Which AI Image Model Ages Characters Most Realistically? Progressive Aging Test

In my previous post, I tested how well different AI image models could age a character using only a single reference image (what I call the base method ). This time, I wanted to try a different approach. Instead of always referencing the original image, what if I aged the character progressively? In other words, each newly generated image becomes the reference for the next age group. Would this produce more realistic aging while preserving the character's identity? Let's find out. Reference Image This is my reference image of an 18-year-old woman. Image Models Tested 1. Grok Imagine Quality 2. Wan 2.7 Pro 3. Seedream 5 4. GPT Image 2.0 Prompt For every model, I used the same prompt: Half-body frontal portrait of a (25, 35, 45, 55)-year-old woman wearing a short black dress, medium-length hair, looking directly at the camera, white background. Note: I initially used a black spaghetti yoga top with black tight shorts, but Seedream 5 produced less convincing aging results. My guess is that sportswear tends to make the character appear younger, especially for the 45- and 55-year-old versions. Switching to a simple black dress produced noticeably better results. Method Let's call this the progression method . Starting with the original 18-year-old image, I generated the 25-year-old version. I then used the 25-year-old image as the reference to generate the 35-year-old version, followed by the 45-year-old version, and finally the 55-year-old version. Unlike the base method, each generation builds upon the previous one, allowing age-related features to accumulate more naturally over time. For easier comparison, I've included the original 18-year-old base image alongside every age group so readers can compare how well each model preserves the character's identity while aging her. Age 25 Reference Image (left), Seedream 5 (middle), Wan 2.7 Pro (right) Grok Imagine Quality (left), GPT Image 2.0 (right) Age 35 Reference Image (left), Seedream 5 (middle), Wan 2.7 Pro (right) Grok Imagine Quality (left), GPT Image 2.0 (right) Age 45 Reference Image (left), Seedream 5 (middle), Wan 2.7 Pro (right) Grok Imagine Quality (left), GPT Image 2.0 (right) Age 55 Reference Image (left), Seedream 5 (middle), Wan 2.7 Pro (right) Grok Imagine Quality (left), GPT Image 2.0 (right) Real-World Test Generating portraits is one thing, but how well do these aged characters hold up in an actual scene? To find out, I took each model's 45-year-old version and used it as the reference image in GPT Image 2.0 (my preferred image model) to generate a realistic cinematic scene. The goal was simple: Can the generated character still convincingly look like the same 45-year-old woman? I used the following prompt: Sophisticated Chinese woman standing alone beside a floor-to-ceiling window in a luxurious penthouse at night, city skyline glowing outside, holding her phone while waiting for a message that never comes, elegant evening dress, quiet loneliness, cinematic blue-hour lighting, emotional realism. Using Seedream 5's 45-year-old version Using Wan 2.7 Pro's 45-year-old version Using Grok Imagine Quality's 45-year-old version Using GPT Image 2.0's 45-year-old version Conclusion After comparing all four models, I found that Seedream 5 produced the most convincing aging overall when using the progression method. It consistently added believable age-related facial features while maintaining the character's identity across different age groups. Compared with the base method from my previous test, I prefer the progression method when using Seedream 5, as it produces more natural and convincing aging overall. That said, you may have a different preference, so I encourage you to check out the base method comparison and decide which approach you like better. Interestingly, I found Wan 2.7 Pro performed quite well with the base method. When I used the generated characters in the real-world scene test, the character created from Wan 2.7 Pro no longer resembled its original 45-year-old portrait as closely as the other models. The remaining three models maintained the character's identity reasonably well, although GPT Image 2.0 still made her appear younger than her intended age. My Personal Ranking 1. Seedream 5 2. Grok Imagine Quality 3. Wan 2.7 Pro 4. GPT Image 2.0 GPT Image 2.0 continues to struggle with older characters. Its 45- and 55-year-old generations often appeared closer to someone in their late 30s or early 40s. Grok Imagine Quality performed noticeably better with the progression method than it did in my previous test. While its results can occasionally be inconsistent, regenerating the image often produces a much stronger outcome. Wan 2.7 Pro aged the character reasonably well, but when I later used those images as references inside GPT Image 2.0, the character's identity drifted more noticeably than the other models. One interesting observation was that clothing appears to influence the aging results for some image models. Using simple everyday clothing (such as a plain black dress) produced more convincing results than sportswear. With Seedream 5 in particular, yoga attire consistently made the character appear slightly younger than the intended age. My Recommended Workflow If your goal is to create an older version of a character while maintaining both realism and identity consistency, this is the workflow I recommend: 1. Use Seedream 5 (or Grok Imagine Quality ) to age the character. 2. Use the aged image as the reference in GPT Image 2.0 to refine the clothing, hairstyle, lighting, and overall presentation while preserving the aged appearance. This combination produced the strongest overall results in my testing. I hope you found this comparison useful. If you've experimented with other AI image models for character aging, I'd love to hear about your experience. You may also be interested in my previous comparison, where I tested the base method using a single reference image.