Self-Host Studio is a self-hosted platform for AI and business workflow automation. It runs LLM pipelines, GPU image and video generation, and the everyday business processes around them, with approval steps where you want a person in the loop and a record of every run. Unlimited runs, because they happen on your machines.
Everything ships as three named artifacts. Two of them are one install away on PyPI.
The platform. Deterministic, repeatable workflow runs, approval gates for human sign-off, organizations with isolated credentials, audit logging, and provider integrations defined by schema.
The way in. An operator CLI that installs, configures, upgrades, backs up, and restores your deployment. Nobody clones a repo or builds containers to get started.
The extra machines. A standalone package that runs worker engines on other hardware, pointed at your Studio API. Put image, video, and audio generation on the box with the GPU.
Studio Console handles the whole setup. You need Docker with Compose v2; the first run opens a setup wizard, then starts the stack.
# install the operator CLI (or: pip install studio-console) uv tool install studio-console # first run opens the setup wizard studio-console
Deployment guides, workflow templates, and the developer are in the community.
Join Self-Host HubCloud workflow tools chain HTTP calls and charge you per operation. AI frameworks wire up models and agents. Self-Host Studio is the layer underneath, where your jobs run, your compute routes, your files persist, and each organization's data stays its own.
You give it a prompt. An LLM writes the script. A GPU model generates your scenes. A TTS engine narrates. A video compositor renders the final output. Every step is logged. Every file is stored. And plenty of production workflows have no LLM step at all: lead follow-up, client onboarding, record syncs, notification bots. Same engine, same tracking.
A real production pipeline running in Self-Host Studio, not a mockup. It crosses LLM, GPU image generation, text-to-speech, video compositing, and delivery in a single tracked run you own end to end.
You build a pipeline once as a template. Install it into any organization, bind your own credentials, adjust parameters, and it becomes a workflow. Every run creates an instance, a frozen execution with per-step tracking and output records.
Browse and install pre-built prompt templates: story generators, scene designers, content summarizers, tone rewriters, and more. Each template is categorized, versioned, and ready to drop into your workflows. Build your own and share them across your organizations.
Self-Host Studio isn't competing on node count. It's the platform that runs your AI-heavy, compute-heavy, multi-organization workloads.
Jobs dispatch to workers by capability. General workers make your HTTP calls. GPU workers run ComfyUI. Video workers run FFmpeg. Audio workers run TTS. When you need a new modality, you add a label, not an architectural rewrite.
Every organization you create is fully separate: credentials, workflows, and files. Scale from one org to a thousand. White-label branding means your clients see your platform, not ours.
Providers extend Self-Host Studio without touching core code. Each is a self-contained JSON package: manifest, auth config, and service definitions. The platform renders your configuration UI, routes your jobs, and processes results directly from the schema.
You get full step-by-step run tracking: per-step status, file outputs, input data inspection, rerun controls, and timing. You see everything: which worker ran, what was sent, what came back, and how long it took.
Your clients see a clean progress screen: steps completing with checkmarks, generated images appearing in real time, and a progress bar at the bottom. They click "Run" and watch their results come in. No training required.
Self-Host Studio scales with you: solo creator, agency, or SaaS platform.
You've outgrown scripts and copy-paste workflows
Deterministic, repeatable runs. Your API keys, your GPU, your production system.
Per-operation pricing is compressing your margins
Deliver infrastructure your clients own. Your IP compounds on top of their deployment.
You're facing 6–12 months building a backend from scratch
Organizations, credential isolation, and audit logging, already built.
Self-Host Studio isn't competing with cloud automation tools. It's where you move when those tools become your bottleneck.
They chain HTTP calls, charge per operation, and hold no state, so you end up renting Airtable just to store records and fire triggers
Self-Host Studio has storage, run records, and scheduling built in, and runs the compute-heavy steps on your own hardware.
They orchestrate models and agents, but give you no visual builder or organizations
Self-Host Studio is your platform, not another model wrapper.
API, job queue, GPU scheduling, audit logging: that's 6–12 months of your time
That backend already exists, and Studio Console installs it in one command.
You're generating content on someone else's infrastructure under someone else's terms
Run the same generation on your own GPU, with no per-image fee.
Studio Console sets up a multi-container stack on your server, or a single container when you want the smallest footprint. Upgrades, backups, and restores come from the same tool.
Start on one machine. When you need more compute, install studio-workers on another box and point it at your API. Your workflows don't change.
You can't redistribute Self-Host Studio as a competing platform, but you can build your business on it: sell your consulting, your templates, and your workflow configurations, and commercialize anything your pipelines produce.
Output rights follow the licenses of the third party models, nodes, and API services your workflows use.
As your business grows, the platform grows with you. New providers, templates, and prompt libraries ship through the marketplace, and updates bring performance improvements and current AI capabilities.
The engine is free. The ecosystem is where your value compounds.
Join the Self-Host Hub community for everything you need to deploy:
One membership. Unlimited runs on your own hardware.
Join Self-Host Hub