What this flow does
Upscale + Face Swap combines two finishing steps into one:
Upscale upgrades clarity, detail, and overall fidelity.
Face Swap replaces faces in every target image with the same source face, ensuring consistent identity across the entire batch.
You can use it when you already know which face your Persona should have and you want the fastest route to final, consistent images.
When to use it (and when not)
Use this flow when:
You’ve generated a set (by Prompt, Reference, PhotoShoot, or uploads) and you’re ready to lock identity.
You want fewer clicks and a single render per image instead of separate Upscale and Face Swap steps.
You need batch consistency across dozens or hundreds of images.
Consider other options when:
You still plan to do heavy edits (dress-up, background replacements). Do those first, then run this flow last.
The face must remain exactly as in the original photo (no swap needed). Use Upscaler — Safe Face modes instead.
Modes (choose your Upscale engine)
You can select which Upscale engine runs inside the flow:
Basic: fast and budget-friendly; great for large batches and social-ready quality.
Premium: higher realism and micro-detail; slower and more expensive than Basic.
For the most stable identity preservation, use Basic Safe Face or Premium Safe Face modes. These modes are optimized to protect facial identity during enhancement. Choose between them based on your quality and speed needs.
How to use (step-by-step)
Open “Face Swap + Upscale”.
Upload Face Image. Supports a single image or selection from Personas. Use a sharp front or slight 3/4 portrait with minimal occlusion.
Upload Images (targets). You can batch up to 1000 images.
Version → choose Basic or Premium.
Start the flow. Track progress in the task list; partial failures can be retried.
Review results → Download or send winners to other tools (e.g., Carousel, Video).
The cost per image is shown on the button before you launch.
Source face guidelines (for best swaps)
Lighting & sharpness: neutral light, visible eyes, no heavy shadows.
Pose: frontal or gentle 3/4. Avoid extreme angles or covered faces.
Expression: neutral to slight smile; consistent across your Persona library.
One identity per batch: if you need multiple Personas, run separate jobs.
Target image tips
The face should be visible (not fully blocked by hands/hair).
Avoid tiny faces in the frame; crop a little tighter if the subject is very small.
Group photos: the tool swaps faces it can detect; review results and re-run on missed frames if needed.
Best practices
Run this flow last in your pipeline. Generate → curate → Upscale + Face Swap → publish.
If you also need strict face preservation without swapping, use Upscaler Safe Face modes instead.
Keep aspect ratio and color grade consistent across a campaign to strengthen visual identity.
For extremely soft, tiny originals, consider generating higher-quality bases first; upscaling cannot invent lost composition.
FAQ
What’s the difference between Basic and Premium here?
Only the Upscale engine changes: Basic is faster/cheaper; Premium adds more realism and micro-detail.
Will the face always match my source?
Yes—that’s the point of the flow. Quality still depends on source clarity and how visible the face is in the targets.
Can I run this on images from any tool?
Absolutely. Send results from Face Gen, Generation by Prompt/Reference, PhotoShoot, or uploads.
How many images can I process?
Up to 1000 targets per job.
What about ethics and rights?
Only swap faces with consent and follow local laws/platform rules. Don’t impersonate or mislead.
Troubleshooting
Face looks off / uncanny → Use a sharper, better-lit source; avoid sunglasses; try a frontal portrait.
Skin tones mismatch → Premium upscale tends to harmonize lighting/tones better. Also try targets with similar lighting to the source.
Artifacts after upscaling → Switch from Basic to Premium; if you need minimal changes elsewhere, go to Upscale tool and use Natural or Ultra Clarity modes after the face swap.
Why this flow is a great default
It removes two manual hand-offs, locks identity with one consistent face, and ships polished images in a single render. For most teams, it’s the fastest path from “good generation” to publish-ready content.