Q: Body and Multiple Models
I get that your platform was primarily built for headshots, but can you say more about how we work on having the body in the images the platform generates match the subject we are using (training on, or model already created)? Like is there any way to have an accurate body shape by training with full body pictures? And/or through a specific prompt or setting? Can you please more about how to do this? Thanks!
If we train more than one model, can we use 2 or 3 in generating an image including those models?
Creative_Bot_PhotoGPTAI
Jun 24, 2025A: Thanks for the great question!
You're right that PhotoGPT AI is primarily optimized for high-quality headshots, but it’s absolutely possible to influence body shape representation in your generations.
If you'd like the model to capture body type more accurately, you’ll need to ensure all training images are full-body shots. This helps the model learn proportions and overall physique better. That said, since the model would then be trying to learn a broader range of features (rather than focusing just on facial details), you may notice a slight dip in facial resemblance compared to when the training is focused solely on headshots.
Regarding multiple people in one image: it's not currently possible to merge two separately trained models into one frame. However, we do offer a beta workflow that lets you train a single model with two people together in all images, this is currently the most reliable way to generate accurate visuals involving more than one person. You can check out the guide here:
👉 https://www.photogptai.com/guides/two-character-model-training-(beta)
Hope this helps! And feel free to reach out on Discord (https://discord.gg/rENRJw439H) if you’d like guidance as you get started.
SC2022PLUSThank you! What if I want the best of both worlds (accurate facial AND body)? Can we train the model with all faces to begin, and then "re-train" by adding body shots? I don't know if re-train means you start all over or are refining the model further (please clarify).
Also, what about training primarily on face and using a prompt to refine the body shape for accuracy- is that an option?