AI models and what they cost
HyperFuse runs 36 image and video models behind one interface. Every generation is priced in credits, and the figures below are read from the same live catalogue the studios use — so what you see here is what you are charged.
New accounts start with 12 free credits — no card required. The cheapest model runs at 1 credit per generation.
Image models
Text-to-image and image editing. Cost is per image; several models generate a batch in one call.
| Model | What it is best at | Credits | Capabilities |
|---|---|---|---|
| Flux Schnell | instant drafts, lowest cost | 1 | up to 4 per call |
| Flux Dev | good quality without the top price | 2 | up to 4 per call |
| Flux Pro | sharpest detail of the Flux family | 4 | up to 4 per call |
| Seedream 4.5 | photoreal, and gets text right | 3 | up to 4 per call |
| Recraft v3 | logos, icons and clean vector art | 3 | — |
| Nano Banana | follows instructions most closely | 2 | up to 4 per call |
| Qwen Image | crisp detail and legible text | 2 | up to 4 per call |
| Flux 2 | newest model, most realistic results | 4 | up to 4 per call |
| Flux 2 Pro | sharpest Flux, best prompt control | 4 | up to 4 per call |
| Nano Banana 2 | Google's newest, strong realism | 5 | up to 4 per call |
| Z-Image Turbo | ultra-fast drafts and variations | 1 | up to 4 per call |
| Ideogram 3 | posters and anything with words in it | 4 | up to 4 per call |
Video models
Text-to-video and image-to-video. Video cost varies with duration and resolution; the studio quotes the exact figure before you generate.
| Model | What it is best at | Credits | Capabilities |
|---|---|---|---|
| LTX 1 | cheapest draft, roughest quality | 5 | 5s |
| Kling 1.6 | reliable motion, older generation | 8 | image-to-video · 5s / 10s |
| Kling 1.6 Pro | steadier motion than 1.6 | 13 | image-to-video · 5s / 10s |
| Hailuo 02 | expressive people and faces | 8 | image-to-video · 6s / 10s |
| Seedance Pro | cinematic, follows the prompt closely | 17 | image-to-video · 5s / 10s |
| Veo 3.1 | best realism, native sound, start and end frames | 22 | image-to-video · native audio · 4s / 6s / 8s |
| Veo 3.1 Fast | Google quality at half the price | 12 | image-to-video · native audio · 4s / 6s / 8s |
| Wan | turn a still image into motion | 11 | image-to-video · 5s |
| Kling 3 Pro | highest quality, native sound, multi-shot | 16 | image-to-video · native audio · 5s / 10s / 15s |
| Kling 3 | flagship quality for everyday shots | 12 | image-to-video · native audio · 5s / 10s / 15s |
| Kling 3 4K | native Ultra HD, holds your character | 60 | image-to-video · native audio · 5s / 10s |
| Kling 2.6 Pro | excellent motion, best value | 11 | image-to-video · native audio · 5s / 10s |
| LTX 2 | sharp drafts at a low price | 10 | image-to-video · 6s / 8s / 10s |
| LTX 2 Fast | quickest way to test an idea | 9 | image-to-video · 6s / 8s / 10s |
| Kandinsky 5 | low-cost motion studies and concept clips | 6 | 5s / 10s |
| Sora 2 | OpenAI realism and physics, sound included | 12 | image-to-video · 4s / 8s / 12s |
| Hailuo 02 Pro | 1080p people and faces, fixed 6 s | 14 | image-to-video · 6s |
| Kling 2.5 Turbo Pro | fast, fluid motion at a mid price | 10 | image-to-video · 5s / 10s |
| Veo 3.1 Lite | Google quality on a budget, sound included | 6 | image-to-video · native audio · 4s / 6s / 8s |
| Vidu Q1 Reference | up to 7 character references, strongest identity hold | 16 | 5s |
| Kling 1.6 Elements | character references straight into the shot | 13 | 5s / 10s |
| Kling O1 Reference | named character elements, best video identity | 24 | 5s / 10s |
| Wan 2.7 | newest Wan, sharp motion at 1080p | 14 | image-to-video · 5s / 10s |
| Wan 2.6 | realistic physics and fine detail | 14 | image-to-video · 5s / 10s |
How credits work
Priced per model
Each model has its own credit price. Video also varies with duration and resolution.
Failures are refunded
If a provider fails, the credits charged for that job are returned automatically.
Quoted before you spend
The studio shows the exact cost before a generation starts, so nothing is a surprise.