Read the change in shot size
The supplied source frames move from a closer view of the two performers to a wider view that includes the stage floor. Keep both identity checks and framing checks in your review: facial details become smaller in the wide shot, while clothing and body movement become more visible. In the portrait version the complete landscape source is centered on an orange canvas. That preserves the source composition but adds large upper and lower borders. Choose 16:9 if filling the player with the original performance matters more than a tall output file.
Read the Hotel Lobby Verse template as an exchange
Watch a candidate performance with attention to who appears active at each moment. A duo can feel like an exchange even when one person moves more than the other. The useful question is whether you can follow the relationship between the visible people and the sound. That requires watching the whole excerpt rather than selecting a source because its name sounds appropriate.
For your own notes, divide the fifteen seconds into an opening, a middle and an ending. Record whether both faces are visible and whether there is a clear change in emphasis. These are viewing checkpoints, not an instruction to cut the source into three separate shots. The site's template remains a continuous performance.
Do not assign fixed roles before seeing the configured preview. The left slot is not automatically the main vocalist, and the right slot is not automatically a reaction role. A particular passage may distribute attention differently. Confirm what the source actually presents before deciding which photograph belongs where.
Choose references that do not fight the performance
A clear portrait gives a better basis for judging identity than an exaggerated expression chosen to imitate a single cover frame. A very wide smile, deeply tilted head or partly hidden mouth can make it harder to tell which differences come from the reference and which arise during animation. Choose a photograph with readable features first.
You do not need to photograph the subject rapping. The source clip supplies the performance, and the uploaded image is a visual reference. Posing someone with a microphone in front of their mouth can remove facial information without adding a useful instruction. A straightforward portrait is often the more informative input.
If two candidate photographs show the same person clearly, compare the angle and lighting against the preview's demands. A reference with a severe shadow across one side of the face may leave uncertainty when the generated subject turns. This does not establish a universal winning pose; it identifies a variable worth controlling in your first attempt.
Keep the two photographs distinct and single-subject. A collage of several expressions is not a substitute for the expected two separate inputs. Nor does uploading an image of a pet or illustrated character become supported because the selected template is called Verse. The product remains a human-portrait duo workflow.
Look closely when bodies, hands and faces overlap
The shared center of a two-person performance deserves attention. When subjects approach the same area, a hand, shoulder or microphone may partly cover a face. Inspect whether the source leaves enough visible information to follow each person. For your result, check whether identities remain stable after the obstruction clears.
Separate a source obstruction from a generated artifact. If a hand covers a mouth in the original excerpt, the same obstruction appearing in the output is not automatically an error. If facial features change after that hand moves away, that is a different issue. Compare the same moment rather than comparing an input portrait to an unrelated output frame.
Do not use the poster image as a verdict on this behavior. It may show a clean moment while the difficult interaction happens later. If continuous recognition is the main purpose of your clip, review the most challenging visible moment before accepting the whole result. A brief problem can be easy to miss on the first watch.
Evaluate sound and mouths as related but separate checks
Hotel Lobby keeps the source audio. It does not create a voice from either photograph or make the people recite a caption. Choosing Verse changes the selected source performance when configured; it does not unlock custom lyrics, a new verse or a personal spoken message.
First listen for audio continuity without concentrating on the faces. Then replay while watching how the visible mouth motion relates to that same sound. These checks answer different questions. A file can contain the intended soundtrack while its visible articulation still looks wrong. It can also have convincing movement but an export or playback problem.
Use ordinary speed for the main judgment. Slow playback can help locate a mismatch, but viewers will normally see the completed clip at its intended pace. Note whether a problem is sustained, isolated or visible only after repeated frame inspection. That distinction helps you decide whether the clip is suitable for your intended audience.
If the source preview itself has a sound issue, changing the portraits is not a useful fix. Choose another verified source or wait until that preview is corrected. Paying for another output does not repair an audio defect already present in the selected performance.
Check the vertical version around moments of emphasis
A tall frame can concentrate attention on faces, but a performer may move toward an edge at the exact moment you wanted viewers to notice. Watch those moments in the actual 9:16 source version. Ask whether the action and the other person remain in view together.
For the 16:9 version, look at how much of the frame the people occupy on your likely viewing device. More visible width is useful only if the audience can still read the faces and interaction. A wide frame shown very small inside a social post may lose the detail that mattered in your original plan.
Leave text decisions until you know the movement. A line positioned above a shoulder in the opening may overlap a face later. If your caption explains who the people are, put it where it remains readable without obscuring the performance, or keep it in the post description.
Avoid stretching either version to fill a different shape. That changes facial proportions and makes it harder to judge whether a problem came from generation or later editing. Preserve the source ratio when comparing candidate results.
Use a short observation log after generation
Write down three concrete observations: where each person is easiest to recognize, where the visual exchange becomes unclear and whether the sound plays as expected. Include approximate timestamps for problems. This produces something actionable if you decide to change an input or a template.
Suppose one person looks recognizable until turning away and returning. Compare a clearer reference of that same person before changing both portraits. Suppose both faces are readable but the selected excerpt feels wrong for your intended caption. In that case, compare another source preview rather than assuming a higher quality tier will change the scene's timing.
These are hypothetical review situations, not results from a Verse test. No success percentage, measured face score or direct Standard-versus-Pro comparison is available here. Standard costs 300 credits and Pro costs 400; each fresh generation is a separate spending decision.
Choose Verse for the preview you can actually inspect
The best reason to choose the Hotel Lobby Verse template is that its verified performance fits your pair and your planned post. The preset's label alone cannot tell you that. Look for the moment you want viewers to understand, check both frame variants and confirm the face mapping before generating.
If you mainly need a straightforward first comparison, the Classic preparation guide covers baseline setup. If your main concern is how the short clip ends or repeats, the Encore guide covers those decisions. Those editorial lenses do not assert differences in unverified source footage.
Keep your approved portraits ready while a preview is unavailable. Once it is configured, use the matching source to make the final choice. A careful selection takes less effort than trying to explain an unsuitable passage with a longer caption after paying for a render.