wonda

AI-powered content generation from your terminal

Images, video, music, audio, editing, and social publishing — all via CLI.

---

You don't need to learn this CLI. Your agent already knows how to use it.

and any agent that can run shell commands.

Install

curl -fsSL https://wonda.sh/install.sh | bash

brew tap degausai/tap && brew install wonda

npm i -g @degausai/wonda

Get started

wonda auth login          # Authenticate (opens browser)
wonda skill install -o .  # Install skill file for your AI assistant

Then ask your agent: "Use wonda to generate a product video of this image."

Made with wonda

Product videos, UGC-style content, ad creatives — generated, edited, and published from the terminal.

Pricing

An account is required. Sign up at wonda.sh.

Generations cost credits. Top up anytime:

wonda topup    # Opens Stripe checkout
wonda balance  # Check remaining credits

Use wonda pricing estimate to check costs before generating.

Commands

Generation

Editing

TikTok/Reels-style video editing operations — designed for short-form social content.

Analysis

Publishing

Reddit

Marketing & Analytics

Media & Workflows

Examples

Generate an image

wonda generate image \
  --model nano-banana-2 \
  --prompt "Product photo of headphones on marble" \
  --wait -o photo.png

Generate a video from a reference image

MEDIA=$(wonda media upload ./product.jpg --quiet)
wonda generate video --model sora2 \
  --prompt "Slow orbit, dramatic lighting" \
  --attach "$MEDIA" --duration 8 --wait -o video.mp4

Add animated captions (TikTok-style)

wonda edit video --operation animatedCaptions --media "$VID_MEDIA" \
  --params '{"fontFamily":"TikTok Sans","position":"bottom-center","highlightColor":"#FFD700"}' \
  --wait -o captioned.mp4

Full pipeline: generate → music → captions → publish

# Generate a product video
VID=$(wonda generate video --model sora2 --prompt "Ocean waves" --wait --quiet)
VID_MEDIA=$(wonda jobs get inference "$VID" --jq '.outputs[0].media.mediaId')

# Add background music
MUSIC=$(wonda generate music --model suno-music --prompt "lo-fi ambient" --wait --quiet)
MUSIC_MEDIA=$(wonda jobs get inference "$MUSIC" --jq '.outputs[0].media.mediaId')
MIXED=$(wonda edit video --operation editAudio --media "$VID_MEDIA" --audio-media "$MUSIC_MEDIA" \
  --params '{"videoVolume":80,"audioVolume":30}' --wait --quiet)
MIXED_MEDIA=$(wonda jobs get editor "$MIXED" --jq '.outputs[0].mediaId')

# Burn in animated captions
FINAL=$(wonda edit video --operation animatedCaptions --media "$MIXED_MEDIA" \
  --params '{"fontFamily":"Montserrat","position":"bottom-center"}' --wait --quiet)
FINAL_MEDIA=$(wonda jobs get editor "$FINAL" --jq '.outputs[0].mediaId')

# Publish
wonda publish tiktok --media "$FINAL_MEDIA" --account tiktok_acct_123 \
  --caption "Summer vibes" --privacy-level PUBLIC_TO_EVERYONE

Publish to Instagram

wonda publish instagram \
  --media med_abc123 \
  --account ig_acct_456 \
  --caption "New drop. Link in bio."

Output formats

All commands output JSON to stdout. Errors go to stderr.

# Default — formatted JSON
wonda generate image --model nano-banana-2 --prompt "A cat"

# Quiet — just the ID, useful for shell variables
JOB=$(wonda generate image --model nano-banana-2 --prompt "A cat" --quiet)

# Field selection
wonda jobs get inference "$JOB" --fields status,outputs

# Built-in jq (no external dependency)
wonda generate image --model nano-banana-2 --prompt "A cat" --wait \
  --jq '.outputs[0].media.url'

When stdout is piped, JSON mode is enabled automatically.

AI agent integration

Just point your agent at wonda — it reads --help, finds the built-in skill file, and figures out model selection, prompt strategies, and content workflows on its own.

wonda skill install              # Sync skill file to ~/.wonda/skill/
wonda skill install --all -o .   # Install main + all content skills locally
wonda skill list                 # Browse available content skills
wonda skill get product-b-roll   # Fetch a specific content guide

The skill file auto-syncs in the background. No configuration needed — your agent discovers it automatically.

Platforms

macOS · Linux · Windows — x64 + ARM64

License

Proprietary — see wonda.sh for terms.