What this workflow creates
Correct closed eyes, red-eye, or an unnatural expression in a portrait. One image per request.
- Who it’s for
- Individual users, event photographers
- What it does
- Correct eyes and expressions
- Where to use it
- Personal photos and activity records
One image per request.
What you need
- Single photos of eyes or expressions to be repairedRequired · 1 image
- Preserve the same person’s identity, face shape, age, facial-feature placement, head pose, and lighting.
Provide every required input. Add optional inputs only when useful, and mark conditional inputs as applicable only when they apply to this request.
How to use this workflow
Open the generation form
Select “Use this workflow” to see the required inputs. You’ll need to sign in before generating.
Add the required inputs
Prepare: A single photo of the eyes or expression required to be repaired (1 photo). Complete every required field, add optional inputs only when useful, and mark each conditional input as applicable or not applicable.
Review text and output settings
Choose the aspect ratio, resolution, and quality that fit this use case. The generation form shows the available options.
Generate and review
Select “Generate image,” then review the result and open the full-size image. Check it against the detailed requirements below. One image per request.
Do I need to write a prompt?
No prompt needed—upload the required inputs.
Upload the images listed below; you do not need to add a written prompt. You do not need to write generation instructions. Use the form fields to provide the images, exact copy, data, or description requested by this workflow.
Detailed result requirements
A single photo of the eyes or expression required to be repaired (1 photo)
With your eyes closed, give priority to another photo with the best expression in the same scene; the size/resolution and position of the characters in the two pictures are similar.
No universal public standard is available. Check the requirements of the platform or service you will use.
Set the output canvas from the selected platform placement or production template, not from a reference image. Choose the option that matches this request: Single photos with natural expressions working canvas — 4:5 (recommended). Use a different ratio only when a verified placement or production template requires it. If the platform does not require a fixed aspect ratio, use 4:5 as the recommended working canvas for single photos with natural expressions. Compose the subject with appropriate breathing room. Create one aspect-ratio-specific deliverable per request; generate each additional placement as a separate output. Recompose for the selected ratio while preserving the complete subject, all text, and every required safe area. Never stretch the subject or treat cropping important content as adaptation.
No universal public standard is available. Check the requirements of the platform or service you will use.
Automatically align and mask the expression area; for red eyes, adjust the pupil area and darkening level.
No universal public standard is available. Check the requirements of the platform or service you will use.
No universal public standard is available. Check the requirements of the platform or service you will use.
No universal public standard is available. Check the requirements of the platform or service you will use.
No universal public standard is available. Check the requirements of the platform or service you will use.
Save the layered version and export the flat finished product separately.
Don't generate eyes out of thin air and claim to be real restorations; Smart Portrait may be distorted on inputs such as profile, glasses, occlusion, etc.
Official and platform requirements
No universal public standard is available. Check the requirements of the platform or service you will use.
FAQ
What can I create with this workflow?
Correct closed eyes, red-eye, or an unnatural expression in a portrait. One image per request.
Do I need to write a prompt?
Upload the images listed below; you do not need to add a written prompt. You do not need to write generation instructions. Use the form fields to provide the images, exact copy, data, or description requested by this workflow.
What do I need to upload?
“Single photos of eyes or expressions to be repaired”: 1 image; required. JPG, PNG, and WebP are supported, up to 10 MB per image.
What does one generation produce?
One image per request.
Common search terms
Public evidence and sources
Three directly relevant public sources have been found, and four types of specific constraints have been extracted. The following content only restates information that appears clearly from the source; no filling in of blank fields will be made.
