Two-Character Model Training (Beta)

Our beta two-character pipeline lets you build a shared model for two people—perfect for co-founder shoots, couples, or dynamic marketing duos.

Step 1: Gather Balanced Photos

Collect 20–25 photos for each subject. Keep backgrounds simple and avoid group shots so PhotoGPT can isolate each person. Upload images into their respective slots (Person A and Person B).

Step 2: Define Identifiers

Give each subject a memorable identifier (brandx ceo, brandx cto). During inference you'll reference both identifiers inside one prompt.

Step 3: Train the Model

Kick off training from the dashboard. The beta pipeline alternates between both datasets so the final checkpoint understands their unique features and proportions.

Step 4: Prompt the Duo

Use both identifiers in a single prompt:

brandx ceo and brandx cto, studio lighting, modern tech office

Add pose or wardrobe instructions to choreograph interactions. For example, brandx ceo holding a tablet while brandx cto reviews a laptop.

Step 5: Iterate & Share

Once you're happy with the results, save presets so your team can reuse the composition. Mark favorite renders and export them in social or print-ready formats.

Limitations During Beta

• Stick to two people per model.
• Avoid heavy occlusion or props that hide facial features.
• Expect slightly longer training times compared to single-subject DreamBooth.

We're actively improving the pipeline—send feedback through the dashboard to help shape the next release!