Which agents and editors are supported?

First-class support for Claude Code, Codex, OpenCode, OpenClaw and Hermes. The CLI runs in any shell, so Cursor, Windsurf, Gemini CLI, GitHub Copilot and any command-capable agent work too.

How much does it cost?

Free to use. All generation consumes the credits in your PixVerse account and reuses your membership benefits — pricing, models and every capability are aligned with the website.

Can I use it in CI or headless environments?

Yes. Device-flow auth works over SSH, and every command supports JSON output, deterministic exit codes and non-blocking task polling.

Give your agent the full power to create viral content.

Q: What is PixVerse CLI & Skills?

The PixVerse CLI is a UI-free version of the PixVerse platform — everything you can do on app.pixverse.ai works from the command line with the same models and quality. PixVerse Skills teaches coding agents like Claude Code, Codex, OpenCode, OpenClaw and Hermes to use those capabilities.

PixVerse CLI & Skills turns your own agent into an AI content studio: full video, image, voice and music generation, every top AI model in one place, and workflows built to create at scale.

Install Guide Docs for Agents

~/studio — pixverse · video

$

Works with

Claude CodeCodexOpenCodeOpenClawHermesCursorWindsurfGemini CLIGitHub Copilot

# 01 · Install

Up and running in three steps.

Built for agents. Let your agent handle the tedious work — the creative spark stays yours.

01

Install the CLI

One global npm package for all platforms.

02

Sign in

Sign in with your own PixVerse account — keep every membership benefit and credit, and share all your assets and creation history.

03

Teach your agent

Install the PixVerse Skills pack to teach your agent how to get the most out of everything PixVerse can do.

# 02 · Capabilities

Every top model. One command away.

PixVerse V6, Seedance, Kling, Veo, Sora, GPT Image, Nano Banana, ElevenLabs, MiniMax and more — one auth, one credit balance, one interface your agent already knows.

PixVerseSoraVeoGeminiKlingSeedanceGrokGPT ImageSeedreamQwenMiniMaxElevenLabs

Core generation

create video01

Video

Text-to-video and image-to-video across 19 models — PixVerse V6, Seedance 2.0, Kling, Veo 3.1, Sora 2 and more, up to 15s and 4K.

$pixverse create video --prompt "…" --model v6 --quality 1080p

create image02

Image

GPT Image 2, Nano Banana 2, Seedream and Qwen — up to 4K, every aspect ratio, image-to-image included.

$pixverse create image --prompt "…" --aspect-ratio 16:9

create voice03

Voice

Studio-grade TTS with MiniMax Speech 2.8 and ElevenLabs voices — narration and dialogue in one flag.

$pixverse create voice --text "…" --voice-id luna

create music04

Music

Full tracks from a prompt with MiniMax Music, ElevenLabs Music and Google Lyria — lyrics, auto-lyrics or instrumental.

$pixverse create music --prompt "…" --auto-lyrics

Workflows & scale

create reference05

Consistent characters

Keep the same subject across shots with up to 7 reference images — mix in motion and audio refs on Seedance 2.0.

create transition06

Keyframe transitions

Turn keyframes into one continuous shot — storyboards become film.

create motion-control07

Motion transfer

Drive a character image with any motion reference video.

create extend08

Extend & modify

Continue a video past its last frame, or rewrite it with a new prompt.

create template09

Viral templates

Apply trending PixVerse templates to your media with one command.

--count 4 --off-peak10

Batch & off-peak

Queue up to 4 variants per run, render on off-peak pricing, and poll everything with task status --json.

# 03 · Automation

Built for pipelines, not just prompts.

Batch by default

Render variants from a CSV or a for-loop with --count, --seed and --no-wait. Off-peak pricing keeps large runs affordable.

$ pixverse create video --prompt ./hooks.txt --count 4

CI-friendly

Device-flow auth that works over SSH, JSON output and deterministic exit codes — video belongs in your build artifacts.

$ pixverse task wait t_8f3 --json

Agent-native

Structured output and self-describing skills mean your agent picks up new capabilities without a line of glue code.

$ pixverse auth status --json

# 04 · For agents

Too long to read? Let your agent read and set it up.

Everything here — install steps, commands, models, flags and limits — is one plain-markdown file your agent can read in a single fetch and set up for you.

Open llms.txt

llms.txt

# PixVerse CLI & Skills — agent reference

> Full video, image, voice and music generation
> for terminals and coding agents — every top
> AI model behind one command.

## Install

```sh
npm install -g pixverse
pixverse auth login
```

# 05 · FAQ

Questions, answered.

PixVerse CLI is a UI-free version of the PixVerse platform — everything you can do on app.pixverse.ai works from the command line with the same models and quality. PixVerse Skills is a companion pack that teaches coding agents like Claude Code, Codex, OpenCode, OpenClaw and Hermes to use those capabilities on your behalf.

First-class support for Claude Code, Codex, OpenCode, OpenClaw and Hermes. And because the CLI runs in any shell, any agent that can execute commands works too — Cursor, Windsurf, Gemini CLI, GitHub Copilot and more.

Free to use. All generation consumes the credits in your PixVerse account and reuses your membership benefits — same pricing, same models, every capability aligned with the website.

No — pixverse auth login uses an OAuth device flow you can confirm from any browser, even over SSH. No manual key handling required.

Yes. Every command supports --json output and deterministic exit codes, --no-wait submits without blocking, and task status/wait handles polling — videos and images render straight into GitHub Actions or any scheduler.

Every render lands in your PixVerse asset library, so nothing is lost when a job outlives your shell session. Download with pixverse asset download, and organize with saved folders.