egaki

Egaki is a command-line interface (CLI) and MDX-based framework designed for AI-powered media generation, enabling users to create images, videos, and speech from code. The tool addresses the need for programmable media creation by allowing direct generation and editing of visual and audio content through simple commands or scriptable MDX files. It is suitable for developers, creators, and anyone seeking to automate or customize media production workflows using code.

The platform supports a range of media generation features. For images, Egaki provides capabilities such as text-to-image synthesis, editing existing images, inpainting with masks, adjusting aspect ratios, generating multiple images at once, and deterministic output. Image generation can be piped to stdout or output as JSON for automation purposes. For video, Egaki enables text-to-video and image-to-video workflows, video editing, video extension, and reference-to-video (R2V) generation, as well as exporting multiple videos. Videos can be scripted using MDX, where each heading defines a timed scene section, and animation primitives like Opacity, TranslateX, and Scale allow for motion effects. Export to MP4 is handled via in-browser WebCodecs rendering, and the framework supports visual style interpolation, cross-scene transitions, and built-in components for effects such as blur reveals and staggered fades.

Egaki also offers speech and audio features, including text-to-speech synthesis, voice cloning, audio stem separation using demucs, and transcription with word timestamps. The tool supports reading input from stdin, controlling speech speed, and generating captions for video. Multiple providers are supported for both image and speech generation, with authentication handled via provider keys, ChatGPT OAuth, or an Egaki subscription. The platform is extensible, with a catalog of supported models and detailed CLI documentation available.

Delivery is primarily through the CLI and MDX scripting, with project setup involving configuration steps such as Vite and TypeScript integration. Pricing is model-dependent, and users can bring their own provider keys or use an Egaki subscription for authentication. The tool is positioned as an AI media generation framework for code-driven content creation.