music, sound effects and word-timed captions, and build from about 450 catalog blocks and components for titles, transitions, charts and maps, kinetic type and 3D. On your own clips it transcribes speech on your computer, cuts long pauses, adds animated captions, removes the background with the free CLI cut-out and grades colour, all as layers over your untouched original.
npx hyperframes preview opens the open-source browser editor with a timeline, keyframes, a code view and a live preview. The Studio app puts the agent beside the preview and lets you point at the picture to ask for a change (how it works). Renders on your computer are free with no watermark, in MP4, WebM or MOV up to 4K, plus GIF, PNG frames and HLS from the CLI.
85,000+ stars and 148 contributors (counting anonymous ones) on 15 November 2026, version 0.9.55 released the same day, repository opened on 10 March 2026. It scores 4.4 rather than 4.8 because it is under a year old.
HyperFrames pros:
- An agent makes the whole video, not only the captions or the cuts
- Works from Claude Code, Codex, Cursor and Gemini CLI, a desktop app with a built-in agent, and a connector in Claude, ChatGPT and Grok
- Free local renders with no watermark, up to 4K, plus transparent video
- Your original footage never changes
HyperFrames cons:
- Not a drag-and-drop editor for cutting hours of footage by hand
- The Studio app itself is not open source
Choose HyperFrames if: you want your AI agent to make the video, from a launch film to a captioned talking-head cut, and keep editing it with you. Download HyperFrames Studio to try it.
What's new in HyperFrames: On 13 November 2026, the Studio app added a built-in HyperFrames agent that needs no install; the open-source framework shipped v0.9.55 on 15 November 2026.
Available for: macOS, Windows and Linux (CLI on Node.js 22 with FFmpeg; Studio app for Mac, Windows x64 and Linux).
#2 Concat: Best CapCut-style app with local AI
Our rating: 4.5/5
- AI help: 4.7/5
- Agent access: 4.6/5
- Track record: 4.0/5
- Cost to use fully: 4.7/5
License and pricing: AGPL-3.0, free, "No watermarks. No account. No paywall. Ever." A commercial license is offered for closed-source use. Concat on GitHub.
Standout feature: Auto-captions "made on your machine", text-to-speech with voice cloning, and background removal, in a multitrack editor for desktop and phones.
Concat's model list includes whisper.cpp for captions, Chatterbox Turbo for voices, RobustVideoMatting and rembg for cut-outs, Real-ESRGAN for enhancing and Spleeter for voice isolation. Optional models are about 2.8 GB and run on your machine. A built-in MCP server (concat --mcp) and a JSON-RPC API let agents and scripts drive it. The project opened on 25 August 2026 and already has 17,802 stars.
Concat pros:
- Several local AI tools with no account or key
- A built-in MCP server for agents
- Desktop, Android and sideloaded iOS builds
Concat cons:
- Labeled beta: "it works, and it still has edges"
- Windows and macOS builds are unsigned
Choose Concat over HyperFrames if: you want to cut social clips by hand in a familiar CapCut-style timeline, with AI captions and voices on your own machine.
What's new in Concat: On 11 November 2026, version 0.2.10 added voice isolation, in, out and loop animations and per-clip mask stacks; 0.2.11 followed on 13 November with fixes.
Available for: Windows x86_64, macOS (Intel and Apple Silicon), Linux, Android, and iOS and iPadOS as a sideloaded beta.
#3 OpenShot: Best for beginners, and agents that script edits
Our rating: 4.5/5
- AI help: 4.0/5
- Agent access: 4.3/5
- Track record: 4.8/5
- Cost to use fully: 4.7/5
License and pricing: GPL-3.0 or later, free. OpenShot on GitHub.
Standout feature: OpenShot's words: "our new command-line interface (CLI) lets you or an AI agent like Claude Code or Codex build entire videos."
OpenShot 4.0 (30 August 2026) isolates subjects with machine-learning models that run on your computer. OpenShot 4.1 (25 October 2026) added the command line, which returns JSON, thumbnail strips, waveforms and timeline previews and works headless or alongside the open editor. OpenShot labels it experimental.
