AI Models

Generate short sound effects from a text description — impacts, whooshes, fire, explosions, UI clicks, and other foley.

Generate layered scene audio from a text description — room tone, ambience, and a sequence of events in one pass, optionally guided by up to three reference clips. Use this instead of Sound Effects when you want an atmosphere rather than a single isolated hit.

Repair photos at their existing resolution: sharpen out-of-focus and motion-blurred shots, clean high-ISO noise, restore damaged or degraded images, and correct exposure, white balance and colour.

Repair and retime video without enlarging it: stabilize shaky footage, remove noise and motion blur, interpolate to a higher frame rate or slow motion, colorize black-and-white sources, and convert SDR to HDR.

Generate text. Supports max_tokens and system_prompt.

Generate text. Supports temperature, max_tokens, and system_prompt.

Generate text. Supports temperature, max_tokens, and system_prompt.

Generate text. Supports max_tokens and system_prompt.

Generate text. Supports temperature, max_tokens, and system_prompt.

Generate text. Supports temperature, max_tokens, and system_prompt.

Fast LTX 2.5 video generation with native audio, up to 4K and 20 seconds

High-fidelity LTX 2.5 video generation with native audio, up to 1080p and 10 seconds

Wan 3 Prime, the faster Wan 3 endpoint. Hyper-real video up to 30 seconds with native audio, first/last frame control, and reference-to-video from up to 10 images, 5 clips, and 5 voice references.

Generate complete songs up to five minutes long from a music description and lyrics. Structure lyrics with tags like [verse] and [chorus] on their own lines.

Native-audio video with first/last frame control, up to 20 seconds and 1080p.

Automatically cut filler words ("um", "uh", "you know") and overlong silences out of a finished talking video. The result is a tighter re-cut of the same footage — no re-shoot, no re-render, and the original audio and framing are preserved between cuts.

Fast 768P video with native stereo audio, strong prompt adherence, and text / image / reference-to-video modes.

Seedance 2.5 renders single takes up to 30 seconds with native stereo audio, keyframe transitions, ultrawide aspect ratios, and Omni Reference across up to 30 images, 10 video clips, and 10 audio clips.

Generate text. Supports max_tokens and system_prompt.

Turn an image or a prompt into a navigable 3D world. Returns a Gaussian splat scene you can fly a camera through — generate a location once, find the exact angle, and capture the frame as a reference for video generation.

Strong text rendering and prompt adherence with flexible resolutions up to 2048×2048.

Generate text. Supports temperature, max_tokens, and system_prompt.

Generate text. Supports temperature, max_tokens, and system_prompt.

Generate text. Supports temperature, max_tokens, and system_prompt.

Flagship Seedream image model with multi-reference editing, strong prompt adherence, and up to 2K output.

Hunyuan video generation

Kling v2.6 Pro video generation

Fast high-fidelity text-to-image. Choose the LoRA endpoint for custom LoRAs, or the uncensored endpoint (no content filtering, prompt-only).

LTX-2 19B Distilled high-quality video generation

Fast LTX 2.3 video generation with audio, up to 4K and 20s duration

High-fidelity LTX 2.3 Pro video generation with audio, up to 4K and 48fps

Luma Dream Machine Ray 2 video generation

Magi video generation

Flux Pro Kontext Max text-to-image

Flux Krea high quality generation

FLUX.1 SRPO [dev] high-quality image generation