Our price is simple – $0.04 per gigabyte, whether that’s images or videos.
You can calculate our charges by simply multiplying the number of gigabytes by $0.04.
AWS charges will also apply, and will vary depending on what infrastructure you use. Here are some approximate values based on some plausible assumptions*:
| Files | Model size | Time to process | Our charges | AWS charges | Total charges |
|---|---|---|---|---|---|
| 1000 JPG images1 | Nano | 0.1 sec | $0.04 | $0.01 | $0.05 |
| 1000 large JPG images2 | M | 0.2 sec | $0.55 | $0.01 | $0.56 |
| 1000 RAW images3 | X | 0.3 sec | $5.44 | $0.01 | $5.45 |
| 1 hour 4k/60 video4 | X | 67 sec | $0.40 | $0.61 | $1.01 |
| 1 hours HD video5 | M | 34 sec | $0.05 | $0.30 | $0.35 |
How so fast and so cheap?
Combining the efficiency of YoloX and the power of a cluster of 8 Nvidia A100 GPUs in a p4d.24xlarge instance on AWS gives the best price and performance on the market by far – the main cloud vision APIs charge around 20x more, and are slower with lower accuracy.
Why so cheap?
Sentigral is designed for scale.
When we began working on the product, we believed at least an order of magnitude of cost reduction was necessary for developers to routinely include Computer Vision in all the apps that should include it.
If computer vision, and eventually other forms of AI, can be in the hands of lots of developers, it gives more choice to consumers, and in turn more ability for them to control how AI operates in their lives.
*AWS charges are estimates, and are subject to change. Based on a p4d.24xlarge instance – others are available.
- 1000, 1MB compressed JPG files, 1GB total
- 1000, 13.6MB JPG files, 45MP, 100%. 13.6GB total
- 1000, 135MB 45 Megapixel RAW files. 135GB total
- 4K 60FPS H.264 MP4. Total 216,000 frames, 10GB per hour.
- 1080p 30FPS H.264 MP4. Total 108,000 frames, 1.2GB per hour.
*AWS charges are estimates, and are subject to change. Based on a p4d.24xlarge instance – others are available.