open-higgsfield README

The project's README from github.com/wide-trace/open-higgsfield. As one markdown file: README.md.

OpenHiggsfield AI — Open-Source Alternative to Higgsfield AI

The free, open-source alternative to Higgsfield AI. Generate images and videos with 32 models from one prompt bar — no closed ecosystem, no studio subscription.

🌐 Try it Online — No Install Required

Hosted version: openhiggsfield.ai

Image and Video in one studio, in the browser — no Node.js, no setup. Add your platform key (id:secret) to start generating. The studio itself is free.

Why OpenHiggsfield AI instead of Higgsfield AI?

Next.js 16 App Router on Vercel · React 19 · plain CSS · Zustand · pnpm

Features

Generate

State and errors

Architecture

Each generate is one object: { model, prompt, media, settings }.

Getting started

pnpm install
pnpm dev            # http://localhost:3000

Open the studio, press Add key, and paste your platform key as id:secret.

Environment

HF_API_BASE_URL=                      # generation API origin, server only
OPEN_HIGGSFIELD_READ_WRITE_TOKEN=     # Vercel Blob read-write token

Commands

Command What it does
pnpm dev Dev server on port 3000
pnpm build Production build
pnpm start Serve the production build
pnpm brand Rebuild the icons and OG card in public/

Layout

src/
  app/          /  is the full-viewport studio and the only page
                /api/blob issues upload tokens
                base.css owns the document canvas
  generation/   generate requests, server actions, API mapping, catalog, stores
  openhiggsfield/
                the studio surface: composer, gallery, viewer, model picker,
                settings, asset picker, selection bar — and openhiggsfield.css

Design principles

Dark studio ground, a single lime accent #d1fe17, Inter throughout. The chrome stays neutral so the generated work is the only color on the surface.

  1. The tool disappears into the task — expression never obscures state or affordance.
  2. Accent is state, not decoration — selection, primary action, liveness only.
  3. Data is data — settings, counts and durations read in tabular numerals. One typeface throughout; no monospace anywhere.
  4. Motion conveys state — the generation lifecycle, the arrival of a run. Nothing loops decoratively.
  5. Every control ships all its states — hover, focus, active, disabled, loading, error, empty.
  6. The catalog is the source of truth — the studio renders what the model declares, never a parallel hardcoded list.

Built for people who work in long sessions, iterating on prompts, inputs and settings.

open-higgsfield (wide-trace/open-higgsfield)

Searching Database…

4.4K ·GitHub
Open-higgsfield
What is this?OpenHiggsfield AI is a free, open-source, self-hosted studio for generating images and videos through one prompt bar, using 32 different AI models (8 image, 24 video) that each get their own settings panel and land in one shared gallery.
Is this a framework I build on, or a ready-to-use app?It is a ready-to-use web application (a Next.js studio), not a library or framework — you clone it, self-host it, and use it directly as an image/video generation tool.
How do I get it running?Use the standard Next.js scripts: `pnpm dev` runs a dev server on port 3000, `pnpm build`/`pnpm start` build and serve production, or skip install entirely and use the hosted version at openhiggsfield.ai.
Do I need my own API key to generate anything?Yes — you must supply your own generation platform key in the format id:secret; the app opens a modal demanding it and generation won't start without it.
Any catches with history and saved files?Generation history lives only in the browser's IndexedDB and is capped at 60 records (favorites are exempt); since media URLs point to the generation platform's own CDN, old entries can go dead once that CDN expires them.
Does my history or key get sent to your servers?History stays client-side in your browser's IndexedDB; the platform key you enter is stored only in an httpOnly cookie set by a server action, and the browser never talks to the generation API directly.
How many models does it support and what are they?32 models total — 8 for images and 24 for video — including Soul 2, Soul Cinema, Kling 3 variants, Wan, Flux, Ideogram, Recraft, LTX, MiniMax, PixVerse, Grok and Qwen, all searchable in one catalog.
Can I redo a previous generation with the same settings?Yes, the gallery's 'Reuse' action restores the exact model, settings and prompt of any past run so you can re-render it instead of retyping everything from scratch.
What happens if I delete something by mistake?Deletions are reversible for 6 seconds through an undo bar before they become permanent.
Why build this instead of just paying for Higgsfield AI?To avoid vendor lock-in and subscription fees — OpenHiggsfield AI is free, self-hosted, and uses your own platform key instead of a paid closed-ecosystem studio.
What are similar projects?waoowaoo, unreel, artcraft-services, Open-Generative-AI, moyin-creator, roomGPT, printfilm, html-anything
Which topics is it in?indexeddb, multi-model, ai-art, vercel, ai-tools, content-creation, video-generation, image-generation, nextjs, generative-ai, css, web-app
Creator
wide-trace
Size
0.5 MB
Stars
4.4K
Updated
CommitSep 17
Age
1mo
Commits · 90 Days