The truly free, open-source, local-first, cross-platform CapCut replacement. Scriptable through a JSON-RPC API & MCP.
Third-party notices
FFmpeg
Concat links FFmpeg's libraries - libavformat, libavcodec, libavfilter,
libavdevice, libavutil, libswscale and libswresample - through the
ffmpeg-the-third crate. The Slint app spawns no ffmpeg or ffprobe
process.
FFmpeg is licensed under the LGPL-2.1-or-later; builds that include x264
(which the H.264 export uses) are GPL-2.0-or-later. Concat's own sources are
AGPL-3.0-or-later, and section 13 of the GPL-3.0 and AGPL-3.0 expressly
permits linking the two, so a distributed build may be conveyed on those
terms. Which FFmpeg a binary carries depends on the machine that built it:
Homebrew's on macOS, a BtbN gpl-shared build
(https://github.com/BtbN/FFmpeg-Builds) on Windows and Linux and in CI, and
on Android and iOS an LGPL build without x264, compiled from FFmpeg's source
by src/scripts/ffmpeg-mobile.sh and linked statically.
FFmpeg source code: https://ffmpeg.org/download.html
whisper.cpp
Transcription compiles whisper.cpp (https://github.com/ggml-org/whisper.cpp,
MIT) and ggml into the app through the whisper-rs crate. Whisper models
originate at https://huggingface.co/ggerganov/whisper.cpp (MIT); Concat
mirrors them and downloads them on demand from that mirror, and never
bundles them. See "The model mirror" below.
The one model Concat does bundle is Silero VAD v6.2.0 in whisper.cpp's
format (ggml-silero-v6.2.0.bin, from https://huggingface.co/ggml-org/whisper-vad,
MIT; Silero VAD itself is MIT, https://github.com/snakers4/silero-vad,
Copyright (c) 2020-present Silero Team). It finds where the speech is, so
captions are timed from it.
Slint (used under GPL-3.0-only)
The concat crate (src/crates/concat) builds against
Slint, which its authors offer under
any one of three licences, at the user's choice: a Royalty-free licence, a
paid commercial licence, or GNU GPL-3.0-only.
Concat uses Slint under the GPL-3.0-only option. That choice is deliberate and it is recorded here because nothing in the source tree would otherwise say which of the three applies. The Royalty-free and commercial options are not used: both are aimed at shipping proprietary applications, and neither can grant downstream recipients the freedoms Concat's own licence promises them.
Section 13 of the GPL-3.0 exists for exactly this combination:
Notwithstanding any other provision of this License, you have permission to link or combine any covered work with a work licensed under version 3 of the GNU Affero General Public License into a single combined work, and to convey the resulting work.
So the combined binary is conveyable. Slint's portion remains GPL-3.0-only, Concat's portion remains AGPL-3.0-or-later, and the AGPL's section 13 network requirement applies to the combination as a whole. Anyone forking Concat who would rather not be bound by the GPL must take Slint under one of its other two licences and remove or replace Concat's AGPL-licensed code accordingly; the two cannot be mixed.
Slint pulls in its Skia renderer (see src/crates/concat/Cargo.toml) along
with winit and their transitive crates, which are
predominantly MIT/Apache-2.0/BSD licensed. The Skia renderer crate,
i-slint-renderer-skia, is built from a patched copy in
src/patches/i-slint-renderer-skia; its changes are listed in
src/patches/README.md and marked Concat: in the source.
cargo tree -p concat gives the resolved set of any given build.
Fonts
The window and the title painter embed the interface's font into the binary
(src/crates/concat/build.rs, EmbedResourcesKind::EmbedFiles, and
include_bytes! in concat-text), so a distributed binary carries it and its
licence travels with it. The files and the full licence text are in
src/crates/concat-text/fonts/.
- Hanken Grotesk: Copyright 2021 The Hanken Grotesk Project Authors
(https://github.com/marcologous/hanken-grotesk), SIL Open Font License
1.1. The Regular, Medium, SemiBold, Bold and Italic static instances are
embedded. See
fonts/LICENSE-HankenGrotesk.txt.
These are the base fonts every title can use, embedded by concat-text
(BASE_FONTS) and registered with the window for the text presets. Each
file's licence is beside it in src/crates/concat-text/fonts/, as
LICENSE-<Family>.txt; OFL is the SIL Open Font License 1.1.