OpenShot pros:
- A beginner-friendly timeline
- An official command line built for agents
- Developed since 2008, per its site, on Linux, Mac and Windows
OpenShot cons:
- One named AI feature, subject isolation
- The agent command line is experimental
Choose OpenShot over HyperFrames if: you want a classic beginner timeline you click through yourself, with an agent able to script it later.
What's new in OpenShot: On 25 October 2026, OpenShot 4.1 added direct text editing, Auto-Optimize for 4K and larger, the experimental command line and Linux ARM64 builds; 4.1.1 on 2 November added AMD AMF and HEVC export presets.
Available for: Linux, Mac and Windows.
#4 OpenChatCut: Best for chat-driven edits on a timeline
Our rating: 4.5/5
- AI help: 4.7/5
- Agent access: 4.7/5
- Track record: 4.0/5
- Cost to use fully: 4.4/5
License and pricing: AGPL-3.0 for the open-source edition, free; OpenChatCut Pro is a separate desktop product with a 7-day trial, after which the Agent and MCP need a license. OpenChatCut on GitHub.
Standout feature: "Let Codex, Claude Code, and the built-in agent read, edit, and export real video projects that remain fully editable."
A built-in agent and outside agents over a local MCP server edit one multitrack project, with word-level transcription, text-based cuts, pause handling, speakers and linked captions. It scores 4.7 on AI help rather than 4.9 because its pages describe editing and generating media, not designing scenes and motion. It can generate images, video, speech, music and sound effects through providers you connect, and npx skills add 0xsline/OpenChatCut installs 26 skills.
OpenChatCut pros:
- A real multitrack timeline that agents edit
- Transcript cuts and caption translation
- Open-source installers for Mac, Windows and Linux
OpenChatCut cons:
- Generation needs provider keys you pay for
- "These are early builds"; macOS packages are not signed
Choose OpenChatCut over HyperFrames if: your work is cutting recordings on a timeline and you want an agent making those cuts while you watch.
What's new in OpenChatCut: On 15 October 2026, version 0.2.17 let the built-in agent browse and import local media in the browser-hosted editor.
Available for: macOS (Apple Silicon and Intel), Windows x64 and Linux x64, or from source in the browser.
#5 WeftCut: Best for an agent driving a full editor over MCP
Our rating: 4.4/5
- AI help: 4.7/5
- Agent access: 4.6/5
- Track record: 3.8/5
- Cost to use fully: 4.6/5
License and pricing: MIT, free; it needs your own agent, such as Codex, Claude or any MCP client. WeftCut on GitHub.
Standout feature: "WeftCut exposes the editor itself as a tool surface": a localhost MCP server publishes every editing tool, and an Agent view shows each change as a timeline diff.
WeftCut transcribes to caption layers in the cloud or on your machine with whisper.cpp or FunASR, detects pauses and shot boundaries, describes clips with a vision model, and renders "Motifs", layers written in HTML, CSS or WebGL. The installer bundles no model weights.
WeftCut pros:
- Every tool open to your agent
- A timeline diff of each agent change
- Code-written layers on the timeline
WeftCut cons:
- "A one-person project", before version 1.0
- The macOS build is not notarized
Choose WeftCut over HyperFrames if: you want your agent inside a traditional editor and want to see each of its timeline changes as a diff.
What's new in WeftCut: On 10 November 2026, version 0.7.0 rendered Motifs across up to eight capture hosts, taking two Motifs "from under 4 fps each to 10-30 fps"; 0.7.1 followed on 11 November.
Available for: Windows 10 and 11 (64-bit), macOS Apple Silicon (13 or later) and Linux.
#6 Kdenlive: Best classic editor with AI subtitles
Our rating: 4.4/5
- AI help: 4.5/5
- Agent access: 3.5/5
- Track record: 4.8/5
- Cost to use fully: 4.7/5
License and pricing: GPL-3.0, free. Kdenlive on GitHub.
Standout feature: Speech to Text writes subtitles with Whisper or VOSK on your computer, SeamlessM4T translates them, and the Object Mask effect uses SAM2 to isolate a person or object.
