Phosphene：Mac 本地 AI 视频创作

Free, open-source app for Apple Silicon Macs: local AI video with sound, songs and images. Includes LTX-2.5 and Hailuo H3.

Phosphene AI video with sound, songs and images, made locally on your Mac. “The Commuter”: a 75-second One Shot rendered overnight on one Mac with Hailuo H3, from a single designed still. More clips. Website · Docs · Benchmarks · Changelog · @PhospheneAI Current release: v4.19.13. See what's new in the changelog. Phosphene is a free, open-source app for Apple Silicon Macs, built on Apple's MLX by Mr Bizarro. Renders run offline, and your prompts, pictures and videos stay on your Mac. Optional extras download when a feature first needs them. What it does Video with LTX-2.5: turn text or a picture into video with sound from the same prompt. Set first and last keyframes, Extend a clip, Retake one moment, or lip-sync a face to your voice or song. Upscale & Face Fix and Motion Control are included. Adapts from 16 GB of memory. Scenes with Hailuo H3: make scenes with dialogue you can hear. Use keyframes, extend a shot, lip-sync, or add sound to a silent clip. About 75 GB, one click from the Pinokio sidebar. Needs 36 GB+ memory. Its licence excludes the EU, UK, South Korea and US. Music Studio: YuE2 makes songs with vocals from lyrics and a style. Make another take or work on the score. About 11 GB, installed inside the panel. Needs 24 GB+ memory. Images: Qwen-Image-Edit and FLUX.2 via mflux make images and edit them in words, like “make the jacket red”. Use up to three reference pictures, then send a still straight to video. Storyboard, One Shot and the Editor: build a film from one prompt or shot by shot. Cut it on the Editor timeline with music and room tone. Train a character: train a face and voice LoRA on your Mac from your own photos and recordings. Scripts and agents: queue renders through the local HTTP API. There's also an image layout API for agents. Prompt help: Google's Gemma runs locally for Enhance, training captions and the storyboard planner. Enhance sharpens a prompt. Hardware Apple Silicon only, M1 or newer. Intel Macs and Windows are not supported. Memory What to expect 16–24 GB LTX-2.5 video from text or a picture at smaller sizes, and images. Songs from 24 GB. 32–36 GB Everything above. Hailuo H3 from 36 GB. 48–64 GB LTX keyframes, Extend, trained characters with Q8 weights, and the High tier. 96 GB+ Most modes at full size (available from 80 GB). Size limits come off at 120 GB. Phosphene is built on an M4 Max with 64 GB. See the benchmarks. Disk: a fresh install downloads 27.5 GB before Start, then about 10.7 GB in the background after the first start. Hailuo H3 (~75 GB) and the music engine (~11 GB) are optional. Install Today, install through Pinokio: Install Pinokio. Open Discover → Download from URL and paste https://github.com/mrbizarro/phosphene. Press Install. Press Start, then open the panel. To update, press Update in Pinokio, then restart Phosphene. See the full install guide for details, or the manual install to do it by hand. A free native Mac app is coming soon. It can reuse models your Pinokio install already downloaded. Linux on Apple Silicon is maintained by @joshuaswarren on a best-effort basis. It is not installed by Pinokio. Where we're going A free native Mac app, so you won't need Pinokio. Pro, for people who use Phosphene every day or earn money with it. More models. Bringing new open models to the Mac first is the point. New ports reach everyone free; Pro may get them up to 24 hours early. Everything available today stays free. Pro We love open source. Phosphene will always be open source and well maintained. If you can afford it, Pro keeps the lights on: faster renders, priority support and more. Pro is coming soon. Nothing is on sale yet. Join the waitlist. Privacy No prompt, picture, video or filename is uploaded; the panel listens on 127.0.0.1 only, with no login, so don't expose it to a network. Phosphene checks GitHub for updates and contacts Hugging Face or CivitAI only when you download models or LoRAs. Anonymous usage counts (version, hardware class such as “M4 Max / 64 GB”, and render stats) are on by default; one click in Settings turns them off, and every field is documented. Contributing Bug reports and PRs are welcome. A report with your chip, macOS version, Phosphene version and the error line is worth as much as a patch. Contributors are credited in the release notes. See CONTRIBUTING.md. Credits Lightricks for LTX-Video 2.x, and @dgrauet for ltx-2-mlx, the MLX port Phosphene builds on. MiniMax for Hailuo H3, and Apple ML research for MLX. LightX2V for lightx2v v1.3, the adapter for Best on High; TaoLive (TaoMate-H3) and @kijai for the 3-step Fast adapter. @filipstrand for mflux; Multimodal Art Projection for YuE2 and @vanch007 for its MLX port. mlx-community for Gemma on MLX, and @cocktailpeanut for Pinokio. Thanks to @blackest, @tanis2000, @joshuaswarren, @sahilkashyap64, @Morac2, @ybekocak, @PhantombrainM, @Viktorminator, @Townesvz, @Sebartus, @mrapolo2020, @Techy2023, and everyone who filed a bug. License Phosphene's own code is MIT. The AI models are not part of this repo and each has its own licence. Phosphene shows a model's licence before download or first run when the licence requires it. Ideogram 4, YuE2's weights, FLUX.1 Krea [dev] and Bria FIBO are non-commercial. YuE2 lets individual creators sell songs they make. Hailuo H3's licence excludes the EU, UK, South Korea and US. Check a model's licence before using its output commercially. Model licence texts.