Concat video editor with a timeline, captions and a preview window open

Concat: The Free Open-Source CapCut Alternative That Runs on Your Own Machine

Concat: The Free Open-Source CapCut Alternative That Runs on Your Own Machine

TL;DR: Concat is a free, open-source (AGPL-3.0) video editor for Windows, macOS, Linux and Android, with iOS builds in beta. It covers the jobs people open CapCut for: auto-captions, text-to-speech, voice cloning, background removal, keyframes and 4K export, all of it on your own machine. CapCut Pro is $19.99 a month or $179.99 a year as of 15 October 2026. Concat costs nothing, with no watermark and no account. It is also a beta, and its installers are not code-signed.

The AI features are what turned video editing into a subscription. CapCut's free plan still cuts and exports video, but the caption, voice and generation tools that made the app famous sit on paid plans: $19.99 a month, $179.99 a year, or $79 a month for Ultra, metered in monthly AI credits. DaVinci Resolve is free until you want the Studio-only tools, which are $295 once. Every editor in this corner of the market found the same door.

Concat is a project that went the other way. It puts the same class of features into a desktop app and runs them locally: captions from a Whisper model on your own disk, voices synthesised or cloned on your own hardware, a cutout model that removes a background without a single frame leaving the machine. The repository appeared on 25 August 2026, and by mid-October it had around 6,500 GitHub stars and 113,000 downloads across seven releases. Version 0.2.7 landed on 13 October 2026.

What Concat Actually Does

This is a desktop non-linear editor, not a browser toy. The feature list reads like a list of things you would otherwise pay for:

  1. Cut on a timeline with split, trim, ripple delete, merge, freeze frame and speed controls, with a magnetic timeline you can switch on or off.
  2. Generate captions locally with Whisper. You pick a model size and a language build, and styled captions land on the timeline as text you can edit.
  3. Turn typed words into speech with the built-in voices, then clone a voice from a few seconds of someone talking and use it in the same project.
  4. Remove backgrounds from people or objects, or paint the mask yourself by hand.
  5. Keyframe position, scale, rotation, opacity, volume and effect parameters, with a curve editor for the timing.
  6. Apply more than 170 effects, filters, transitions and text animations, rendered live in the preview.
  7. Style titles with fonts, stroke, shadow and a background plate, starting from presets.
  8. Clean up audio with one switch: denoise, voice enhance, loudness levelling. The joke effects are in there too.
  9. Export as H.264, HEVC or AV1, up to 4K at 60fps and 10-bit colour.
  10. Work in multi-track and multi-timeline projects with blend modes, crop and flips.

For anyone who has watched a Shorts workflow end to end, that list is the workflow. The difference is where each step happens.

The Local Part Is the Whole Point

Every AI feature in Concat runs against a model that lives on your disk. The models download once from Settings, and after that the app works with the network unplugged. Their sizes are small enough to plan around:

Model family Size What it does
Captions (Whisper) 78MB, 148MB or 488MB each Tiny, Base or Small, in English or multilingual builds
Text-to-speech (Kokoro) 132MB or 350MB The built-in speakers
Voice cloning (Pocket TTS) 98MB A voice from a few seconds of recording
Voice cloning (Chatterbox) about 1.1GB The studio-grade read, in nine files
Cutout 15MB, 179MB or 40MB People, objects, or a mask you paint
Enhance 5MB Picture restoration

Install every optional model and you are looking at roughly 3.4GB. The app itself starts at about 500MB with the smallest caption model.

Nothing in that pipeline needs a server, an account or a credit balance, which is the practical difference between Concat and a cloud editor. Client footage under an NDA, footage of your own kids, an unreleased product demo: none of it should sit in someone else's storage bucket for a caption track. There is also no meter running while you experiment, so you can try six takes of the same thirty seconds without doing arithmetic first.

One honest note about where the intelligence comes from. Concat did not train these models. It wires open ones together (Whisper for speech, Kokoro plus the Pocket and Chatterbox voices for speech synthesis, and open cutout models for masking) behind a native Rust engine with a GPU compositor, proxies and hardware decode. That is why 4K timelines scrub smoothly and why the app can be free: the expensive parts are public research already, and the work here is the editor built around them.

The API Almost No Editor Has

Editors traditionally have no automation story. You click, or you do not edit. Concat ships a JSON-RPC API, a gRPC API, an MCP server and a CLI, all documented on its developer docs site, which means a script or an AI agent can drive an edit: import clips, place them, add captions, apply a look, render.

If you already automate with n8n, or you run an agent that produces content, the difference between a tool that exposes an MCP endpoint and one that does not is the difference between a pipeline and a chore. Of the five editors in the table below, Concat is the only one that speaks MCP.

