Images and video
Make images and video clips, and see what each one will take from your usage before you generate it.
- Make an image
- Choose one of 38 models, a shape, and a size. Add reference pictures from an upload or from Files. Each result is saved to Files.
- Make a video
- Switch Image to Video, choose a model and a clip length, and press Generate. A clip takes a minute or two and is saved to Files.
- Extend, recast, or lip sync
- Extend continues a video with Quality or Turbo. Recast puts the people in your photos into a video. Lip Sync makes the face in a picture speak or sing an audio clip.
- Enhance an image
- Enhance upscales the image with Topaz and adds the result as a new pass. The original stays as it is.
- See the price before you generate
- The Images composer shows what a pass will take from your usage. It reads About when the final amount depends on what the model returns.
- Ask for it in a chat
- Turn on Image Generation or Video Generation in Skills & Plugins, and Chat, Code, and Agents can make them when you ask.
Video models and edits
- MiniMax H3 Max Turbo: fast, low-cost clips from text or a starting picture.
- MiniMax H3 Max: the strongest H3 clips, with up to 4 reference pictures.
- FLUX 3 Max: clips up to 20 seconds with sound, from text or a starting picture.
- H3 Max Recast: puts the people in your reference photos into a video.
- H3 Max Lip Sync: makes the face in your picture speak or sing an audio clip.
- Extend: continues a video with Quality, or with Turbo for about half the price.
Questions
The Images studio has 38 image models, including GPT Image, Nano Banana, Flux, Grok Imagine, Seedream, Recraft, Qwen, Krea, MAI Image, and Muse. Each model lists its own shapes, sizes, and settings.
Yes. MiniMax H3 Max Turbo, H3 Max, and FLUX 3 Max make clips from a prompt or from a starting picture. H3 clips run 5 to 15 seconds, and FLUX 3 Max clips run 5 to 20 seconds.
Yes. The Images composer shows what the next pass will take from your usage before you press Generate. It reads About when the final amount depends on what the model returns.
Yes. When a clip could take more than $1 of your usage, Chat tells you the price and waits for your yes. Video Generation must be on in Skills & Plugins.
It upscales the image with Topaz Wonder 3.5 and adds the result as a new pass. Each side is doubled unless that would pass 4K, and the original stays as it is.
On Fal, which hosts the models. If Fal can't take an image request, the request goes to the AI Gateway instead, and Images tells you. Video has no backup host.