Create images and videos with over 600 AI models without subscription fees or content filters
Enter text in a browser or desktop app, and models like Flux, Sora, and Kling generate images and videos. You can run generation tasks on your own computer without paying for paid platform subscriptions or dealing with content filters. This tool combines Muapi.ai's cloud API with a local inference engine to handle various generation tasks in a single interface.
A generation environment spanning cloud and local
The tool calls over 600 cloud models via Muapi.ai and supports local model inference in the desktop app. Cloud models include the latest options such as Flux, Midjourney, Sora, and Veo, allowing generation without content filters or prompt refusals. The desktop app provides two local engines, sd.cpp and Wan2GP, enabling image and video generation without an internet connection. On Apple Silicon Macs, it supports Metal GPU acceleration for significantly faster processing speeds than the CPU.
Various studios and workflows
Beyond simple text-to-image conversion, it offers over 14 specialized studios for video, audio, lip sync, cinema, and marketing. The image studio automatically switches to an image-to-image transformation model when a reference image is uploaded and supports multi-image input with up to 14 reference images simultaneously. In the workflow studio, you can visually build and execute multi-stage pipelines connecting image, video, and audio models using a node-based editor. You can also use community templates or plan and execute generation tasks interactively through the agent studio.
Local inference engine setup and hardware requirements
Local inference is available only in the desktop app, where you choose between two engines based on your situation. The sd.cpp engine runs on Mac, Windows, and Linux, supporting image models like SD 1.5, SDXL, and Z-Image. Using the Z-Image model is recommended with 16GB or more of RAM; on Macs with 8GB of RAM, the system may freeze, so using the SD 1.5 model is safer. The Wan2GP engine handles video models (Wan 2.1/2.2, Hunyuan, LTX-2) and large image models (Flux, Qwen-Image) that require an NVIDIA or AMD GPU. This engine can be installed on the same computer or connected to another GPU machine via a remote MCP server.
Installation methods and platform-specific notes
The desktop app provides installers for macOS, Windows, and Linux, requiring no Node.js or terminal setup. On macOS, the app may be blocked by Gatekeeper due to the lack of Apple certification, so you must run the xattr -cr command in the terminal or select 'Open Anyway' in System Settings. On Windows, a SmartScreen warning may appear; click 'More info' and then 'Run anyway'. On Ubuntu 24.04 or later, the app may fail to start due to AppArmor security policies, so you should install the .deb package or disable user namespace sandboxes using the sysctl command.
Things to check before use
The tool is distributed under the MIT license, and data remains on your computer when self-hosted, ensuring privacy. However, when using cloud models, processing goes through Muapi.ai's API, so you must verify API key management and cost structures. If you want local inference, you must check your hardware specifications, especially GPU and RAM capacity, as certain models require downloading large weight files. Additionally, the tool does not provide guidelines on legal liability or ethical use of generated content, so you must judge the purpose of use and copyright issues yourself.
