What is the SANA image generator?
SANA is NVIDIA's linear-attention diffusion model that generates a 1K image in about a second — the cheapest and one of the fastest options here. Type a prompt, pick a size and count, and download the results — every run is saved to your history. Sign in to get 200 free credits every month.
What is SANA good at?
- Near-instant generation (around a second per image)
- Lowest price tier — practically free to iterate
- Good quality for the speed, especially illustrations
How to generate an image
- Sign in (free) — every account gets 200 free credits per month.
- Describe the image you want in the prompt box — for example "a minimal product photo of a ceramic mug on a linen cloth, soft morning light".
- Choose a size (square / portrait / landscape / wide) and how many images to generate — the exact credit cost is shown before you run.
- Press generate and wait a few seconds to a minute depending on the model.
- Download the results. Every run is saved to your history for re-download; failed runs are fully refunded.
SANA pricing
SANA costs about 3 credits per 1024×1024 image (1 credit ≈ ¥1), up to 4 images per run. The 200 free credits every account gets each month are enough to try it properly, and failed generations are fully refunded.
Which AI image model should you use? (comparison)
All models below are commercially usable and share the same credits. Prices are per 1024×1024 image.
| Model | Made by | Best for | Credits / image (Free plan) | Credits / image (Paid plans) |
|---|---|---|---|---|
| FLUX.1 [schnell] | Black Forest Labs | Rapid drafts, thumbnails, bulk generation | 3 | 3 |
| FLUX.1 [dev] | Black Forest Labs | High-quality general-purpose generation | 21 | 9 |
| FLUX.1 Krea [dev] | Black Forest Labs × Krea | Natural photorealistic visuals | 21 | 9 |
| FLUX1.1 [pro] | Black Forest Labs | Production-grade visuals | 34 | 14 |
| FLUX1.1 [pro] ultra | Black Forest Labs | High-resolution hero images | — | 20 |
| Qwen-Image | Alibaba (Qwen) | Images with (Japanese) text | 17 | 7 |
| SANA (this page) | NVIDIA | Instant, ultra-cheap drafts | 3 | 3 |
| Stable Diffusion 3.5 Large | Stability AI | Broad style range | — | 22 |
| Nano Banana | Google (Gemini 2.5 Flash Image) | Instruction-accurate scenes | 32 | 13 |
| Ideogram V3 | Ideogram | Logos & posters with exact text | — | 20 |
| Recraft V3 | Recraft | Design & brand assets | 32 | 13 |
| GPT Image 2 | OpenAI | Knowledge-heavy, instruction-heavy prompts | 43 | 17 |
| Seedream 4.0 | ByteDance | Quality per credit | 24 | 10 |
| Seedream 4.5 | ByteDance | Portraits & cinematic looks | 32 | 13 |
| FLUX.2 [pro] | Black Forest Labs | Flagship production quality | 24 | 10 |
| FLUX.2 [dev] | Black Forest Labs | Cutting-edge open model | 11 | 5 |
| FLUX.2 [klein] | Black Forest Labs | Fast + cheap FLUX.2 quality | 6 | 3 |
| Nano Banana Pro | Google (Gemini 3 Pro Image) | Most demanding scenes | — | 48 |
| Nano Banana 2 | Google (Gemini 3.1 Flash Image) | Pro-like quality, faster | — | 26 |
| Z-Image Turbo | Alibaba (Tongyi) | Fast photoreal on a budget | 5 | 3 |
| Krea 2 | Krea | Opinionated aesthetics | — | 20 |
| Grok Imagine | xAI | Cheap photoreal exploration | 16 | 7 |
| Nano Banana Lite | Google (Gemini 3.1 Flash-Lite Image) | Google quality, fastest and cheapest | 28 | 11 |