How Concat Compares to CapCut and the Free Editors

Tool Price (checked 15 Oct 2026) Licence Works offline Editor builds Scriptable
Concat Free AGPL-3.0 Yes, local AI models included Windows, macOS, Linux, Android, iOS beta CLI, JSON-RPC, gRPC, MCP
CapCut Free tier; Pro $19.99/mo or $179.99/yr; Ultra $79/mo Proprietary Basic editing only, AI features are cloud-side and credit-metered Windows, macOS, Android, iOS No
DaVinci Resolve Free; Studio $295 one-off Proprietary Yes Windows, macOS, Linux, iPad Python and Lua scripting
Shotcut Free GPL-3.0 Yes Windows, macOS, Linux No
Kdenlive Free GPL-3.0 Yes Windows, macOS, Linux No

Two things are worth saying plainly, because a table can imply more than it should.

First, local captions are not unique to Concat. Shotcut has shipped Speech to Text since version 24.10 in late 2024, running Whisper through whisper.cpp, and Kdenlive generates subtitles on your own machine as well. What Concat adds on top of that is the rest of the stack in one app: voice cloning, background removal, the effects and keyframe system, and phone builds that open the same project files as the desktop.

Second, Concat is not a threat to DaVinci Resolve for anyone grading a feature film, or to CapCut for anyone who lives inside templates and cloud collaboration. Winning on price does not make a tool the deepest one in the room. Concat's argument is narrower than that: for short-form edits, in an app that runs on your own machine, it covers what the paid plans charge for.

Where Concat Falls Short

The project is candid about being in beta, and so is this list.

  • It is beta software at version 0.2.7, with about seven weeks of releases behind it. Expect rough edges and the occasional failed export rather than a finished product.
  • The builds are unsigned. Windows SmartScreen stops the first launch until you click through it, macOS may claim the app is damaged until you clear the quarantine flag in Terminal, Android asks you to allow installs from this source, and iOS is not on the App Store at all.
  • The App Store and Microsoft Store releases are on the roadmap under items that need capital. So is a marketplace for custom effects. None of it exists yet.
  • The licence is AGPL-3.0, which is real open source with real copyleft terms. If you plan to embed the editor inside something you ship, read the licence first.
  • Sponsorship is what funds development. The maintainer, jub0t, accounts for the overwhelming majority of commits, with a small group of contributors around him. That is a different risk profile from a company with a support contract.
  • No account also means no cloud project sync, no shared libraries across devices and no collaboration. Two people cannot open the same timeline the way cloud projects allow.
  • The AI models are extra downloads of up to about 3.4GB, and the recommended spec for 4K work and the larger caption models is 16GB of RAM plus a GPU with Metal, DirectX 12 or Vulkan. The CPU fallback works, but it is slower.
  • It carries 118 open issues at the time of writing, which is normal for a project this young and also means there is a queue.

How to Get Started

The download picker on the project site detects your system and hands you the right build, or you can take what you want from GitHub Releases. Each release publishes a SHA256SUMS file if you want to verify what you downloaded.

On macOS, expect the quarantine warning and clear it once:

xattr -dr com.apple.quarantine /Applications/Concat.app

On Linux, the packages install the usual way:

chmod +x Concat-*.AppImage           # AppImage
sudo apt install ./Concat-*.deb      # Debian and Ubuntu
sudo dnf install ./Concat-*.rpm      # Fedora and openSUSE
sudo pacman -U Concat-*.pkg.tar.zst  # Arch

Android takes the APK. iOS and iPadOS take an IPA by sideloading with AltStore, Sideloadly or whatever tool you already use for that.

The first launch has nothing to sign into. You get a start screen with New project on one side and your recent projects on the other. Open Settings, pull down whichever caption model suits your machine, and start cutting. Base is the sensible first pick for captions, and the Small multilingual model is worth the extra download if you work in more than one language.

Who's Behind It

Concat is maintained by jub0t, with contributions from a small group of developers and a Discord server where contributors can claim a role. Paid sponsorships keep the lights on, with Proxyon currently holding the first slot. The project's own roadmap is unusually blunt about what comes next: raise capital, ship to the Apple App Store and the Windows Store, and build a place where people can share and sell their own effects. The costs behind that list are published too: $99 a year for the Apple developer programme, $25 once for Google Play, and $50 to $100 a month for the website and the community asset store.

The Bottom Line

Concat is the free CapCut alternative that behaves like software you own. Captions, voices, cutouts, effects and 4K export all run on the machine in front of you, nothing uploads, and there is no account to create. For short-form work, that covers the features people were paying $19.99 a month for.

Weigh that against the beta label and the unsigned installers. If you need a signed, store-installed editor, or you collaborate through the cloud, wait a release or two and use CapCut's free tier for those parts. If you would rather stop renting your editor, Concat is worth the twenty minutes it takes to install.

Related Tools

Concat is also catalogued in our directory with the model sizes and platform list on one page. For the other pieces of the same workflow, RMBG-2 Studio handles background removal on stills, Voicebox and DramaBox TTS cover voice work, and OBS Studio and Capptivo handle the recording side. If you would rather grade a film than cut a Short, our DaVinci Resolve and Kdenlive pages are the deeper end.

Looking for more free tools? We covered seven free Adobe alternatives rebuilt in Rust, the free open-source alternative to Topaz Gigapixel, and a stack of free alternatives to popular paid subscriptions.

Enjoyed this article?

Share it with your network