Explore AI model collections

Alibaba's Wan 3.0 image-to-video starts from an opening frame and directs 2 to 30 seconds of motion with generated audio. An optional closing frame pins where the shot lands, at 480p, 720p or 1080p.

ByteDance's Seedance 2.5 image-to-video starts the clip on a still you supply and animates it, with sound in the same pass. Clips run 4 to 30 seconds at 480p, 720p or 1080p, and an optional closing still makes the shot arrive on a set frame.

Seedance 2.5 Spicy is ByteDance's uncensored image-to-video model, animating a still you supply as the first frame, with sound in the same pass. Clips run 4 to 30 seconds at 480p, 720p or 1080p native, plus an upscaled 4K tier.

Seedream 5.0 Pro image editing takes one to ten reference images plus an instruction and returns one image. Revise a single photo or assemble subject, wardrobe and set into one frame, in fifteen aspect ratios at 1K, 1.5K or 2K. Uncensored.

Qwen Image 2.1 image editing revises one to ten reference images from a written instruction and returns one frame at 1k, 1.5k or 2k. It also does local edits with a black-and-white mask, where white is repainted.

Wan 2.7 Pro image-to-image edits or composites one to nine reference images under a single instruction, at a 2k canvas of about 4 megapixels. It is built to hold compound, multi-part edit prompts, and this endpoint is uncensored.

Z-Image Spicy Pro is an uncensored text-to-image endpoint with exact width and height set independently up to 2560 pixels per side. The prompt has no character cap and the prompt expander is off by default.

Seedream 5.0 Pro image editing takes one to ten reference images plus an instruction and returns one image. Revise a single photo or assemble subject, wardrobe and set into one frame, in fifteen aspect ratios at 1K, 1.5K or 2K. Uncensored.

Alibaba's Wan 3.0 image-to-video starts from an opening frame and directs 2 to 30 seconds of motion with generated audio. An optional closing frame pins where the shot lands, at 480p, 720p or 1080p.

Kling 3.0 Motion Control copies the moves, expressions and gestures of a 3 to 30 second driving clip onto one character photo, at 720p or 1080p. The result runs exactly as long as the driving video.

Lip Sync redraws the mouth of the person in a video to match a new speech track of 3 to 60 seconds, leaving performance, lighting and framing as shot. It returns an MP4 with your audio, keeping resolution up to 720p.

Kling Avatar 2.0 talking avatar turns one portrait and an audio file into a clip of that person speaking or singing it. Render at 720p or 1080p; the clip length follows your audio, up to about 5 minutes.

Seedream 5.0 Pro text-to-image is ByteDance's high-detail image model, turning a prompt of up to 5,000 characters into one frame. Choose from fifteen aspect ratios at 1K, 1.5K or 2K, in JPEG or PNG. Uncensored.

Qwen Image 2.1 text-to-image is Alibaba's open-weight model: a 5,000-character prompt becomes one image at 1k, 1.5k or 2k in fifteen aspect ratios, as JPEG, PNG or WebP. Transparent backgrounds are available, and there is no content filter.

Z-Image Spicy Pro is an uncensored text-to-image endpoint with exact width and height set independently up to 2560 pixels per side. The prompt has no character cap and the prompt expander is off by default.