Higgsfield CLI

Generate images, videos, 3D assets, audio, and finished-video analysis from the terminal using 40+ Higgsfield AI models — Nano Banana Pro, Nano Banana 2 Lite, Gemini Omni Flash, FLUX.2, Soul V2, Veo 3.1, Kling v3.0, Seedance 2.5, Marketing Studio, Virality Predictor, and more. Train face-faithful Soul characters and produce branded marketing assets without leaving your shell.

Contents

• Install
• Quickstart
• Examples
• Models
• Workflows
• Video Explainer
• Games
• Websites
• Commands
• Flags
• Updating
• Uninstall
• Troubleshooting
• Support
• License

Install

macOS / Linux — curl

curl -fsSL https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh | sh

macOS / Linux — Homebrew

brew install higgsfield-ai/tap/higgsfield

Cross-platform (incl. Windows) — npm

npm install -g @higgsfield/cli

Manual

Download an archive matching your OS and architecture from Releases, extract, and place the binary in your $PATH.

Quickstart

Authenticate:

higgsfield auth login

Generate an image and wait for the result URL:

higgsfield generate create nano_banana_2 --prompt "a quiet beach at sunrise" --wait

Examples

Nano Banana Pro

higgsfield generate create nano_banana_2 \
  --prompt "modern architecture, glass facade, golden hour light" \
  --aspect_ratio 16:9 \
  --resolution 2k \
  --wait

GPT Image 2.5

Recommended default for image generation, design, and on-image text.

higgsfield generate create gpt_image_2_5 \
  --prompt "clean infographic showing global energy mix, flat icons, muted palette" \
  --aspect_ratio 3:4 \
  --quality high --resolution 2k \
  --wait

Kling v3.0

higgsfield generate create kling3_0 \
  --prompt "slow camera push through a forest clearing at dawn" \
  --start-image ./first.png \
  --duration 5 --mode pro --sound off \
  --wait

Seedance 2.5

SOTA default for general video generation.

higgsfield generate create seedance_2_5 \
  --prompt "drone shot over a mountain valley at sunrise" \
  --aspect_ratio 16:9 --duration 5 \
  --resolution 1080p --mode t2v --bitrate_mode high \
  --wait

Virality Predictor

brain_activity is the technical job set type for Virality Predictor. It analyzes a finished video for hook strength, attention, retention, and viral potential, then prints scores plus an Open report link.

higgsfield generate create brain_activity --video ./ad.mp4 --wait
higgsfield generate get <job_id>
higgsfield generate wait <job_id>

Draw To Video

Edit a video from a source clip plus an edited sketch frame:

higgsfield generate workflow draw_to_video \
  --video ./source.mp4 \
  --sketch ./frame.png \
  --timestamp 3.2 \
  --prompt "make the jacket red" \
  --wait

Reframe

Reframe a source video for a new aspect ratio:

higgsfield generate workflow reframe \
  --video ./source.mp4 \
  --aspect-ratio 9:16 \
  --resolution 720p \
  --wait

Voice Change

Replace the voice on a source video with a chosen voice:

higgsfield generate workflow voice-change \
  --video ./source.mp4 \
  --voice_type preset \
  --voice_id <voice_id> \
  --wait

Dubbing

Dub a source video into another language (--target_language is an ISO-639-3 code, e.g. eng, spa, fra, deu, jpn; run higgsfield workflow get dubbing for the full list):

higgsfield generate workflow dubbing \
  --video ./source.mp4 \
  --target_language spa \
  --wait

Voices

List available voices (presets + your custom voices) to get a voice_id for text2speech_v2 and voice-change. Use a voice's id as --voice_id and its type (preset/element) as --voice_type:

higgsfield voices list
higgsfield voices get <voice_id> --json

Video Explainer

The explainer skill builds matched 10-second blocks: resolve a style, generate all narration with Seed Audio first, generate the corresponding Gemini Omni clips second, then assemble the ordered pairs with explainer_video:

higgsfield preset list video-explainer --json
higgsfield preset resolve video-explainer <preset_id> --json
higgsfield voices list --json

higgsfield generate create seed_audio \
  --prompt "<Block 1 narration>" \
  --voice_type preset --voice_id <voice_id> --wait --json

higgsfield generate create gemini_omni \
  --prompt "<Block 1 visual prompt>" \
  --image <resolved_style_media_id> \
  --duration 10 --resolution 720p --aspect_ratio 16:9 --wait --json

higgsfield generate create explainer_video \
  --items @blocks.json --width 1280 --height 720 --wait --json

Repeat the audio/video calls once per block. blocks.json maps every clip job to its matching audio job in playback order. See MODELS.md for the assembler schema and optional subtitle fonts.

