nvidia/sana-sprint-1.6b 🔢📝❓ → 🖼️

▶️ 897.4K runs 📅 Mar 2025 ⚙️ Cog 0.16.0 🔗 GitHub 📄 Paper ⚖️ License
low-latency one-step-diffusion real-time text-to-image

About

SANA-Sprint: One-Step Diffusion with Continuous-Time Consistency Distillation

Example Output

Prompt:

"a tiny astronaut hatching from an egg on the moon"

Output

Example output

Performance Metrics

0.14s Prediction Time
0.15s Total Time
All Input Parameters 
{
  "seed": -1,
  "width": 1024,
  "height": 1024,
  "prompt": "a tiny astronaut hatching from an egg on the moon",
  "output_format": "jpg",
  "guidance_scale": 4.5,
  "output_quality": 80,
  "inference_steps": 2,
  "intermediate_timesteps": 1.3
}
Input Parameters
seed Type: integerDefault: -1
Seed value. Set to a value less than 0 to randomize the seed
width Type: integerDefault: 1024Range: 256 - 4096
Width of output image
height Type: integerDefault: 1024Range: 256 - 4096
Height of output image
prompt Type: stringDefault: a tiny astronaut hatching from an egg on the moon
Input prompt
output_format Default: jpg
Format of the output images
guidance_scale Type: numberDefault: 4.5Range: 1 - 20
CFG guidance scale
output_quality Type: integerDefault: 80Range: 0 - 100
Quality when saving the output images, from 0 to 100. 100 is best quality, 0 is lowest quality. Not relevant for .png outputs
inference_steps Type: integerDefault: 2Range: 1 - 4
Number of sampling steps
intermediate_timesteps Type: numberDefault: 1.3Range: 1 - 1.5
Intermediate timestep value (only used when inference_steps=2, recommended values: 1.0-1.4)
Output Schema

Output

Type: string • Format: uri

Example Execution Logs 
Using seed: 50650
Using intermediate_timesteps: 1.3 with 2 inference steps
Set timesteps: tensor([1.5708, 1.3000, 0.0000], device='cuda:0')
  0%|          | 0/2 [00:00<?, ?it/s]
100%|██████████| 2/2 [00:00<00:00, 47.77it/s]
Version Details
Version ID
038aee6907b53a5c148780983e39a50ce7cd0747b4e2642e78387f48cf36039a
Version Created
July 23, 2025
Run on Replicate →