AI models

Every model is available on every plan, in both Create Studio and the Node Editor. The model picker is the same control in both places: models grouped by company, with a hairline between groups, a NEW badge on recent arrivals, and the same name for the same model everywhere. Pick a model in Studio and it behaves exactly as it does inside an Image Generate or Video Generate node.

Image models

From OpenAI, Google, BytePlus, Black Forest Labs, Runway and others. Some take reference pictures, some edit an existing picture, some render a transparent background.

Video models

Kling, Seedance, Veo, Runway, WAN, MiniMax and more. They differ in what they start from: references, a start frame, or a clip to edit or extend, and in whether they make sound.

Text models

The language models behind Text Generate, from OpenAI and Google, with their attachment limits.

Engines inside tools

Upscale, Remove Background, Alpha Matte, Video Upscale and Motion Control each have their own small set of models that never appear in a generate menu, because they take no prompt.

Each model carries a declaration of what it can do, and every surface reads it. A control exists only when the selected model supports it, and a missing control means a missing feature, not a hidden setting. That is why the composer and the nodes change shape when you switch models.

Shape and size

A list of aspect ratios, or exact pixel sizes with custom dimensions on models that take them. Some offer several resolutions.

Quality tiers

Low, Medium, High and, on some models, higher steps still. They use the vendor's own levels, and the picker shows a note when one model's High matches another's Medium.

References

Whether the model takes reference pictures, how many, in which formats and up to what size. Models that take none show no reference strip.

Variants

How many results one press can make. Each is a separate generation, charged in full.

Video inputs

Clip length choices or an automatic length, sound on or off, a start frame with or without an end frame, editing or extending an existing clip, and references the prompt names by number.

Constraints

Video models can tie settings together: a resolution that rules out sound, a start frame that fixes the length. The node resolves those rules before showing the menus and says why an option is greyed.

Every price you see, in the picker, on a Run button and on a model page, comes from one price book kept in the database, the same rows the server charges from. A model's base price is derived from what it really costs to run; options such as resolution, quality, clip length, variants and reference images add to it. Prices can change without a release, so the button is the thing to trust.

The product must stay exact

Pick a model that takes reference pictures and keep the product photo among them. The reference count in the picker tells you which ones do.

Change a picture you already have

Pick an edit-capable image model and attach the picture. The prompt then says what changes rather than what to make.

Speed over polish

Each family tends to have a fast or lite version beside its flagship. Use it for drafts and switch up for the final.

A clip with sound

Look for models that declare audio. On some it depends on the resolution; the node tells you.

Continue a shot

Pick a video model with a start frame, and feed it the last frame of the clip before it.

Not sure

Run the same prompt on two or three models in Studio with one variant each. The picker makes switching cheap, and the results sit side by side on the canvas.

Models come and go. A retired model leaves the pickers, so nothing new is made with it, but it does not break what you built. A saved workflow that names it is resolved to the closest live sibling, the one that took its place, never a cheaper or weaker model, so a re-run costs what you would expect. A cloned workflow template whose model has gone opens that node with the picker empty, for you to choose.

Each family has a page on the site, such as Kling, Seedance, Nano Banana, Veo, Seedream, WAN, Runway, GPT Image and FLUX: what the model is, its versions, its current price, how to use it in Studio and the editor, what it makes, and the questions people ask. New models are added as they come out, with no plan gate.