higgsfield-ai/cli

Higgsfield CLI – AI generation from the terminal

What it is – A cross‑platform command‑line tool that lets you call 40 + Higgsfield AI models (image, video, 3D, audio) directly from your shell. It handles authentication, job submission, cost‑estimation, and result retrieval, and even lets you spin up full‑stack React‑based web apps or publish browser games – all without leaving the terminal.

Installation

• macOS/Linux via a one‑liner curl … | sh
• Homebrew tap (brew install higgsfield-ai/tap/higgsfield)
• Any platform via npm (npm install -g @higgsfield/cli)
• Manual download of pre‑built binaries.

Typical workflow

1. higgsfield auth login – store your API token.
2. Create a job, e.g. higgsfield generate create nano_banana_2 --prompt "a quiet beach at sunrise" --wait.
3. The CLI polls the service, prints a URL (or JSON with --json) when the asset is ready.

Key capabilities

Area	What you can do from the CLI
Images	Generate, up‑scale, out‑paint, background‑remove, style‑transfer with models such as Nano Banana Pro, FLUX.2, GPT‑Image 2.5, Soul V2, etc.
Videos	Text‑to‑video, style‑guided clips, re‑framing, voice‑change, dubbing, virality analysis (Virality Predictor) using models like Seedance 2.5, Kling v3.0, Gemini Omni.
3D assets	Convert multi‑image sets or text prompts into 3D meshes, rigging, or Tripo‑style assets.
Audio	Text‑to‑speech (ElevenLabs, Minimax, etc.), music generation, voice‑swap.
Workflows	Higher‑level pipelines (draw‑to‑video, reframe, voice‑change, dubbing) that bundle multiple model calls.
Websites	Scaffold a React 19 + TanStack Start app, push code via a temporary git token, and deploy as a Cloudflare Worker‑backed site. Choose a template (app‑detail, preset, studio) or a plain website.
Games	Upload a ZIP containing a browser game and publish it to the Higgsfield marketplace.
Soul ID	Train a personal “Soul” character from a few photos and reuse it for face‑faithful image generation.
Cost estimation	higgsfield generate cost … predicts credits required for a given model or workflow.
Account & workspace	View credit balance, transaction history, and switch billing workspaces.

Model catalogue – The higgsfield model list command shows all 40+ models. They are grouped into Image (23), Video (22), 3D (5) and Audio (5). Each model has its own job_set_type (e.g. nano_banana_2, seedance_2_5, text2speech_v2) and a set of parameters documented in MODELS.md.

Example commands

# Simple image
higgsfield generate create gpt_image_2_5 \
  --prompt "clean infographic showing global energy mix" \
  --aspect_ratio 3:4 --quality high --resolution 2k --wait

# Video with voice‑over
higgsfield generate workflow dubbing \
  --video ./ad.mp4 --target_language spa --wait

# Deploy a full‑stack app
higgsfield website create \
  --type app --category ads-marketing --template studio --subdomain my‑ads
# Clone, edit, push, then deploy
higgsfield website repo-access <id>
# …git work…
higgsfield website deploy <id>

Why it matters – By exposing a rich set of generative AI services through a lightweight CLI, Higgsfield CLI makes it easy to script content pipelines, integrate AI assets into CI/CD, or prototype products without writing any HTTP client code. It also bundles end‑to‑end tooling for web‑app scaffolding and marketplace game publishing, turning a single binary into a full creative‑production workstation.

License – Open‑source under the repository’s LICENSE (see LICENSE file).