Local-first AI video studio · macOS & Windows

Generate, edit and dub AI video on your own desktop

Mosael is a desktop app that takes a video from idea to upload. Generate images, video and music with your own API keys for Seedance, Kling, Veo, Wan and more, or with the ComfyUI you already run. Cut on a multi-track timeline, clone a voice, translate and lip-sync a video, make a photo talk, and turn a product photo into a narrated short with ready-made workflow templates.

Download Mosael View source on GitHub

Free for personal, non-commercial use · macOS (Apple silicon) · Windows 10/11 x64 · Commercial license

Block out cameras and moves in 3D, then hand the shot to an image or video model as a reference.

Explore 3D previs

What is Mosael

One app between the idea and the upload

Making an AI video usually means a generator in one tab, an editor in another, a voice tool, a subtitle tool and a folder full of downloads. Mosael puts generation, editing, voice, automation and publishing in one desktop app. You connect the model providers you already pay for — Mosael doesn't resell credits — and your projects and media stay on your computer by default.

01 / AI generation

Seedance, Kling, Veo and more — with your own keys

AI Studio generates images, video, music and sound effects from text, a first and last frame or reference images. Pick any model from the providers you connect — Volcengine Ark, Kling, Google, Alibaba Bailian, MiniMax, OpenAI, Evolink and others — and the controls follow the model you picked. Results land in your media library and remember what they were made from.

Text-to-video, image-to-video, first/last frame
Images, music and sound effects in one panel
Spending recorded from what providers report

Explore AI Studio

Mosael AI Studio generate panel picking a LoRA from a ComfyUI workflow, with previews, base model and trigger words

02 / ComfyUI

Your ComfyUI, as a model, a tool and an app

Connect a ComfyUI server on this computer or on your LAN. Every saved workflow becomes a model you can pick in AI Studio and on boards, and a tool the agent and your workflows can run. Browse the server's model files by type with preview images, base model and trigger words, download what's missing, import workflows and fill in missing models and nodes. It doesn't replace ComfyUI's own interface; it puts your workflows next to cloud models, the editor and automation.

Saved workflows become models and tools
Model library with previews and trigger words
Workflow library: import, complete missing models and nodes
App forms and a workbench: only the fields others fill in, models docked by the canvas

Explore the ComfyUI integration

Mosael's ComfyUI model library listing checkpoints, LoRAs and VAEs with previews, base model and file size

03 / Editing & voice

A real timeline, with subtitles and your own voice

Cut picture and sound on multiple tracks, edit by deleting words from the transcript, import or export .srt / .vtt subtitles and stack bilingual captions. Dub a subtitle track line by line with a voice you cloned locally (F5-TTS or Fish Speech), with CosyVoice in your own Alibaba Cloud account, or with free Edge voices. Grade with curves, presets and LUTs, then export with loudness normalization.

Transcript-based editing
Bilingual subtitles, .srt / .vtt in and out
Local voice cloning and line-by-line dubbing
Color grading with LUTs

See editing and voice-over

Mosael editor Voice panel dubbing a subtitle track with a locally cloned voice (F5-TTS), above video, audio and subtitle tracks

04 / Digital humans & dubbing

Make a photo talk. Translate a video, lips included.

Turn a portrait and a voice-over into a talking video with OmniHuman, Kling, HeyGen, Hedra or Wan, or re-sync the lips of existing footage. The translated-dubbing templates transcribe, translate, lay subtitles, dub every line in a cloned voice and can re-sync the speaker's lips. Real people need a consent declaration first, and exports that contain AI-generated content are labelled.

Talking photos and lip-sync
Video translation: subtitles, dub and lip-sync
Consent checks and AI-generated labels

Explore digital humans

Making a character asset talk: type the line and pick a video model that does talking photos

05 / Workflow automation

Build the pipeline once, run it again and again

Connect models, media and tools on a visual node canvas. Mosael checks every required input before anything runs and points out references that could never resolve. Run a flow by hand, on a schedule or from a webhook, or let the agent call it — and start from an official template instead of a blank canvas.

Visual node editor with readiness checks
Manual, scheduled or webhook runs
Official templates you can import and edit

Explore visual workflows

Mosael workflow canvas connecting generation, media, and notification nodes

06 / AI infinite canvas

Lay the whole project out on an infinite canvas

Put scripts, references, characters, 3D shots and generations on one board. A connection passes text and images into a generation cell; cells transcribe, translate, denoise or split a grid on their own; a timeline cell gives you a rough cut without leaving the board. Comments and markers keep review anchored.

Generate right on the canvas
Cells that do their own work
A rough cut on the board

Explore the infinite canvas

Mosael Infinite Canvas with images, video, audio, and notes connected across an open surface

07 / 3D previs

Block out the shot in 3D before you generate it

Stage a scene, light it and keyframe cameras and objects; watch the shot and the camera path side by side. Send the frame, a first and last frame or a camera preview video to the image or video model you choose, and exchange scenes with Blender in both directions.

Shot preview alongside the camera path
One keyframe timeline for cameras and objects
Frames and camera moves as generation references
Exchange scenes with Blender in both directions

Explore 3D scenes and animation

Mosael 3D scene with gallery models, a global camera path, shot preview and keyframe animation

Templates

What you can make today

Each template is a complete workflow: import it, point it at your own models and footage, and run. Templates carry no accounts, keys or someone else's media.

Browse all workflow templates

Models

Bring the models you already pay for

Connect providers with an API key or a subscription sign-in. You pay the provider directly; Mosael doesn't resell credits and records the usage each run reports. Which models you can use depends on your account and region.

