Training guide

How to prepare your training photos

The model only knows what you show it. Pick, crop and upload your photos this way and the headshots come back looking like you.

Go to dashboard Back to home

Do
Good light
Daylight or a bright room, no harsh shadows.
Do
Plain background
A wall or a quiet scene, not a crowd.
Do
Face and shoulders
Front on, eyes open, a natural expression.
Avoid
Group shots
One person in frame, every time.
Avoid
Sunglasses and hats
Nothing hiding your face or hairline.

01 / Three steps

How to prepare your images

Pick, edit and upload your training photos. Ten minutes here makes the difference between a passable set and a good one.

Pick your photos

Choose high-resolution images from your camera roll or desktop. If using iCloud or Google Photos, download the highest quality version first. Put them in one folder and duplicate the set so you can edit copies and keep originals.

Edit your photos

Crop each image to face and shoulders (e.g. with Preview on Mac: Selection tool then Crop). Name them 1 to 10 so you can keep track. If files are too big, compress with the Pacing Compressor or reduce size; if too small, consider AI upscaling (e.g. Upscayl) and check the result.

Upload

Upload 4 to 10 photos, each under 3 MB, in JPG or PNG. Head to your dashboard, start a new model, and add your prepared images. Our AI will train on them and deliver your headshots.

Recommended steps

Find high-resolution image files directly (camera roll, downloads, desktop, etc.).
If using iCloud or Google Photos, download the highest quality version (e.g. “Original” or “Full resolution”) before editing.
Add the chosen photos to a single folder, then duplicate the images and make your edits on the copies so you keep originals.
Crop and resize as needed (see below). Name them 1–10 (e.g. 1.jpg, 2.jpg) so you can keep track.

What kind of photos to upload

Use 4 to 10 clear portrait photos of just you, mostly head and shoulders, with a couple from the waist up if you have them.
Aim for variation: different outfits, poses, days, backgrounds, and lighting. Every image should add new information about you.
Avoid multiple photos from the exact same moment (same outfit, background and lighting), otherwise the model may over-learn that look.
Always choose photos where you are alone in the frame and not making exaggerated or “funny” faces.

Selection and crop with Preview (Mac)

Open the image in Preview.
Use the Selection tool (rectangular marquee) to select the area you want (e.g. face and shoulders).
Crop: Tools → Crop, or ⌘K.
To get a square (1:1): hold Shift while dragging the selection, or use Tools → Adjust Size and set the same width and height.
Export or Save. Prefer “Save as” to keep the original; use JPEG for smaller file size.

Cropping, resizing, and spotting bad quality

You can crop high-resolution images without losing enough detail for the AI. You cannot fix a photo that is already low-resolution by cropping. You’ll just get a smaller, still low-res image.

How to spot a bad-quality photo: Very low resolution (e.g. under 512×512), very small file size (e.g. under 100 KB), or visibly blurry/compressed. If you only have small or low-res photos, you can try upscaling (e.g. with Upscayl) or see the optional section below and then check the result before uploading.

Optional: upscale small photos with AI

If your photos are too small (e.g. under 512px or very low file size), you can try upscaling them with an AI tool before uploading. Upscayl is a free, open-source upscaler for desktop (Linux, macOS, Windows). Always review the upscaled image. Results can sometimes look over-smoothed or show artefacts. Use only when you don’t have a higher-resolution source.

If images are too big

Each image must be under 3 MB. If your files are larger, either reduce their dimensions (e.g. in Preview: Tools → Adjust Size) or compress them with our own tool.

Pacing tool

Pacing Compressor

Drop your photos in, pick a size, download the smaller files. It runs in your browser, so nothing is uploaded anywhere. Reads and writes JPEG, PNG, WebP and AVIF.

Open Pacing Compressor

For training: keep each photo under 3 MB and save it as JPEG or PNG, then upload those files.

File size limits

Per file: 3 MB maximum per image (recommended by our AI provider for best training).
Total: No combined limit; you upload 4–10 photos and each one must be under 3 MB.
Supported formats: JPG, PNG.

Example training images

These photos are examples of the kind of images that work well for training: clear, well-lit headshots that tend to produce better results.

Training photo FAQ

Each image must be under 3 MB. There is no total combined limit; you upload 4 to 10 photos and each one must be under 3 MB. This is the limit recommended by our AI provider for best training results. If a photo is over 3 MB, reduce its size or compress it (the Pacing Compressor runs in your browser) before uploading.

Yes, but prefer high-resolution originals. If you use iCloud or Google Photos, download the full-quality version first, then crop. Cropping a high-resolution image is fine; cropping a low-resolution one loses detail and can hurt AI training. On Mac, Preview is ideal for selection and crop; on Windows, use Photos or another editor that lets you export at full quality.

Check resolution (e.g. at least 512x512, ideally 1024x1024 or higher) and file size. Very small files (e.g. under 100 KB) or very low pixel dimensions usually mean the image is too low-res for good training. Blurry or heavily compressed images also tend to produce worse results. When in doubt, use a higher-resolution source or consider upscaling (e.g. Upscayl) and then check the result before uploading.

Optional. Tools like Upscayl can upscale low-resolution images with AI. If you do this, always review the upscaled result before uploading; sometimes upscaling can introduce artefacts. Use it only when the original is genuinely too small (e.g. under 512px) and you have no higher-res source.

Naming them 1.jpg, 2.jpg, etc. (or example-1.jpg, example-2.jpg) helps you keep track of which photos you have selected and in what order. It also makes it easy to replace or add images in your folder without confusion. The app does not require specific names; this is just a workflow tip.

Either reduce dimensions in an image editor (e.g. Preview on Mac: Tools then Adjust Size) or compress them. The Pacing Compressor is our own tool: it runs in your browser, nothing is uploaded, and it keeps the quality you need for training. Keep each image under 3 MB and ideally at least 512x512 (or 1024x1024) for good training.

No. The training guide is a public page and anyone can view it, logged in or not.

Ready to upload? Head to your dashboard and start a new model.

Go to dashboard

Ready to elevate your professional image?

Join professionals using our AI headshot tool. Need more help with your marketing? We're a full-service agency ready to support your growth.

What makes a good training photo