What are we creating today?
Video, image and speech models with a fixed price per run — in a native Mac app, over an API, or straight from your agent.
Apple Silicon · signed and notarized · free to start
Build, generate, and market anything with skills, connectors, and automation
The same catalogue behind a trailer, an explainer, a batch pipeline, or an agent that renders on your behalf.
Product launch trailer
Write the beats, render each shot at a known price, cut them together in one job.
Ad variants from one brief
Same prompt, different hooks and aspect ratios, priced per run before you commit.
Narrated explainer
Speech billed per character drives the visuals, so the voice track leads the cut.
Talking-head explainer
Record the voiceover first and let it drive a presenter, lip sync included.
Batch renders on a schedule
Submit jobs, get a webhook when each finishes. No session to babysit.
Tools your agent can reach
Connect an MCP server once and call it from any client, without pasting keys around.
Game trailer from key art
Start from a still you already own and move the camera through it.
Character turnaround
Hold a character across shots by driving every render from the same reference frame.
The models people ask for
All 76 models →A price before the render, not after
Every model in the catalogue quotes a fixed price per run, resolved before the job is submitted. A request we cannot price is refused rather than guessed at — no meter left running because a loop misbehaved.
See pricing →One router, not one vendor
Media and text models reach you through the same router and the same account. Which company hosts a model is our problem to solve, not a decision you have to re-make every time the market moves.
Browse the catalogue →Where you already work
A native macOS app for hands-on work, and a plain HTTP API with webhooks for pipelines. Agents reach your other tools through the MCP Gateway, which is a separate product on the same account.
See the MCP Gateway →Three steps, and none of them is provisioning a GPU
Choose the model, or let the agent
Name the model yourself in the catalogue, the API or MCP; nothing picks one behind your back. Hand the brief to the video agent instead and it chooses. Either way the run price is on screen before anything starts.
Run it from anywhere
Render in the macOS app or POST to the API. Same catalogue, same price, either way.
Get the file and the receipt
Outputs arrive as expiring signed links; a webhook fires when the job lands. Every run is itemised in usage.
Weighing it against Higgsfield, fal.ai, or going direct?
We wrote the comparison down, including the parts where another tool is the better buy.
Latest writing
All posts →How to make a game trailer with AI, shot by shot
A trailer is six to ten short clips in a deliberate order, not one long generation. Start from key art you already own, write the shot list before you render, and price the whole thing per second first.
Camera control with Wan: the words that actually move a camera
Writing "cinematic" does nothing. Describing a shot and describing a move are different sentences, and the negative constraints do most of the work.
Agentic AI for content creation: what an agent can actually take off your hands
Variant sweeps, retries and filing are delegable. Taste, the brief and clearance are not. A fixed price per run is what makes an autonomous batch a budget rather than a bill.
Start with a render, not a rental
Free to start. Top up when you need more, or move onto RH Pass once the volume is steady.