Browse by capability
You usually arrive at a model by the thing it has to do. Each page below lists the models on our gateway that handle a specific capability, with prices.
Readable text in image
6 models
Most diffusion models mangle letters. These models render crisp, specifiable typography directly in the image, which unlocks posters, ads and diagrams without a design tool.
Instruction-based image editing
8 models
Upload an image, describe the change in a sentence — restyle, relight, swap a background — and keep the subject intact. These models accept an image reference plus an edit instruction.
Realistic human motion
3 models
Human performance is the hardest thing to fake in generated video. These models are known for natural body mechanics, expressive faces and clean physics.
Video with native audio
8 models
A clip that already has sound skips a post-production step. These models generate the audio track and the video together.
Image-to-video animation
1 model
Start from a still — a product shot, a portrait, a landscape — and animate it with a motion prompt. Every video model below accepts an image as the first frame.