- Inter: Copyright 2016 The Inter Project Authors (https://github.com/rsms/inter). OFL. See
fonts/LICENSE-Inter.txt. - Montserrat: Copyright 2024 The Montserrat Project Authors (https://github.com/JulietaUla/Montserrat). OFL. See
fonts/LICENSE-Montserrat.txt. - Space Grotesk: Copyright Florian Karsten. OFL. See
fonts/LICENSE-SpaceGrotesk.txt. - Archivo Black: Copyright 2017 The Archivo Black Project Authors (https://github.com/Omnibus-Type/ArchivoBlack). OFL. See
fonts/LICENSE-ArchivoBlack.txt. - Bebas Neue: Copyright 2010 Dharma Type. OFL. See
fonts/LICENSE-BebasNeue.txt. - Anton: Copyright 2020 The Anton Project Authors (https://github.com/googlefonts/AntonFont). OFL. See
fonts/LICENSE-Anton.txt. - DM Serif Display: Copyright 2014-2018 Adobe, with Reserved Font Name 'Source'; Copyright 2019 Colophon Foundry. OFL. See
fonts/LICENSE-DMSerifDisplay.txt. - Abril Fatface: Copyright 2011 TypeTogether. OFL. See
fonts/LICENSE-AbrilFatface.txt. - Bangers: Copyright 2010 The Bangers Project Authors. OFL. See
fonts/LICENSE-Bangers.txt. - Permanent Marker: Copyright Font Diner. Apache License 2.0. See
fonts/LICENSE-PermanentMarker.txt. - Pacifico: Copyright 2018 The Pacifico Project Authors. OFL. See
fonts/LICENSE-Pacifico.txt. - Lobster: Copyright 2010 The Lobster Project Authors, with Reserved Font Name "Lobster". OFL. See
fonts/LICENSE-Lobster.txt. - Caveat Brush: Copyright 2015 Google Inc. OFL. See
fonts/LICENSE-CaveatBrush.txt. - Staatliches: The Staatliches Project Authors. OFL. See
fonts/LICENSE-Staatliches.txt. - Black Ops One: The Black Ops One Project Authors. OFL. See
fonts/LICENSE-BlackOpsOne.txt. - Germania One: The Germania One Project Authors. OFL. See
fonts/LICENSE-GermaniaOne.txt. - Atomic Age: The Atomic Age Project Authors. OFL. See
fonts/LICENSE-AtomicAge.txt. - Hammersmith One: The Hammersmith One Project Authors. OFL. See
fonts/LICENSE-HammersmithOne.txt. - Secular One: The Secular One Project Authors. OFL. See
fonts/LICENSE-SecularOne.txt. - Londrina Solid: The Londrina Solid Project Authors. OFL. See
fonts/LICENSE-LondrinaSolid.txt. - Bungee Hairline: The Bungee Project Authors. OFL. See
fonts/LICENSE-BungeeHairline.txt. - Rubik Scribble: The Rubik Scribble Project Authors. OFL. See
fonts/LICENSE-RubikScribble.txt. - Shadows Into Light Two: The Shadows Into Light Two Project Authors. OFL. See
fonts/LICENSE-ShadowsIntoLightTwo.txt. - Pixelify Sans: The Pixelify Sans Project Authors. OFL. See
fonts/LICENSE-PixelifySans.txt. - Google Sans Code: The Google Sans Code Project Authors. OFL. See
fonts/LICENSE-GoogleSansCode.txt. - Kosugi Maru: Copyright MOTOYA Co., Ltd. Apache License 2.0. See
fonts/LICENSE-KosugiMaru.txt. - Frijole: The Frijole Project Authors. OFL. See
fonts/LICENSE-Frijole.txt. - Emilys Candy: The Emilys Candy Project Authors. OFL. See
fonts/LICENSE-EmilysCandy.txt. - Mystery Quest: The Mystery Quest Project Authors. OFL. See
fonts/LICENSE-MysteryQuest.txt. - Zen Tokyo Zoo: The Zen Tokyo Zoo Project Authors. OFL. See
fonts/LICENSE-ZenTokyoZoo.txt.
The Noto faces in src/crates/concat-text/fonts/ are the fallback for the
scripts the other faces do not draw, in the interface and in titles and
captions alike, compiled into the binary once
(src/crates/concat-text/src/lib.rs, SCRIPT_FALLBACKS). Each is the
Regular face from the notofonts project's own release, cut down with
fonttools' pyftsubset; the release, the SHA-256 of the archive it came
in and the command that cut it are recorded beside each face there. A subset is a Modified
Version under the OFL, and keeps the family name the release gave it, as
the OFL allows for Noto, which reserves no font name.