Kdenlive is a classic multitrack timeline editor for long projects you cut by hand. Main development happens on KDE's GitLab; the GitHub mirror shows 6,116 stars and 235 contributors.
Kdenlive pros:
- Three AI tools: subtitles, translation and masks
- A full professional timeline
- A decade of releases
Kdenlive cons:
- Speech models must be installed in the settings first
- No official route for outside agents; community MCP servers only
Choose Kdenlive over HyperFrames if: you edit hours of footage by hand and want AI subtitles and masks inside a classic editor.
What's new in Kdenlive: On 7 November 2026, Kdenlive 26.08.3 improved recovery of corrupted projects, autosave and undo and redo speed.
Available for: Windows 10 or newer, Linux (AppImage, Flatpak) and macOS 13 or newer on Apple Silicon and Intel.
#7 auto-editor: Best for cutting silence automatically
Our rating: 4.4/5
- AI help: 3.5/5
- Agent access: 4.3/5
- Track record: 4.8/5
- Cost to use fully: 4.9/5
License and pricing: Unlicense (public domain) for everything in the repository; its separate GUI and online apps carry proprietary assets. auto-editor on GitHub.
Standout feature: A first pass that cuts "dead space", usually silence, by audio loudness or motion, then exports to Premiere, Resolve, Final Cut Pro, Shotcut, Kdenlive or EDL.
The cuts themselves use no AI model. A whisper subcommand transcribes with whisper.cpp models you download, and the subtitles can drive cuts. npx skills add WyattBlue/auto-editor installs skills for agents.
auto-editor pros:
- One binary, free, no keys
- Exports a timeline to your editor of choice
- Agent skills for each job
auto-editor cons:
- Command line only in the open-source repository
- AI is limited to the whisper command
Choose auto-editor over HyperFrames if: you need a first pass that removes silence from long recordings before you edit elsewhere.
What's new in auto-editor: On 13 November 2026, version 31.9.0 added Final Cut Pro XML import, every Premiere Pro transition, speed ramps and caption tracks.
Available for: Windows, macOS and Linux.
#8 Shotcut: Best lightweight classic editor
Our rating: 4.3/5
- AI help: 4.3/5
- Agent access: 3.5/5
- Track record: 4.8/5
- Cost to use fully: 4.7/5
License and pricing: GPL-3.0, free. Shotcut on GitHub.
Standout feature: "Codec independent so does not rely on system codecs", and it "can run as a portable app from external drive".
Shotcut turns speech into subtitles with whisper.cpp, with a GPU mode on Windows and Linux since 26.9, and reads text aloud with text-to-speech.
Shotcut pros:
- Codec independent, so it opens many formats
- Speech-to-text with GPU support
- Portable install
Shotcut cons:
- AI help covers speech and subtitles only
- No official route for outside agents
Choose Shotcut over HyperFrames if: you want a small classic editor that doesn't rely on system codecs.
What's new in Shotcut: On 27 September 2026, version 26.9.27 added Auto Duck, Adjustment Clip and Vulkan GPU speech-to-text, plus download security fixes.
Available for: Windows (including Windows on ARM), Mac and Linux.
#9 OpenReel Video: Best editor in the browser
Our rating: 4.3/5
- AI help: 4.4/5
- Agent access: 4.6/5
- Track record: 3.6/5
- Cost to use fully: 4.6/5
License and pricing: MIT, free; the AI Editor runs on your own OpenAI, Anthropic or compatible key, and "provider usage may cost money". OpenReel Video on GitHub.
Standout feature: An AI Editor panel that makes reversible timeline edits, in a browser editor with no installation.
The desktop build also supports MCP clients, so outside assistants can work on your project.
OpenReel Video pros:
- Nothing to install
- Reversible AI edits
- MCP on desktop
OpenReel Video cons:
- Alpha; the browser can clear its storage, so keep backups
- No release since 29 August 2026
Choose OpenReel Video over HyperFrames if: you want to edit in a browser tab with an AI helper and install nothing.
What's new in OpenReel Video: On 29 August 2026, version 1.0.0-alpha.17 improved caption workflows.
Available for: Web, plus alpha desktop builds for macOS, Windows and Linux.
