The artists behind this site

Every picture and video in the walls on this site was made by a member of the Civitai community and posted publicly there: 84 pictures and 46 videos by 83 artists. The artists own their work. It is shown here with their name on it, and it comes down whenever they ask. Each name links to their profile, and each piece links to where they posted it. Go and see the rest of their work.

Is your work here?

If you'd rather it wasn't, ask for it to be removed and it will come off the site within three days, no questions asked. If you'd like your credit changed, the same address works.

Want Outmake to feature your art, for free?

Outmake wants to celebrate the people who make things with it. Over time, every picture and video on this site will be made in Outmake, by artists who are proud to show it here, each with their name on it.

Send a link to your work. Every submission is reviewed, and you'll hear back.

Submit your art

The voices on this site

Ono Anna, Sohee, Aiden, Serena, Dylan, Uncle Fu: the voices you can play in the Voice section, and the ones in the launch film, are Qwen3-TTS 12Hz 1.7B CustomVoice by the Qwen team at Alibaba (Apache 2.0). They ran on a Mac through the 8-bit MLX conversion that runs on a Mac by PowerBeef02. Ten languages from one model, with a warmth that carried a whole film.

The models in Outmake

Each of these teams released their models for anyone to download and run. That choice is what makes a local AI app possible. Every model's licence is shown on its page in Outmake before you download it.

Google

Mistral AI

Devstral Small 2 24BApache 2.0

Z.ai

GLM 4.7 FlashMIT

OpenAI

huihui-ai (base: Alibaba Qwen)

mlabonne (base: Alibaba Qwen)

Open Qwen 8BApache 2.0

huihui-ai (base: OpenAI)

Open GPT-OSS 20BApache 2.0

huihui-ai (base: Mistral AI)

Open Mistral 24BApache 2.0

Tongyi (Alibaba)

Z-Image TurboApache 2.0

Black Forest Labs

Lodestone (Chroma)

Chroma1 HDApache 2.0

Cagliostro Research Lab

Animagine XL 4.0CreativeML Open RAIL++-M

CircleStone Labs

Anima TurboCircleStone Labs Non-Commercial

Wan (Alibaba)

Lightricks

LTX-2.3LTX-2 Community License

The engines

The open-source code that runs every model in Outmake.

llama.cpp by ggml-org and its contributors
Runs the chat models. It made large language models practical on an ordinary computer, and much of local AI builds on it.
MIT
stable-diffusion.cpp by leejet
Runs the picture models, in plain C++, fast on Apple silicon.
MIT
MLX, mlx-swift and mlx-swift-lm by Apple
The machine learning framework built for Apple silicon's shared memory, under the voice engine.
MIT
mlx-audio-swift by Prince Canuma
The voice engine Outmake's speech helper is built from. One person's project, carrying a lot of voices.
MIT
sherpa-onnx by k2-fsa
Speaks the downloadable voices and hears dictation, with a huge, carefully packaged model library.
Apache 2.0
espeak-ng by the eSpeak NG contributors
Turns text into pronunciations for Piper, Kokoro and KittenTTS voices.
GPL 3.0 or later, as a separate program inside sherpa-onnx
swift-huggingface and swift-transformers by Hugging Face
Download and tokenize models from Swift.
Apache 2.0
TAEHV and TAEF2 by Ollin Boer Bohan
Tiny decoders that show a picture or video while it is still being made.
MIT

Voices in Outmake

The voice models in the app, and the people whose recorded voices they learned from.

Pocket TTS by Kyutai
The voice Outmake starts with: small, quick and natural.
CC BY 4.0
VCTK Corpus by Centre for Speech Technology Research, University of Edinburgh
CSTR VCTK Corpus, Centre for Speech Technology Research, University of Edinburgh. Ten of Outmake's built-in voices come from speakers p253, p260, p286, p292, p307, p333, p335, p336, p374 and p376, in Kyutai's enhanced versions.
CC BY 4.0
Unmute Voice Donation Project by Kyutai and the people who donated their voices
One built-in voice, from donation 4b70. Thank you to whoever recorded it.
CC0 1.0
Voice-Zero by Owen Tyme, from LibriVox readers
One built-in voice, from Peter Yearsley's LibriVox recording. LibriVox volunteers read public-domain books aloud for everyone.
CC0 1.0
Kokoro-82M by hexgrad
82 million parameters that sound like far more.
Apache 2.0
Qwen3-TTS by the Qwen team at Alibaba
Expressive voices in ten languages from one model.
Apache 2.0
Supertonic 3 by Supertone
Fast, clear speech.
OpenRAIL-M, with the use restrictions it lists
KittenTTS Nano and Mini by KittenML
Voices small enough for any Mac.
Apache 2.0
Piper voices by Rhasspy, and the people in each recording
A long-running open voice project, and the voices of Jenny (Dioco), Cori, Kristin and the other readers behind it.
Per voice: Jenny dataset licence; Cori and Kristin public domain (LibriVox); Northern English Male CC BY-SA 4.0; Lessac personal use only; Ryan CC BY-NC-SA 4.0; Alan and Amy not stated

