Model directory
Video, image and music models side by side. Prices come from the same pricing table the Generate button charges against, so what you read here is what you pay.
Cheapest way in
The lowest-priced model in each group, at its smallest size and shortest length. Enough to test a prompt before spending on a full render.
Text-to-video and image-to-video. Prices below are the cheapest combination of resolution and length each model offers; picking a longer clip or a higher resolution raises the cost before you generate, never after.
3 models
Kuaishou's video models. Kling 3.0 Pro renders at 4, 8 or 12 seconds; the 2.5 Turbo Pro and 2.6 rows run on Kling's own 5- and 10-second steps.
3 models
ByteDance's video family. Seedance 2.5 is the newest, 2.0 sits in the middle, and 1.5 Pro is the cheapest way to test an idea before committing to a full render.
1 model
xAI's Imagine video model. Priced near the bottom of the video range, which makes it the one to iterate on before moving a prompt to a heavier model.
1 model
Google DeepMind's Veo, in its 3.1 Fast configuration.
1 model
MiniMax's Hailuo H3, with separate text-to-video and image-to-video entry points.
1 model
HappyHorse 1.1, a stylised video model sitting between the fast and the premium tiers on price.
Text-to-image and image-to-image. Most families ship both directions under one brand, and the image-to-image variants accept reference images.
3 models
ByteDance's image family. Every row has a matching image-to-image variant that takes up to 10 reference images, so an edit keeps the look of what you feed it.
3 models
Google's Nano Banana line. Nano Banana 2 for everyday work, Pro when you need the extra detail, and the original as the cheapest way to iterate on a prompt.
2 models
GPT Image 2 and 1.5. The 1.5 rows let you choose a medium or high quality tier at generation time and price each one separately; both generations do text-to-image and image-to-image.
Full songs from a prompt — lyrics, vocals and instrumental in one generation.
1 model
Suno V5 writes and performs the whole track — lyrics, vocals and instrumental — from a single description.
Some jobs cut across every model family — changing a photo you already have, turning a still into a clip. These guides line up each model that does the job next to what it charges, then hand you to the create page. They are guides, not free tools: the generation itself spends credits.
Image models
Start from a picture instead of a blank prompt: upload it, describe the change, and get a new version back — a different style, a fixed detail, a cleaner background. The guide compares every model that accepts a reference image.
