README

OpenHiggsfield AI — Open-Source Alternative to Higgsfield AI

 The free, open-source alternative to Higgsfield AI. Generate images and
 videos with 38 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?

- Free & open-source — no studio subscription, no vendor lock-in
- Self-hosted — clone it, run it, change it
- Your key — generate with your own platform key
- 38 models — 8 image, 30 video, one catalog, one composer

---

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

---

 Features

 Generate

- One composer for Image and Video. A single prompt bar drives both; the
  model you pick decides image or video. ⌘/Ctrl + Enter submits.
- 38 models in the catalog — 8 image, 30 video: Soul 2, Soul Cinema, Seedance
  2.5 (Edit / Extend), Seedance 2.0 (Fast / Mini), Kling 3 (Turbo / Std / Pro / 4K / Motion), Wan, Flux,
  Ideogram, Recraft, LTX, MiniMax, PixVerse, Grok, Qwen and more. Searchable
  picker.
- Per-model settings. Aspect ratio, resolution, duration, output format,
  audio, batch size, prompt enhancement — each model declares its own allow-list
  and the studio renders exactly that. No parallel hardcoded list.
- Media inputs by role. Start frame, end frame, references, video and audio,
  each with the per-role cap the model declares. Files upload to Vercel Blob and
  become public URLs the generate request can carry.
- Asset picker. Attach from your uploads library or from any finished run in
  history — two tabs over one library, filtered to the role's kind.
- Batch. Up to 4 results per press. Models with a native count setting use it;
  the rest are submitted once per result, each clearing its own tile.
- Live run lifecycle. Skeletons open in the grid on submit, the request is
  polled every 4s until a terminal status (10-minute deadline), and each finished
  result blooms into place on its own clock.

 Gallery

- Four scopes — Image, Video, Assets (every finished run) and Favorites —
  as an arrow-key-navigable tab rail.
- Masonry grid of real runs at their true aspect ratio, newest first, with a
  gradient placeholder while media loads.
- Per-tile actions: reuse, favorite, delete, select.
- Reuse restores model, settings and prompt, so the same run can be
  re-rendered, not just re-typed.
- Viewer. Full-size media with prompt (copy in one click), model, resolved
  settings, timestamp, download, favorite and Recreate.
- Selection mode. Click a tile's checkbox to enter; shift-click extends a
  range. Bulk download (sequential, with progress and a report of any files the
  CDN refused), bulk favorite/unfavorite, bulk delete. Esc exits.
- Undo. Deletion is reversible for 6 seconds via a bar with a draining
  hairline, i

Top issues