Skip to main content
Participant
July 8, 2026
Answered

The output video did not use the faces of the people in the photo but substituted people unknown. This made the output unusable

  • July 8, 2026
  • 1 reply
  • 14 views

I have already explained the problem. The video made from a photo did not use the faces of the people in the phot making the output unusable 

    Correct answer Hajar Toutsi

    Hi Computerhero3,

    Thanks for letting us know what’s happening on your side.

    To help explain why this happened: Firefly is a generative AI tool, which means it is primarily designed to interpret and reimagine images as a whole rather than make precise, targeted adjustments automatically. Because of how it reimagines the scene, it can sometimes replace details like faces unless it is given strict instructions.

    The good news is that the more precise and detailed you are in your instructions, the closer the output will be to your specific vision. Here are a few strategies to help prevent the AI from altering your subjects' faces in the future:

    Tips for preserving original elements:

    • Be explicit: Do not assume the AI will preserve specific elements automatically. Spell out exactly what should stay unchanged. For example, try including a phrase like: "Keep the exact faces, identities, and all elements of the people in the image intact while adding..."

    • Avoid negative prompting: Firefly does not currently support negative prompting. Instead of telling the tool what not to do (e.g., "Do not change the faces"), focus entirely on telling it what it should do.

    • Check your image quality: Lower resolution photos give the model less detail to work with, which often leads to unintended changes or substituted faces. Ensure you are uploading the highest resolution photo possible.

    • Choose the right model: Some models handle visual effects and constraints better than others, so you may want to experiment with a different model version if the current one continues to struggle with your specific prompt.

    If the output still feels off, you can submit feedback directly by clicking the thumbs‑down icon when hovering over the result. That feedback goes straight to the team to help improve Firefly’s behavior.

    If you’d like, share the exact prompt and reference image you used, and we can help refine it together.

    Hajar
    [Moved to questions]

    1 reply

    Hajar Toutsi
    Community Manager
    Hajar ToutsiCommunity ManagerCorrect answer
    Community Manager
    July 8, 2026

    Hi Computerhero3,

    Thanks for letting us know what’s happening on your side.

    To help explain why this happened: Firefly is a generative AI tool, which means it is primarily designed to interpret and reimagine images as a whole rather than make precise, targeted adjustments automatically. Because of how it reimagines the scene, it can sometimes replace details like faces unless it is given strict instructions.

    The good news is that the more precise and detailed you are in your instructions, the closer the output will be to your specific vision. Here are a few strategies to help prevent the AI from altering your subjects' faces in the future:

    Tips for preserving original elements:

    • Be explicit: Do not assume the AI will preserve specific elements automatically. Spell out exactly what should stay unchanged. For example, try including a phrase like: "Keep the exact faces, identities, and all elements of the people in the image intact while adding..."

    • Avoid negative prompting: Firefly does not currently support negative prompting. Instead of telling the tool what not to do (e.g., "Do not change the faces"), focus entirely on telling it what it should do.

    • Check your image quality: Lower resolution photos give the model less detail to work with, which often leads to unintended changes or substituted faces. Ensure you are uploading the highest resolution photo possible.

    • Choose the right model: Some models handle visual effects and constraints better than others, so you may want to experiment with a different model version if the current one continues to struggle with your specific prompt.

    If the output still feels off, you can submit feedback directly by clicking the thumbs‑down icon when hovering over the result. That feedback goes straight to the team to help improve Firefly’s behavior.

    If you’d like, share the exact prompt and reference image you used, and we can help refine it together.

    Hajar
    [Moved to questions]