Refine an avatar portrait with a clear prompt

“Make it better” leaves almost every detail open. A useful portrait edit identifies both the change and the parts you want to retain.

Personify’s Avatars area lets you start with a description or a saved portrait, then request another image in conversation. Each request produces a new portrait to inspect. Typing does not continuously reshape the image in real time.

Start with a portrait that suits the role

For an invented gardening brand, you might describe a fictional adult presenter in a plain cotton shirt beside a bright kitchen window. Specify the framing and expression if they matter to the video you have in mind.

Keep the first request focused. Asking for a new face, a different camera angle, elaborate clothing and a detailed product interaction all at once makes it harder to decide which part needs another attempt.

Choose a reference you have permission to use. Permission for a real person’s likeness is a separate consideration from whether you can upload an image file.

Name the change and the details to retain

An original follow-up prompt for that fictional presenter could be:

“Make the hair shoulder-length and softly wavy. Keep the face, expression, cotton shirt, camera angle and kitchen-window background as close to the reference as possible.”

Another request might change only the clothing: “Replace the shirt with a plain dark-green overshirt. Keep the face, hairstyle, expression, framing and light consistent.”

These prompts communicate priorities. They do not guarantee that every retained detail will remain identical. Check the result rather than assuming that an unchanged part of the prompt means an unchanged part of the image.

Compare before continuing

For an edit with a previous image, use Before and After to inspect the new portrait against its starting reference. Look at face shape, expression, hair, clothing and lighting separately.

If the new hair works but the face has drifted, decide whether to return to the earlier image before requesting another edit. During the conversation, you can select a saved result as the source for the next request. Check which portrait is selected so the next instruction starts from the image you intended.

Review the quote before another generation. If an image is ready but has not saved, use the save retry rather than requesting a new image to solve a saving problem.

Choose the portrait for the video

Save the portrait you want, then check the avatar selected in the video brief. A portrait is a visual reference; it is not evidence that the presenter has used the product or achieved a customer result.

Finally, review the generated video separately. A still image cannot establish the quality of the voice, movement or lip sync. Selecting a portrait is one creative decision in the finished video, not the final quality check.