Dictation and listening

The speech recognition models you can choose for dictation.

Parakeet V3 by NVIDIA
Accurate, fast dictation in European languages.
CC BY 4.0
Whisper Turbo by OpenAI
The open speech model that showed everyone what was possible.
MIT
Qwen3-ASR by the Qwen team at Alibaba
Small and multilingual.
Apache 2.0
Omnilingual ASR by Meta
Speech recognition for more than 1,600 languages, many of them never covered before.
Apache 2.0
Moonshine by Moonshine AI (Useful Sensors)
Built for fast transcription on small devices.
MIT

Building blocks

Libraries and ideas the app is built with.

Sparkle by Andy Matuschak, Mayur Pawashe, Kornel Lesiński and the Sparkle contributors
How Mac apps outside the App Store update themselves, for twenty years.
MIT
SwiftTerm by Miguel de Icaza
The terminal inside a chat.
MIT
DynamicNotchKit by Kai Azim
The shape of Outmake's notch is adapted from it.
MIT
LobeHub Icons by LobeHub
The makers' logos in the model pickers.
MIT (the drawings; the marks belong to their companies)
models.dev by SST
What the model pickers say about each model.
MIT
sandbox-runtime by Anthropic
Outmake's sandbox for an agent's commands follows the profile it generates.
Apache 2.0
T3 Code by T3 Tools
No code of theirs is included; the agent setup, terminal drawer and project import follow their ideas. Thank you for building in the open.
MIT
Agent Client Protocol adapters by the Agent Client Protocol project
Let Claude Code and Codex run inside Outmake.
Apache 2.0
Mermaid by Knut Sveidqvist and the Mermaid contributors
Draws the diagrams in replies.
MIT
Swift Argument Parser by Apple
Command-line handling for Outmake's helper programs.
Apache 2.0

Where models come from

The communities that host open models for everyone.

Hugging Face by Hugging Face and its community
Where nearly every open model lives, free to download. Outmake searches it from inside the app.
Civitai by Civitai and its community
Where people share picture models, styles and what they make with them.

This website

What this site and its videos are made with.

Lexend by Bonnie Shaver-Troup and Thomas Jockin, from Andrew Paglinawan's Quicksand, with the weight axis by Font Bureau
Every word on this site, in a typeface designed to make reading easier.
SIL Open Font License 1.1
Astro by the Astro team and contributors
Builds this site into plain, fast pages.
MIT
Tabler Icons by Paweł Kuna and contributors
The icons on this site.
MIT
Remotion by Jonny Burger and the Remotion team
Every video on this site, and the launch film, was made in React with it.
Remotion License
FFmpeg, x265 and SVT-AV1 by the FFmpeg developers, MulticoreWare and the Alliance for Open Media
Encode those videos small enough to load quickly.
LGPL 2.1 or later; GPL 2; BSD 3-Clause Clear
Pillow by Jeffrey A. Clark and contributors
The depth of field and repair passes on the videos.
MIT-CMU

Made by

Dane Bentley

Creator of Outmake. Founder of Ambedo Labs.

ambedo

n. a kind of melancholic trance in which you become completely absorbed in vivid sensory details—raindrops skittering down a window, tall trees leaning in the wind, clouds of cream swirling in your coffee—which leads to a dawning awareness of the haunting fragility of life, a mood whose only known cure is the vuvuzela.

John Koenig

The full licence texts that travel with the app are in Outmake's third-party notices. Outmake is not affiliated with or endorsed by any company or project named here; their names and marks belong to them.