#10 video-use: Best for letting Claude Code cut raw footage
Our rating: 4.3/5
- AI help: 4.7/5
- Agent access: 4.6/5
- Track record: 3.6/5
- Cost to use fully: 4.2/5
License and pricing: MIT, free; it needs an ElevenLabs API key for transcription, "one ElevenLabs Scribe call per source". video-use on GitHub.
Standout feature: "Drop raw footage in a folder, chat with Claude Code, get final.mp4 back."
Claude Code, Codex or any agent with a shell cuts filler words and dead space, grades colour, adds fades, burns in subtitles and generates overlays with HyperFrames, Remotion, Manim or PIL, then checks the render at every cut. It has 31,083 stars.
video-use pros:
- A whole edit from raw takes
- Self-checks every cut
- Works in several coding agents
video-use cons:
- A paid transcription key is required
- No tagged releases; last updated 8 October 2026
Choose video-use over HyperFrames if: you have a folder of raw talking-head takes and want Claude Code to cut them with ElevenLabs transcripts.
What's new in video-use: On 24 September 2026, it added phrase-aware captions; on 8 October the README pointed to a waitlist for a hosted version.
Available for: Any agent with shell access; its manual install uses Homebrew.
#11 LosslessCut: Best for trims with no quality loss
Our rating: 4.1/5
- AI help: 3.0/5
- Agent access: 4.0/5
- Track record: 4.8/5
- Cost to use fully: 4.7/5
License and pricing: GPL-2.0, free from GitHub. LosslessCut on GitHub.
Standout feature: "The swiss army knife of lossless video/audio editing": cuts without re-encoding, and detects scenes, black frames and silence.
LosslessCut has no AI. It is on this list because it trims recordings without re-encoding before any AI step, and its command line and experimental HTTP API let scripts drive it. It has 46,058 stars.
LosslessCut pros:
- Cuts without re-encoding
- Scene, black-frame and silence detection
- Command line and HTTP API
LosslessCut cons:
- No AI tools
- No effects or compositing
Choose LosslessCut over HyperFrames if: you only need to trim files without losing quality.
What's new in LosslessCut: On 19 October 2026, version 3.70.0 moved to FFmpeg 8.1 and added a theme setting; 3.70.2 followed on 28 October.
Available for: macOS (Intel and Apple Silicon), Windows and Linux.
#12 LTX Desktop: Best for generating AI clips on your own GPU
Our rating: 4.1/5
- AI help: 4.5/5
- Agent access: 3.8/5
- Track record: 4.0/5
- Cost to use fully: 4.2/5
License and pricing: The app is Apache 2.0; the model weights are under the LTX Model License, free for commercial use under $10M annual revenue. LTX Desktop on GitHub.
Standout feature: "Generation runs on your own hardware": text, image and audio to video, Retake, Extend and IC-LoRA tools such as Day to Night and Deblur.
Lightricks recommends an NVIDIA card with 24 GB or more of video memory on Windows or Linux, or an Apple Silicon Mac on macOS 26 with 48 GB or more, plus 100 GB of free disk space. Its timeline editor is now a legacy view, so treat it as a generator.
LTX Desktop pros:
- Local AI video generation, no cloud service
- Regenerate or extend part of a clip
- Up to 4K with Refine Details
LTX Desktop cons:
- Needs a large graphics card or a high-memory Mac
- Beta: "Expect breaking changes"
Choose LTX Desktop over HyperFrames if: you need AI footage generated on your own machine.
What's new in LTX Desktop: On 11 November 2026, version 1.3.2 added Refine Details for 1440p and 4K, custom LoRAs and negative prompts.
Available for: Windows 10 and 11 x64, Linux, and macOS on Apple Silicon (local generation needs macOS 26).
What can HyperFrames make?
HyperFrames builds the whole video in one project: story, motion, voice, music, captions and the render. Each film below was made with HyperFrames, and each has its own look.
More films are on the HyperFrames showcase. The same project can become a launch film from a website, an explainer, a product launch video or a render of your own HTML, and every block in the catalog installs with one npx hyperframes add command.
