Two-Photo Mode Guide: Photo Setup, Troubleshooting, and Best Styles
Updated · Published · 7 min read · Let's Mack editorial team
Short answer
Short answer: Two-Photo Mode works best with one clear face per image, compatible head angles, and enough head-and-shoulder context for generated motion. If a result looks mismatched, change the source crop or portrait before spending retries on more styles.
Best for
- troubleshooting mismatched two-photo results
- choosing stronger source portraits
- deciding when to crop or replace an input
- selecting a diagnostic first style
Not best for
- the basic transactional workflow
- guaranteeing exact identity preservation
- misleading or non-consensual public posts
Short answer
Use the Make Two Photos Kiss with AI page for the basic creation flow. This guide is for the next question: what should you change when two separate portraits do not compose cleanly into one generated scene?
Confirm that Two-Photo Mode is the right input
Use Two-Photo Mode when each intended person is in a different image. If both people are already visible together in one clean photo, one-photo mode avoids an unnecessary composition step and is usually the simpler starting point.
For a broader one-photo input checklist, use the photo-to-kiss-video troubleshooting guide.
Check each portrait separately
Before comparing the pair, inspect each source image on its own:
- Face visibility: the full face should be visible without hands, sunglasses, masks, text, or heavy shadows covering key features.
- Sharpness: replace motion-blurred, compressed, or heavily filtered screenshots when a clearer portrait is available.
- Crop: include the head, neck, and some shoulder context instead of using an extremely tight face-only crop.
- Angle: front-facing and slight three-quarter portraits are easier starting points than extreme profiles.
- One main person: crop out extra people so the intended input is unambiguous.
Compare the pair for compatibility
The photos do not need identical lighting or camera equipment, but large mismatches can affect the generated composition. Compare:
- face size within each frame
- head angle and camera height
- brightness and shadow direction
- how tightly each portrait is cropped
- whether one source is heavily filtered and the other is natural
When one portrait is a close selfie and the other is a distant full-body image, recrop or replace the distant image before generating.
Use styles as a diagnostic ladder
- Start with Classic. It is the clearest neutral test of whether the source pair composes coherently.
- If Classic works, choose the final mood. Movie Kiss and Rain Kiss add cinematic motion; Caught in 4K and NPC Energy add deliberate comedy.
- If Classic fails, change an input first. Trying many highly stylized modes cannot repair a face that is hidden, tiny, or badly blurred.
Generative output varies even with good inputs, so this process improves the starting conditions rather than guaranteeing a particular result.
Common problems and the next change to try
One person dominates the frame
Recrop the larger face with more surrounding context, or choose a closer portrait for the smaller face so the pair has more compatible scale.
A face looks unclear or changes too much
Replace heavy filters, blur, or obstructions with a sharper source portrait. Exact identity preservation is not guaranteed, but clearer inputs give the model better conditioning.
The pair looks pasted together
Try photos with more similar camera height, head angle, and lighting. Avoid pairing an extreme side profile with a straight-on portrait for the first test.
The result still fails after a better crop
Replace one input at a time. Keeping one portrait fixed makes it easier to identify which source is causing the mismatch.
A practical retry checklist
Before spending another generation, ask:
- Can I see one unobstructed face in each image?
- Are both faces reasonably similar in scale?
- Do both crops include enough head-and-shoulder context?
- Have I tested the pair in Classic first?
- Am I changing one variable at a time?
Privacy and consent basics
Uploads and generations are excluded from public listings by default. Hosted media may use shareable CDN URLs, so anyone with a valid URL may be able to access it. For signed-in users, Let's Mack stores a private source-image copy for photo history, retries, and provider fallback, and sends the inputs to a configured generation provider. See the Privacy Policy for retention and deletion details.
Keep sensitive macks private, avoid harassment, and do not publicly post a person in an embarrassing or intimate context without permission. Label public posts as AI-generated when the context could be ambiguous, and review Trust & Safety before sharing.
CTA
When both portraits pass the checklist, return to Two-Photo Mode and run a neutral Classic test.
Frequently Asked Questions
Why does a Two-Photo Mode result look mismatched?
Large differences in head angle, face size, lighting, blur, or cropping can make the generated composition less coherent. Try clearer portraits with more compatible framing.
Should I retry the style or replace a source photo?
Replace or recrop the source when a face is blocked, tiny, blurry, or at an extreme angle. Try another style only after both inputs are clear enough for a neutral Classic test.
What is the best first style for troubleshooting?
Classic is the most useful diagnostic first style because it provides a neutral baseline before more cinematic or comedic motion is introduced.
Can I keep a two-photo mack private?
Yes. Uploads and generations are excluded from public listings by default. Hosted media may use shareable CDN URLs, so anyone with a valid URL may be able to access it.