Games

Deploy a browser-game ZIP whose root contains index.html and either logic.js or server.js:

higgsfield game deploy ./game.zip \
  --title "Space Runner" \
  --description "Fast arcade survival game" \
  --thumbnail https://cdn.example/cover.png \
  --favicon https://cdn.example/icon.png \
  --json

Update the same game with --game-id <game_id>. Marketplace publication is a separate action:

higgsfield game publish <game_id> --name "Space Runner" --json

Browse the rigged 3D animation catalog before choosing an animation_action_id:

higgsfield preset list animation-action --query walk
higgsfield preset list animation-action --group Fighting --category Punching --json

Soul ID

Train a Soul ID once:

higgsfield soul-id create --name me --soul-2 \
  --image ./me1.jpg --image ./me2.jpg --image ./me3.jpg
higgsfield soul-id wait <soul_id>

Reuse it in any compatible image model:

higgsfield generate create text2image_soul_v2 \
  --prompt "professional portrait, neutral background, soft daylight" \
  --soul-id <soul_id> \
  --wait

Models

40+ image, video, 3D, and audio models. Per-model parameters, defaults, and enums: MODELS.md. Live catalog: higgsfield model list.

Image (23)

Video (22)

3D (5)

Audio (5)

text2speech_v2 turns text into speech with a chosen voice. Pick the engine with --variant (elevenlabs, minimax, seed_speech, vibe_voice, cozy_voice) and the voice with --voice_type (preset or element) + --voice_id. Discover voices with higgsfield voices list.

higgsfield generate create text2speech_v2 \
  --prompt "Hello from Higgsfield" \
  --variant elevenlabs \
  --voice_type preset \
  --voice_id <voice_id> \
  --wait

Workflows

Workflows are higher-level generation flows with their own parameter schemas. Use workflow list to discover available workflows and workflow get to inspect the parameters before creating a job.

higgsfield workflow list
higgsfield workflow get draw_to_video
higgsfield workflow get reframe --json
higgsfield workflow get voice-change
higgsfield workflow get dubbing

Create workflow jobs through generate workflow:

higgsfield generate workflow draw_to_video \
  --video ./source.mp4 \
  --sketch ./frame.png \
  --timestamp 3.2 \
  --prompt "make the jacket red" \
  --wait

higgsfield generate workflow reframe \
  --video ./source.mp4 \
  --aspect-ratio 9:16 \
  --resolution 720p \
  --wait

higgsfield generate workflow voice-change \
  --video ./source.mp4 \
  --voice_type preset \
  --voice_id <voice_id> \
  --wait

higgsfield generate workflow dubbing \
  --video ./source.mp4 \
  --target_language spa \
  --wait

Estimate workflow cost through generate cost workflow:

higgsfield generate cost workflow draw_to_video --duration 8.2 --resolution 720p
higgsfield generate cost workflow reframe --duration 7.1 --resolution 1080p

voice-change and dubbing do not support cost estimation.

Fetch or wait for workflow jobs with the same job commands used by model generations:

higgsfield generate get <job_id>
higgsfield generate wait <job_id>

Websites

Build and deploy full-stack websites from the terminal. Each site is a React 19 + TanStack Start app, server-rendered as a single Cloudflare Worker, with D1, R2, KV, Durable Objects, and Containers available. higgsfield website create provisions the site and a git repo; you clone it, edit the code under app/, push, and deploy to its live URL. The build runs on the Higgsfield platform from the pushed branch.

create requires --type — what kind of product you're building:

• website — a standalone site with no Higgsfield integration (no "Sign in with Higgsfield", no requests to Higgsfield). Landing pages, portfolios, general tools.
• app — a product tightly integrated with Higgsfield: its users sign in with Higgsfield and generate images/videos through the Higgsfield SDK.

create also requires --category — the content category the site is filed under on the marketplace. It's a slug from a curated taxonomy (e.g. cinematic, ads-marketing, ugc-social, other); run higgsfield website categories to see the full list, then pass the closest one (use other when nothing fits). The taxonomy can grow over time, so the server validates the slug and rejects an unknown one.

--type app also requires --template. The flag has exactly four choices:

Pick the closest of app-detail, preset, and studio. Agents must never choose custom by default.

For every non-custom app template, the starter repo ships real code at app/src/layouts/<template>.tsx, already wired as the home page. Adapt that layout in place and thread real data through it; do not rebuild it, replace it, or swap it for another layout. The shipped UI includes demo placeholders, so it still needs the product's real business logic. After cloning, read both app/src/layouts/AGENTS.md and app/src/components/AGENTS.md before editing.

Template validation happens locally: a missing or invalid app template fails before authentication or any backend call. --type website does not require a template. If --template is supplied for a standalone website, the CLI ignores it and omits it from both the backend request and the create result.