- Noto Sans CJK SC (a subset: GB 2312, Big5 level 1, JIS X 0208, the Hangul of KS X 1001 and the kana): Copyright 2014-2021 Adobe (http://www.adobe.com/), with Reserved Font Name 'Source'. OFL. See
src/crates/concat-text/fonts/LICENSE-NotoSansCJK.txt. - Noto Sans Arabic UI 2.011 (a subset: the Arabic block and the characters the Arabic, Persian and Urdu interface uses): Copyright 2022 The Noto Project Authors (https://github.com/notofonts/arabic). OFL. See
src/crates/concat-text/fonts/LICENSE-NotoSansArabicUI.txt. - Noto Sans Hebrew 3.001 (a subset: the Hebrew block): Copyright 2022 The Noto Project Authors (https://github.com/notofonts/hebrew). OFL. See
src/crates/concat-text/fonts/LICENSE-NotoSansHebrew.txt. - Noto Sans Devanagari 2.007 (a subset: the Devanagari block): Copyright 2022 The Noto Project Authors (https://github.com/notofonts/devanagari). OFL. See
src/crates/concat-text/fonts/LICENSE-NotoSansDevanagari.txt. - Noto Sans Bengali 3.011 (a subset: the Bengali block): Copyright 2022 The Noto Project Authors (https://github.com/notofonts/bengali). OFL. See
src/crates/concat-text/fonts/LICENSE-NotoSansBengali.txt. - Noto Sans Thai 2.002 (a subset: the Thai block): Copyright 2022 The Noto Project Authors (https://github.com/notofonts/thai). OFL. See
src/crates/concat-text/fonts/LICENSE-NotoSansThai.txt.
The licence does not permit selling the font on its own; shipping the
fonts/ directory as it stands satisfies it.
sherpa-onnx and Kokoro voices
Text to speech uses the sherpa-onnx runtime (https://github.com/k2-fsa/sherpa-onnx, Apache-2.0), linked statically on macOS and Linux, and on Windows, Android and iOS shipped with the app as the prebuilt shared libraries k2-fsa publishes. It carries onnxruntime (MIT), piper-phonemize (MIT) and espeak-ng (GPL-3.0-or-later, https://github.com/espeak-ng/espeak-ng) for grapheme-to-phoneme conversion. Because espeak-ng is compiled into the app binary or into the sherpa-onnx libraries shipped with it, distributed builds must comply with the GPL-3.0 for that combined work. Concat's own sources are AGPL-3.0-or-later; section 13 of both GPL-3.0 and AGPL-3.0 expressly permits that combination, so the combined binary may be conveyed on those terms.
Kokoro voice model bundles (Apache-2.0, https://huggingface.co/hexgrad/Kokoro-82M) originate in the sherpa-onnx releases - including espeak-ng's data files. Concat mirrors those bundles and downloads them on demand from that mirror, and never bundles them with the app. See "The model mirror" below.
The Pocket TTS bundle is Kyutai's Pocket TTS (CC-BY-4.0, https://huggingface.co/kyutai/pocket-tts, https://kyutai.org) as exported to ONNX by KevinAHM (https://huggingface.co/KevinAHM/pocket-tts-onnx, CC-BY-4.0) and packaged in the sherpa-onnx releases, with two sample recordings from Kyutai's delayed-streams-modeling repository as its named voices. It is downloaded on demand like the Kokoro bundles. Reading in the voice of a recording is a use the person doing it is answerable for: Concat offers it for the narrator's own voice, and Kyutai asks that no one's voice be cloned without their consent.
Chatterbox Turbo (MIT, https://huggingface.co/ResembleAI/chatterbox-turbo-ONNX, Resemble AI) is driven through ONNX Runtime directly, from Resemble's own ONNX export, downloaded on demand like the other voices. Its GPT-2 tokenizer is read by the tokenizers crate (Apache-2.0). Resemble's Python package stamps a Perth watermark on what it speaks; the ONNX pipeline carries no such step and Concat adds none, which MIT permits and which anyone passing the sound off as a person's should know is not there.