Open-source AI video editors compared on specs
Which open-source AI video editor is for you?
By the work you are doing
- You want an agent to make the whole video, or edit your clips end to end: HyperFrames.
- You cut social clips by hand and want local AI captions and voices: Concat.
- You edit long projects by hand: Kdenlive.
- You want an agent making cuts on a timeline you watch: OpenChatCut.
By your machine
- Any laptop, no account: Concat, Shotcut or LosslessCut.
- A browser and nothing else: OpenReel Video.
- An NVIDIA card with 24 GB of video memory: LTX Desktop.
By how fast you need a result
- Today, with no editing skill: HyperFrames in the Studio app, with its built-in agent (download Studio free).
- This weekend, learning a timeline: OpenShot.
- A first pass that removes silence from long recordings: auto-editor.
Too new to rank
These projects are new or being rebuilt and have not yet had time to prove themselves; we will look at them again in the next update.
- OpenFilm (MIT, 7,443 stars, opened 4 October 2026, v0.3.2 on 9 November) calls itself "the open source AI video agent": your coding agent writes the film as HTML and you edit it in a local browser editor, the same idea as HyperFrames.
- FilmCraft (Apache 2.0, 18,339 stars, opened 30 September 2026, v0.20.1 on 12 November) is an editor written in Rust where every action is a command, with an MCP server so Claude or any agent can cut, trim, grade, mix and export.
- OpenCut (MIT, 98,444 stars) is being rewritten from the ground up, with an MCP server for agents listed as coming; its classic repository is now archived.
Also worth knowing: Diffusion Studio's editor (MPL-2.0, 3,518 stars) is an editor installed through agent skills, OpenMontage (AGPL-3.0, 80,654 stars) is an agent production system that picks between Remotion and HyperFrames for each project, and Pireel (AGPL-3.0) is now an embeddable editor component whose host app supplies the AI models. Palmier Pro is no longer open source: its README says releases after v0.7.6 are proprietary.
Is Remotion open source?
Remotion is source-available, not open source. Its own license FAQ says: "Remotion is source-available software, but it is not open source software according to the Open Source Initiative's Open Source Definition." The code is on GitHub with 73,400+ stars, and individuals, non-profits and companies of up to 3 people can use it free, even commercially. Larger companies need a Company License, from $25 a month per seat or $0.01 per render with a $100 monthly minimum. HyperFrames is Apache 2.0, free for any company.
Remotion leads on React: if your team builds video as React components, it fits your code. The HyperFrames vs Remotion page shows the same video made with both, and our Remotion alternatives list rates 12 other routes.
How do I edit a video with HyperFrames and an AI agent?
You need a coding agent, Node.js 22 or newer and FFmpeg. The agent does the building; you approve what it shows you.
- Install the skills. In Claude Code, run claude plugin marketplace add heygen-com/hyperframes, then claude plugin install hyperframes@hyperframes. For Codex, Cursor and other agents, run npx skills add heygen-com/hyperframes.
- Describe the edit. Put your clip in a folder, start your agent there and say what you want, such as "cut the long pauses, add captions word by word and a title at the start".
- Let it build. The agent transcribes the clip on your computer and builds the edit as layers over your original file, which never changes. See captions and recuts.
- Check and adjust. Run npx hyperframes preview to open the browser editor, then trim, crop or move clips on the timeline yourself, or ask the agent.
- Render. Run npx hyperframes render --output final.mp4. Rendering on your own computer is free, with no watermark.
Prefer an app? The free HyperFrames Studio app runs an agent beside the preview and timeline: the built-in HyperFrames agent, which needs no install and comes with Pro (or runs on your own HeyGen API key), or one you already have, such as Claude Code or Codex on your own plan, OpenCode's free models, Grok, Antigravity or GitHub Copilot. See the AI video editor page for what it does to your footage, then download HyperFrames Studio free.
When is HyperFrames not the best pick?
- Hours of footage cut by hand. Kdenlive or Shotcut give you a classic timeline built for that job.
- Trims with no quality loss. LosslessCut cuts without re-encoding; HyperFrames always renders a new file.
