Firefly vibe code not reacting.
If I put in details for image development, this has no bearing on the output - in fact, the images are identical to the reference image, suggesting the prompts are not being read.
If I put in details for image development, this has no bearing on the output - in fact, the images are identical to the reference image, suggesting the prompts are not being read.
Hi
Thanks for your feedback.
We tested this thoroughly and were able to reproduce the same issue.
Across different prompts, workflows (using Edit tab: prompt & Markup + AI assistant), and models, the results remained largely unchanged, with only minor improvements. This confirms that the problem isn’t due to prompt quality or the way the request was written.
The root cause is a current limitation of generative AI: these models don’t reliably enforce precise, synchronized pose changes across multiple people. They tend to preserve the original image structure and regenerate each person independently, which makes consistent arm and hand alignment difficult to control.

In short, this type of edit isn’t fully achievable with standard prompting or AI editing tools today.
For exact pose matching, the most reliable approach is use manual targeted edits per person using Photoshop.
Thank you for your understanding.
Hajar
Already have an account? Login
Enter your E-mail address. We'll send you an e-mail with instructions to reset your password.