AI Image Size Calculator
Find the best pixel size for SDXL, SD 1.5, FLUX, DALL·E, gpt-image-1 or Midjourney at any aspect ratio, and copy the exact parameter. Free.
- No uploads
- No limits
- No signup
Closest size this model allows: 1.6% narrower than the shape you asked for.
Source: SDXL paper, Appendix I: multi-aspect training resolutions
Same shape on every model
| Model | Size (pixels) | Parameter |
|---|---|---|
| Stable Diffusion XL | 1344 × 768 | "width": 1344, "height": 768 |
| Stable Diffusion 1.5 | 680 × 384 | "width": 680, "height": 384 |
| FLUX.2 (BFL API) | 1360 × 768 | "width": 1360, "height": 768 |
| FLUX1.1 [pro] (BFL API) | 1376 × 768 | "width": 1376, "height": 768 |
| OpenAI gpt-image-1 | 1536 × 1024 | "size": "1536x1024" |
| DALL·E 3 | 1792 × 1024 | "size": "1792x1024" |
| DALL·E 2 | 1024 × 1024 | "size": "1024x1024" |
| Midjourney | — | --ar 16:9 |
- Input:
- Aspect ratio
- Output:
- Size in pixels
- Runs on:
- Your device
- Size limit:
- None
- Account:
- Not needed
How to use AI Image Size Calculator
Pick the shape
Choose an aspect ratio such as 16:9 or 2:3, or type your own.
Pick the model
Select SDXL, SD 1.5, FLUX, gpt-image-1, DALL·E or Midjourney.
Copy the setting
Use the pixel size shown, or copy the ready-made parameter into your tool or API call.
Why AI Image Size Calculator on Toolshed
No uploads
Everything runs in this browser tab. Your files never leave your device.
No limits
No file-size caps, no daily quotas. Process as much as your device can handle.
No signup
No account, no email, no watermark.
Frequently asked questions
Why does the size matter for AI image models?
Models work best near the resolution they were trained at and only accept certain sizes: SDXL was trained on fixed buckets around one megapixel, FLUX needs multiples of 16 or 32, and OpenAI's API accepts only a few sizes. Asking for something else gives errors, stretching or repeated subjects.
What is the best SDXL resolution for 16:9?
1344 × 768, the closest of SDXL's training buckets. For portrait 9:16 use 768 × 1344; for 2:3 use 832 × 1216.
Why is there no pixel size for Midjourney?
Midjourney takes only the ratio, as --ar with whole numbers (1.85:1 becomes --ar 37:20), and chooses the pixels itself. It doesn't publish a size table, so we don't guess one.
Where do these sizes come from?
From each model's own documentation, linked under the result: the SDXL paper, the Stable Diffusion 1.5 model card, Black Forest Labs' API docs, OpenAI's API reference and Midjourney's docs. Everything is calculated in your browser.
Are my files uploaded anywhere?
No. AI Image Size Calculator runs entirely in your browser. Your files are processed on your device and are never sent to a server.
Is there a file size limit?
No fixed limit. The only limit is your device's memory. Very large files work best in a desktop browser.