wide-trace/open-higgsfield

What it solves

OpenHiggsfield AI is an open-source studio for generating images and videos. It removes the need for closed ecosystems and expensive studio subscriptions by allowing users to bring their own platform keys to access a wide variety of generative AI models.

How it works

The project provides a unified interface (a "composer") that connects to a generation API. Users provide their own API keys to trigger generations. The system uses a catalog-driven architecture where each of the 40 supported models (including Flux, Kling, and Veo) declares its own specific settings, aspect ratios, and media input roles, which the UI then renders dynamically.

Who it’s for

Content creators and AI artists who want a single, self-hosted interface to iterate on prompts and settings across multiple image and video generation models without being locked into a specific vendor's platform.

Highlights

• Unified Prompt Bar: A single composer for both image and video generation.
• Extensive Model Catalog: Support for 40 models, including 12 image and 28 video models.
• Dynamic Settings: Per-model settings (resolution, duration, etc.) are rendered based on the model's own declarations.
• Asset Management: Integrated asset picker and gallery with masonry grid, bulk actions, and reuse capabilities to restore previous prompts and settings.
• Self-Hosted: Built with Next.js 16 and React 19, allowing users to clone and run the studio locally.