@nsketch/cli

Generate images, videos, voice, motion transfer, and more with Nsketch AI from your terminal — or let your AI agent (Claude Code, Codex, Cursor) do it for you.

Install

npm install -g @nsketch/cli

Sign in

nsketch auth login

Opens a browser, takes 5 seconds. Credentials are stored in ~/.config/nsketch/credentials.json.

Usage

# What can I use?
nsketch models list --type image
nsketch account status

# Generate an image and wait for the result URL
nsketch generate image --prompt "a corgi surfing at golden hour" --wait

# Seedream 5 Pro at 2K
nsketch generate image --model img-seedream-v5-pro --resolution 2K --prompt "a cinematic mountain village" --wait

# Seedream 5 Pro edit using the reference image's natural aspect ratio
nsketch generate image --model img-seedream-v5-pro-edit --resolution 2K --aspect-ratio auto_2K --image ./reference.png --prompt "make this cinematic" --wait

# Image-to-video (local files are auto-uploaded)
nsketch generate video --prompt "the corgi rides a wave" --image ./corgi.png --wait

# Flux 3 first/last-frame video (use vid-flux-3-draft for the 720p draft)
nsketch generate video --model vid-flux-3 --start-frame ./first.png --end-frame ./last.png --duration 8 --resolution 1080p --prompt "A smooth cinematic transition" --wait

# MiniMax H3 multimodal references (9 images, 3 videos, 3 audio maximum)
nsketch generate video --model vid-minimax-h3 --reference-image ./character.png --reference-video ./motion.mp4 --reference-audio ./music.wav --duration 8 --resolution 2K --aspect-ratio adaptive --prompt "Use @image1 with the motion in @video1 and sound from @audio1" --wait

# Seedance 2.5 multimodal references (30 images, 10 videos, 10 audio maximum)
nsketch generate video --model vid-seedance-2.5 --reference-image ./character.png --reference-video ./scene.mov --reference-audio ./dialogue.mp3 --duration 12 --resolution 720p --aspect-ratio adaptive --prompt "Place @image1 in @video1 and synchronize @audio1" --wait

# Voice, lipsync, motion transfer, upscaling
nsketch generate voice --script "Hello world"
nsketch generate lipsync --video ./clip.mp4 --audio ./voice.mp3 --audio-duration 5
nsketch generate motion-transfer --image ./character.png --video ./dance.mp4 --wait
nsketch generate upscale --image ./photo.jpg --mode crystal

# Check an async job later
nsketch status <request-id> --type video --model <model-id>

# Upload a file and get a hosted URL
nsketch upload ./photo.jpg

Global flags: --json (machine-readable output on stdout), --wait, --wait-timeout 10m, --wait-interval 3s, --api-url (override the API base URL).

For AI agents

All commands support --json. Generations consume account credits; check nsketch account status first. Async jobs return a request id — prefer --wait so the CLI handles polling.