ONNX Runtime
The cutout models run on Microsoft's ONNX Runtime
(https://github.com/microsoft/onnxruntime, MIT), through the ort crate
(https://github.com/pykeio/ort, MIT OR Apache-2.0). On Apple silicon and
the phones it is linked into the app from pyke's builds of it; an Intel
Mac's bundle carries Microsoft's last shared build for that machine in
Contents/Frameworks. The Linux and Windows bundles ship Microsoft's own
shared build of ONNX Runtime 1.28.0 beside the binary - lib/libonnxruntime.so.1.28.0 on Linux,
onnxruntime.dll beside concat.exe on Windows, one runtime that the
ort crate and sherpa-onnx share - unmodified, as Microsoft publishes it
at https://github.com/microsoft/onnxruntime/releases/tag/v1.28.0. Its
licence (MIT, Copyright (c) Microsoft Corporation) and third-party
notices are in that release's archive.
Microsoft Visual C++ runtime (Windows)
The Windows bundles and installers carry the Microsoft Visual C++ runtime
DLLs the program and its DLLs import - vcruntime140.dll,
vcruntime140_1.dll, msvcp140.dll and their kin, whichever the build
finds imported - copied unmodified from the redistributable folder of the
Visual Studio that built the release (VC\Redist\MSVC\<version>\<arch>\ Microsoft.VC14x.CRT). They are Microsoft's, not Concat's, and are not
covered by Concat's licence: they are redistributed under the terms
Microsoft sets for them, the Microsoft Software License Terms for Visual
Studio and its list of distributable code (for Visual Studio 2022,
https://learn.microsoft.com/visualstudio/releases/2022/redistribution),
which permit shipping them alongside an application that uses them.
The cutout models
Remove background runs four models, none of which ship inside the app except the first:
- Google's MediaPipe Selfie Segmentation (Apache-2.0), in the ONNX
conversion published by the ONNX Community
(https://huggingface.co/onnx-community/mediapipe_selfie_segmentation,
Apache-2.0), compiled into the
concat-visioncrate; seesrc/crates/concat-vision/models/NOTICE.md. The answer when nothing has been downloaded. - Robust Video Matting, the MobileNetV3 variant, by Peter Lin and others (https://github.com/PeterL1n/RobustVideoMatting, GPL-3.0), from that repository's releases, downloaded on first use. The person model.
- IS-Net from "Highly Accurate Dichotomous Image Segmentation" by Qin and others (https://github.com/xuebinqin/DIS, Apache-2.0), in the ONNX export the rembg project publishes (https://github.com/danielgatis/rembg, MIT), downloaded on first use. The object model.
- SlimSAM (https://github.com/czg1225/SlimSAM, Apache-2.0), in the ONNX export published at https://huggingface.co/Xenova/slimsam-77-uniform (Apache-2.0), downloaded on first use. The brushes' model.
Downloaded models live in the app's data directory under cutout-models
and are never bundled; the three that are downloaded come from Concat's own
mirror of the sources named above, described below. They are all run by
ONNX Runtime
(https://github.com/microsoft/onnxruntime, MIT) through the ort crate
(https://github.com/pykeio/ort, MIT OR Apache-2.0), with the platform's
own accelerator behind it where the runtime has one: CoreML on macOS and
iOS, NNAPI on Android. How each platform carries the runtime is under
"ONNX Runtime" above.
The model mirror
Every downloaded model named above is mirrored onto a release of Concat's own
repository, and the app fetches it from there; the upstream each was
obtained from is the fallback and is named above in every case. The table of
what is mirrored, and the digest each download is checked against, is
models/manifest.toml; .github/workflows/models.yml is what fills the
mirror.
Mirroring is redistribution, and each model keeps the licence it arrived under - Apache-2.0 for the Kokoro bundles, SlimSAM and IS-Net, CC-BY-4.0 for Pocket TTS, BSD-3-Clause for Real-ESRGAN, MIT for Chatterbox Turbo and the whisper conversions, GPL-3.0 for Robust Video Matting. Those terms are met by the attributions above and by the licence files each mirrored archive carries; a mirrored file is a verbatim copy, never a modification. Nothing in the mirror is bundled with the app, and Concat claims no rights over any of it.
Effect preview photograph
The effect catalogue cards are drawn from a photograph by Sou Jest on
Unsplash (https://unsplash.com/@soujest,
https://unsplash.com/photos/2v921tlqdXE), used under the Unsplash License. The still lives at src/crates/concat-host/assets/card-still.jpg
and is embedded in the app; each card is the effect's own shader run over
it on the GPU when the app first needs the card (concat-host's
cards.rs).