How We Test Prompts

Generating a photo where you realistically appear in another decade requires balancing two competing forces: facial identity preservation and period atmospheric coherence. Here is how we engineer and test every prompt on MyEraFrame:

Step 1: Identity Anchor Calibration

Each prompt incorporates explicit semantic constraints instructing the generative model to lock facial proportions, skin tones, hairline, and distinguishing facial structures. We test prompts against real portrait reference images to ensure subjects are recognizable.

Step 2: Film Stock and Aesthetic Verification

We test prompts to confirm they faithfully reproduce analogue photographic characteristics: Kodak Kodachrome warm saturation for the 1960s–70s, flash-and-film 35mm grain for the 1980s, and disposable camera flash falloff for the 1990s and Y2K.

Step 3: Multi-Model Evaluation

We test across multiple frontier models including ChatGPT (DALL-E 3) and Gemini to note model compatibility, limitations, and ideal aspect ratios.