Video: Seedance (Volcengine Ark)
Kling
Veo
Wan
Hailuo (MiniMax)
Sora (via Evolink)
OmniHuman (Jimeng)
HeyGen
Hedra
Image: GPT Image
Seedream
Qwen Image
Gemini (via Evolink)
Your own ComfyUI workflows
Music & sound: Suno (via Evolink)
Lyria
Kling audio
Volcengine AI music
Alibaba Fun-Music
Voice & transcription: F5-TTS (local cloning)
Fish Speech (local cloning)
CosyVoice
Free Edge voices
OpenAI speech
Volcengine speech
FunASR / WhisperX (local transcription)
Chat & agent: DeepSeek
Kimi
Qwen (Alibaba Bailian)
MiniMax
OpenAI
OpenRouter
Ollama and OpenAI-compatible endpoints
Claude Pro/Max and ChatGPT Plus/Pro sign-in

Read the model connection guide

Local-first

Your projects and media stay on your computer.

Projects, media and rendering stay local by default. Cloud models only receive the inputs you send them, through your own account; disconnect the network and you can keep organizing, editing and exporting.

01
Local by defaultProjects and media do not need an upload.
02
Your own accountsProviders bill you directly for model use.
03
No trackingFeatures do not depend on behavioral data.

More capabilities

Finish the work around the video.

From research and scripts to publishing, the easy-to-forget parts of making a video live in the same app.

AI agent

The agent reads your project and uses tools across media, boards, editing and workflows; actions that need approval show a confirmation card first.

Meet the agent

Media library & documents

Imports, screen recordings, URL downloads and AI generations land in one library; PDFs, Word files, slides and spreadsheets are parsed into text on your machine.

Explore the media library

AI notes

Keep scripts, transcripts and research as notes; select a passage to have AI polish, rewrite or translate it, with every change shown for approval.

Explore notes

Multi-platform publishing

Prepare titles and copy for Douyin, TikTok, Bilibili, Xiaohongshu and YouTube, manage several accounts with browser profiles, and review everything before submitting.

Read the publishing guide

Plugins

Blender, Manim, Remotion, TikHub, Baidu Netdisk, MinerU and more; connect local scripts or MCP servers as new tools for the agent and your workflows.

Browse plugins

Chrome video companion

Read transcripts in the browser's side panel, jump to a word, translate, and import the video or the current frame into your library.

Explore the companion

FAQ

Before you download

What is Mosael?

Mosael is a desktop AI video studio for macOS and Windows. Generate images, video and music with your own model accounts or your ComfyUI, edit on a multi-track timeline with subtitles and voice-over, make talking-photo videos and translated dubs, automate repeated steps with workflow templates, and publish to video platforms.

Is Mosael free?

Downloading Mosael and using it for personal, non-commercial work is free. AI models are billed by the providers you connect, at their own prices; Mosael doesn't resell credits. Commercial use requires written permission from the author — the LICENSE file in the repository is authoritative. For a commercial license, email 1142704468@qq.com or message the author on WeChat.

How to get a commercial license →

Is Mosael open source?

The source code is public on GitHub, where you can read it, learn from it and build it locally, but it is released under a proprietary license, not an open-source one: commercial use and redistribution of the source or installers are not allowed without permission. For a commercial license, email 1142704468@qq.com or message the author on WeChat.

How to get a commercial license →

Which AI models does it support?

Video: Seedance, Kling, Veo, Wan, Hailuo (MiniMax) and Sora via Evolink. Images: GPT Image, Seedream, Qwen Image and Gemini. Music: Suno, Lyria and more. Voice: local cloning with F5-TTS or Fish Speech, CosyVoice, Edge, OpenAI and Volcengine. Talking photos: OmniHuman, Kling, HeyGen, Hedra and Wan. You can also connect your own ComfyUI and any OpenAI-compatible endpoint.

Is it a ComfyUI client? How is it different from a cloud ComfyUI service?

Mosael connects to a ComfyUI server on your computer or LAN: saved workflows become models you can pick and tools the agent can run, with a model library that shows previews and a workflow library for importing and completing workflows. It uses your own ComfyUI and GPU, not rented cloud machines, and ComfyUI's own interface keeps working alongside it.

Do I need a GPU?

Not for Mosael itself. Cloud models run on the provider's servers, and ComfyUI runs on whichever machine you connect — this computer or another one on your network. Local transcription, voice cloning and voice separation run on your computer after their models are downloaded; keep 10 GB or more of disk space free for them.

Does Mosael upload my videos?

Projects and media are stored on your computer by default. Mosael goes online when you call a cloud model, download from a URL, use an online tool or publish; a model call sends that run's inputs to the provider you chose (some providers only accept public links, so files go to object storage you configure first). With a team server, shared data lives on that server.

Can it replace CapCut?

It covers multi-track editing, transcript editing, subtitles, voice-over, color and export, keeps projects on your computer and uses your own AI accounts. If you rely on CapCut's templates, stickers and music library, CapCut is still the better fit; Mosael is stronger at AI generation, voice work and automated pipelines.

Which systems does it run on?

macOS on Apple silicon and Windows 10/11 (x64). There is no Linux installer at the moment. Installers are on GitHub Releases, with a Baidu Netdisk mirror for users in mainland China.

Built by a creator

Made for people who care about the work.

Mosael is a source-available project independently built by KindaHuaX. Share how you create, what gets in the way, and what should come next.

KindaHuaX

Follow @KindaHuaX on X

Mosael WeChat group
Scan to join the product and creator community

Author on WeChat
Product feedback, licensing, and partnerships

Ready to make your next video?

Download Mosael, connect one model provider or your ComfyUI, pick a template and run it on your own footage.