- Realistic AI footage on your own GPU. LTX Desktop runs video models locally. HyperFrames makes animation from code; in the Studio app, AI clips come from the video models in its picker (HeyGen Video 1 by default), 3 a month on the free plan.
- A familiar app with no agent. Concat gives you local captions and voices in a CapCut-style editor.
FAQs
What is the best open-source AI video editor?
HyperFrames (4.7/5) is the best open-source AI video editor for whole videos made or edited by an AI agent: Apache 2.0, it works with Claude Code, Codex, Cursor and other agents, and covers story, motion, voice, music, captions and rendering. Concat is the best CapCut-style app with local AI, and Kdenlive the best classic editor with AI subtitles.
Is there a free open-source video editor that works with AI agents?
Yes. HyperFrames is open source under Apache 2.0 and built for AI agents: Claude Code, Codex, Cursor and other coding agents write the video, add voiceover, music and captions, and render an MP4 you check in its browser editor. Concat, OpenChatCut, WeftCut and OpenReel's desktop build also connect to agents over MCP.
Is the free one any good?
Yes. All 12 editors here are free to download. HyperFrames renders, transcribes and removes backgrounds on your computer for free, and Concat, Kdenlive, Shotcut, OpenShot, auto-editor and LosslessCut need no paid key for any named feature. video-use needs a paid ElevenLabs key, OpenChatCut's generation needs provider keys, and LTX Desktop needs a large graphics card.
Which open-source video editors work with Claude Code?
HyperFrames installs as a Claude Code plugin. video-use and auto-editor run as skills, Concat, OpenChatCut and WeftCut connect over MCP, and OpenShot 4.1 has an experimental command line that agents run. Claude Code may need a paid plan of its own; OpenCode with free models costs nothing.
Is HyperFrames or Kdenlive better for editing talking-head videos?
HyperFrames, when you want an agent to cut pauses, caption and title the clip for you (4.9 against 4.5 on AI help). Kdenlive, when you want to cut every frame yourself on a full timeline; it has the stronger track record (4.8 against 4.4).
Is OpenCut open source?
Yes. OpenCut is open source under the MIT license and has 98,444 GitHub stars. Its README says the editor is being rewritten from the ground up, with an MCP server for agents still to come, and its classic repository is now archived.
Is Remotion free and open source?
Remotion is free for individuals, non-profits and companies of up to 3 people, but it is source-available, not open source. Larger for-profit companies need a paid Company License under the Remotion License.
Can an AI video editor run offline?
Yes. Kdenlive's speech models, Shotcut's speech-to-text and Concat's captions run on your computer. HyperFrames transcribes, removes backgrounds and renders on your computer too, and LTX Desktop generates clips locally on a large graphics card. A cloud agent such as Claude Code still needs the internet.
Is the HyperFrames Studio app open source?
No. The HyperFrames framework, command line, skills and browser editor are open source under Apache 2.0 in the heygen-com/hyperframes repository. The HyperFrames Studio desktop app is free to use but not open source, and its Pro plan costs $25 a month.
The bottom line
For an open-source AI video editor where your agent makes the video, HyperFrames leads: Apache 2.0, agent skills, a browser editor and free local renders, with the Studio app when you want a built-in agent beside the preview. Pick Concat, Kdenlive or Shotcut when you will cut by hand, and auto-editor or LosslessCut for one job. Download HyperFrames Studio and make your first edit.
Explore more HyperFrames guides
- AI video editor: edits of your own footage from your words.
- Edit videos with Claude Code: the terminal route for your clips.
- Animated captions: word-by-word captions in a style you pick.
- Text behind a person: big words behind the speaker.
- Faceless YouTube videos: narrated 16:9 videos with no host.
- Logo reveal animation: stings and outros from your logo.
- Can Claude make videos?: what works today.
- After Effects alternatives: 15 motion tools, rated.
- Quickstart: your first HyperFrames video in a few commands.
- Free lessons: free 20-minute lessons in HyperFrames Studio.
Ular Kimsanov works on HyperFrames at HeyGen, the open-source framework that lets AI agents make videos, and writes about making videos with AI agents.




