Questions, answered

Shots that mix photographed plates, CG and machine learning: a generated set extension relit to match the plate, a depth pass from a model driving fog, an ML matte refined with roto. Triton puts all of that in one node graph, on the same frames and in the same colour space.

Those tools are a thin interface over a model library and stop at the image. Triton is a native app with a timeline, frame cache, OCIO colour, multi-layer EXR, roto, tracking, keying and 3D, so the model output goes straight into finishing the shot.

Older compositing apps add ML as a handful of single-purpose tools. Triton was designed around hybrid work from the start: ML nodes write ordinary channels that every other node can use, and local and cloud models share the same graph, cache and job queue.

The app is usable for real work but still changing quickly. Expect rough edges and frequent updates. Founder licenses bought during Early Access cost $99 and are otherwise the same as Indie, with 12 months of updates.

Free is the whole app for non-commercial use, with renders capped at 1920 x 1080. Indie removes the cap and allows commercial work for individuals and companies under $100k annual revenue.

No. Indie is a one-time payment with 12 months of updates. You keep the last version from that window forever. Renewing for more updates is optional.

With prepaid credits, like a phone top-up. Buy a $10, $25, $50 or $100 pack; credits are dollars and last 12 months. Each successful cloud job is charged for the GPU-seconds it used at the published per-hour rate. Failed jobs and time spent waiting in the queue are free. There is no subscription.

macOS on Apple Silicon and Intel, and 64-bit Windows. Linux ships as an AppImage and is community supported during Early Access.

Not to work. The Free edition needs no account at all. A paid license key is checked when you activate it and then works offline for 30 days at a time. Only cloud GPU nodes need a connection while they run.

Local ML nodes run on your machine through ONNX Runtime and never upload frames. Models are downloaded once from inside the app. Cloud GPU nodes upload only the inputs you connect to them, and only when you run them.

Not yet. OFX hosting is being evaluated; the most requested effects are being built as native nodes first.

FilmSimulation uses spectral profiles adapted from the spektrafilm project's published datasheet data, under CC BY-SA 4.0. The profile files ship as readable JSON with attribution. Stock names describe what a profile emulates; Triton is not affiliated with Kodak or Fujifilm.

Community support runs through the Triton Discord, where bugs and requests are tracked. For licensing or billing, use the contact form. There is no guaranteed response time during Early Access.

Triton is made and licensed by Medeu Global LLC. Orders are processed by Polar.sh as merchant of record, so your receipt and any sales tax or VAT come from Polar.