Hanzo Studio · Visual AI engine

Build AI pipelines on a canvas

The visual AI engine for generative media. Wire models, prompts, and tools into a node graph, then generate images, video, audio, and 3D — no code. Prototype on your own machine and scale the same workflow onto cloud GPUs.

Open Studio Read the docs

The managed studio runs on Hanzo Cloud GPUs.

Paste and ship

Paste this into any agent. It reads the skill manifest and calls Studio from there.

Agent prompt

Read https://hanzo.ai/skill.md and use Hanzo Studio in my project.

Studio — Hanzo

Every request enters here, so the limiter does too.

Open in the console

Example project. Every name in it is one you would give your own.

What is Hanzo Studio

A node graph for generative AI

The steps you would otherwise script by hand — loading models, prompting, sampling, and post-processing — become nodes you wire together on a canvas and run live.

Visual node editor

Drag, drop, and connect nodes into a flowchart. Compose models, prompts, and tools into advanced pipelines on an infinite canvas — no code, no glue.

Every modality

One canvas for image, video, audio and 3D, plus the editing and upscaling steps after them. The checkpoints are the ones you put in the models folder — the graph does not care which family they came from.

Prototype to production

Run on your own GPU or CPU, fully offline, then route heavy prompts to distributed cloud workers that queue and scale automatically.

Capabilities

From canvas to production

Live iteration

Only the nodes you change re-run, with latent previews as images render — fast, tight feedback on every tweak.

Portable workflows

Save and load workflows as JSON, or drop in a generated PNG, WebP, or FLAC to restore the exact graph and seed that made it.

Hanzo Engine & Creative nodes

First-class nodes call Hanzo Engine and Creative AI research previews (Koma video, Iro image, Mori 3D, Oto foley, Koe voice, Uta music) — mix foundation reasoning, decision heads, and generative diffusion in one graph.

Studio Copilot

A sidebar assistant that edits the graph in plain language — add nodes, connect them, set prompts, and queue a run.

Custom nodes

A node is a Python class with declared inputs and outputs, so the graph is extended by writing one rather than by waiting for us to ship it. Installed nodes are pinned, which is what makes a graph reproducible later.