iXzenoSQ: Tips to maximize our own custom AI models?
As I don't want to waste 1500 credits to "test" different trained models, do you have any other tips besides the two below mentioned in the app?
1. Close-up shot with clear facial features
2. Clean background with no distractions
I presume it'd be best to train the AI with *variety* for it to generate the most accurate representation of the model.
So, is it better to upload photos of ourselves with different facial and/or upper-body angles? Sort of like the Character Design Sheets that often help with AI image generations.
How about depth? Should we upload photos of varying depth and sizes, e.g. facial close-ups to below-the-waist body shots?
Or perhaps different hairstyles or accessories (glasses, no-glasses, etc.) for varying looks?
If you could provide a list of more tips to maximize the AI training, that would be appreciated!
Creative_Bot_PhotoGPTAI
May 25, 2025A: Absolutely, great observations and questions! You're spot on about the importance of variety and clarity in your input images. Here are a few additional tips to help you get the most out of your model training:
✅ Best Practices for Model Training
- Diverse Angles
Include a mix of frontal, profile, and ¾ angles. This helps the model understand your facial structure more comprehensively.
- Varying Backgrounds & Lighting
While clean backgrounds are ideal, it’s helpful to mix in a few natural settings with controlled lighting for variety — just avoid cluttered or noisy environments.
- Expression & Pose Variety
Use images with neutral, smiling, and slightly varied expressions. Also include different head tilts and upper body poses to improve adaptability.
- Consistent Identity
All images should be of the same person. Including someone else (even in the background) can lead to degraded results.
- Hair & Accessory Options
If you regularly wear glasses, hats, or other accessories, include both versions — with and without them.
- Clothing Diversity
Having different styles, necklines, and textures in your clothing helps improve prompt responsiveness and realism in outfit rendering.
📚 We also recommend checking out our full guide for image selection here:
👉 https://www.photogptai.com/guides/selecting-the-ideal-input-images
And if you're exploring prompt techniques:
👉 https://www.photogptai.com/guides/how-to-write-prompts
If you're ever in doubt or want feedback before training, feel free to drop us a message or hop into our Discord:
👉 https://discord.gg/rENRJw439H
We’re here to help you get the best possible results!