Getting started
Zilla is a local-first, bring-your-own-keys creative router. You install one desktop app, paste your provider keys, and route every image, video and audio generation through them — at cost, with zero markup. This guide gets you from download to first render in about two minutes.
Install
Download the desktop app for macOS or Windows from the home page and open it. There's no account wall and no cloud round-trip — generations save straight to your machine.
- macOS: open the .dmg and drag Zilla into Applications.
- Windows: run the installer and launch from the Start menu.
- On first launch Zilla creates a local library folder for your outputs — you can change it later in settings.
Add your keys
Zilla never holds credit. You connect your own provider keys and pay the API directly. Open Account → Connected keys and paste a key from any supported provider:
- kie.ai — broadest model coverage (images + video).
- fal.ai — fast inference for many of the same models.
- ElevenLabs optional — speech and audio.
Keys are stored encrypted on your device only and used to call providers directly. They never pass through Zilla's servers — we couldn't see your spend if we tried.
First generation
In the Generate tab, pick a model and output type, type a prompt, and hit Generate. The cost label shows the exact at-cost price before you run anything. Results land in your library instantly and can be assigned to projects.
MCP setup
Zilla runs a local MCP server that exposes every model as an agent tool. Point any MCP client — Claude Desktop, Claude Code, or your own — at it and drive generation, editing and renders straight from your agent. Add this to your client's MCP config:
// add Zilla to your MCP clients { "mcpServers": { "zilla": { "command": "zilla", "args": ["mcp"], "env": { "ZILLA_PORT": "7373" } } } }
Restart your client and the zilla tools appear. The app's header shows an AI indicator when an agent is connected.
Available tools
- zilla_generate_image — render an image from a prompt + model.
- zilla_generate_video — render a clip, optionally with start/end frames.
- zilla_generate_speech — synthesize audio.
- zilla_list_models — list models with live at-cost pricing.
- zilla_check_cost — preview the cost of a run before spending.
- zilla_render_composition — render a multi-clip composition in the editor.
Model catalog
One router, every frontier model — routed through your keys. Pricing is at cost and shown live in the app; the figures below are representative.
Pricing & cost
Generations cost exactly what the provider charges — 0% markup. The only thing Zilla bills is a flat $9/month for the software. Every run is logged with its true cost, split by manual vs MCP, so you always know what you've spent to the cent. See Manage for billing.
Looking for something else? Browse the Help Center or email support@zilla.studio.