nvidia/sana 🔢📝❓ → 🖼️

▶️ 203.2K runs 📅 Nov 2024 ⚙️ Cog 0.13.3 🔗 GitHub 📄 Paper ⚖️ License
high-resolution multilingual text-to-image

About

A fast image model with wide artistic range and resolutions up to 4096x4096

Example Output

Prompt:

"a cyberpunk cat with a neon sign that says "Sana""

Output

Example output

Performance Metrics

1.20s Prediction Time
1.21s Total Time
All Input Parameters 
{
  "width": 1024,
  "height": 1024,
  "prompt": "a cyberpunk cat with a neon sign that says \"Sana\"",
  "guidance_scale": 5,
  "negative_prompt": "",
  "pag_guidance_scale": 2,
  "num_inference_steps": 18
}
Input Parameters
seed Type: integer
Random seed. Leave blank to randomize the seed
width Type: integerDefault: 1024
Width of output image
height Type: integerDefault: 1024
Height of output image
prompt Type: stringDefault: a cyberpunk cat with a neon sign that says "Sana"
Input prompt
model_variant Default: 1600M-1024px
Model variant. 1600M variants are slower but produce higher quality than 600M, 1024px variants are optimized for 1024x1024px images, 512px variants are optimized for 512x512px images, 'multilang' variants can be prompted in both English and Chinese
guidance_scale Type: numberDefault: 5Range: 1 - 20
Classifier-free guidance scale
negative_prompt Type: stringDefault:
Specify things to not see in the output
pag_guidance_scale Type: numberDefault: 2Range: 1 - 20
PAG Guidance scale
num_inference_steps Type: integerDefault: 18Range: 1 - ∞
Number of denoising steps
Output Schema

Output

Type: string • Format: uri

Example Execution Logs 
Using seed: 49609
  0%|          | 0/17 [00:00<?, ?it/s]
 18%|█▊        | 3/17 [00:00<00:00, 22.20it/s]
 35%|███▌      | 6/17 [00:00<00:00, 22.18it/s]
 53%|█████▎    | 9/17 [00:00<00:00, 22.17it/s]
 71%|███████   | 12/17 [00:00<00:00, 22.17it/s]
 88%|████████▊ | 15/17 [00:00<00:00, 22.17it/s]
100%|██████████| 17/17 [00:00<00:00, 23.56it/s]
Version Details
Version ID
c6b5d2b7459910fec94432e9e1203c3cdce92d6db20f714f1355747990b52fa6
Version Created
December 6, 2024
Run on Replicate →