Pass --subdomain to choose the site's subdomain — it becomes the slug, so the live URL is <subdomain>.<host>. Always set one derived from the site's name (lowercase, DNS-safe: letters, digits, single hyphens); omit it only if you want a random subdomain. Reserved labels (e.g. api, www) and already-taken subdomains are rejected — pick another.

Create a standalone website or an app using the closest starter template:

higgsfield website create --type website --category other

higgsfield website create \
  --type app \
  --category product-ecommerce \
  --template app-detail

higgsfield website create \
  --type app \
  --category ads-marketing \
  --template preset \
  --subdomain my-app

higgsfield website create \
  --type app \
  --category cinematic \
  --template studio

App create results preserve the selected template. Human-readable table output includes Category and Template columns (alongside Website ID, Type, Slug, Name, Preview URL, and Production URL), and non-custom apps also print the shipped layout path. JSON output includes "template" for apps:

{
  "website_id": "<website_id>",
  "type": "app",
  "template": "studio"
}

Standalone website results omit template, including when an ignored --template was supplied; table output leaves the Template cell empty, while JSON has no template field.

Continue with the returned website_id:

# 1. Create the site + its git repo using one of the commands above.
#    Prefer a DNS-safe --subdomain derived from the site's name.

# 2. Get the clone URL, branch, and a scoped git token
higgsfield website repo-access <website_id>

# 3. Clone with the token, edit under app/, commit, and push
git -c http.extraHeader="Authorization: token <token>" clone <repo_url> <slug>
cd <slug>
# The scaffolded clone has no git identity — set one or the first commit fails:
git config user.email "agent@higgsfield.ai" && git config user.name "Higgsfield Agent"
# ...edit files under app/ (bun-only repo: bun install / bun add / bunx /
#    bun run typecheck|build — never npm/npx/yarn; app/src/routeTree.gen.ts
#    is generated, never hand-edit it) ...
git add -A && git commit -m "initial build"
git -c http.extraHeader="Authorization: token <token>" push origin <branch>

# 4. Deploy — ships the live site (run it again after every change)
higgsfield website deploy <website_id>

# Publish — list the site on the Higgsfield community feed ("show in feed")
# where others can discover and remix it. Publish does NOT deploy: it lists
# whatever `deploy` last shipped, so deploy first (and again after any change).
# Prints the community-feed listing URL.
higgsfield website publish <website_id>

# Enter the $100k Higgsfield app contest (type: app). The entry PUBLISHES the
# app for you — no separate `publish` needed; it just needs a live deploy and
# filled page metadata. Pass one or more public social links (Instagram /
# TikTok / YouTube / X) promoting it. Re-running overwrites the links.
higgsfield website contest <website_id> --url https://x.com/<user>/status/...

# Check deploy status and the live URL any time
higgsfield website status <website_id>

Rename a site's subdomain (the slug in its public URL). The site is re-deployed under the new subdomain and the old subdomain stops working — share the new URL afterwards. Storage (database, files, config) and the git repo are kept. Blocking like deploy (a couple of minutes); reserved or taken subdomains are rejected — pick another:

higgsfield website rename <website_id> --subdomain my-new-name

Inspect the site's database (read-only) and manage secrets (staged until the next deploy):

higgsfield website db tables <website_id>
higgsfield website db rows <website_id> --table users --limit 20
higgsfield website db query <website_id> --sql "SELECT count(*) FROM users"

higgsfield website secrets set <website_id> --name STRIPE_SECRET_KEY --value sk_live_...
higgsfield website secrets list <website_id>

List the sites you own:

higgsfield website list

List the content categories a site can be filed under (the --category slugs for create), each with its slug, label, and description:

higgsfield website categories

Add --json to any command for machine-readable output.

Commands

Run higgsfield <command> --help for flags and examples (also higgsfield generate create --help, higgsfield soul-id create --help, etc.).

Flags

Flags work across all commands.

Example pipeline:

higgsfield generate list --json | jq -r '.[] | select(.status=="completed") | .result_url'

Updating

# curl
curl -fsSL https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh | sh

# brew
brew update && brew upgrade higgsfield

# npm
npm install -g @higgsfield/cli@latest

Pin to a specific release:

curl -fsSL https://raw.githubusercontent.com/higgsfield-ai/cli/main/install.sh | sh -s -- --tag v1.1.2
# or
npm install -g @higgsfield/cli@1.1.2

Uninstall

# curl install (default prefix /usr/local)
sudo rm /usr/local/bin/higgsfield

# brew
brew uninstall higgsfield

# npm
npm uninstall -g @higgsfield/cli

Troubleshooting

Session expired / Not authenticated — tokens are short-lived. Re-run higgsfield auth login.

Unknown model "<name>" — run higgsfield model list for the current catalog.

Support

Bugs and feature requests: github.com/higgsfield-ai/cli/issues. Please include higgsfield version output and the exact command that failed.

License

MIT