From a difficult question to a confident performance
The scene needs its opening hesitation. The interviewee looks toward the phone or question, apparently considering an answer, while the microphone remains part of the interview setup. Then a decisive hand movement changes the relationship: the subject takes the microphone and treats the moment as an opportunity to perform.
The musical response is delivered with a calm face and measured hand gestures. It is funnier when the character seems completely convinced that singing is the appropriate answer. The final calming motions close the sequence without returning to the original question. You do not need to reproduce the lyrics in a caption to make that structure understandable.
Smew follows the short question-to-response arc in a 12-second take. The wide street composition leaves room for the interviewer and subject's gestures. Motion generation and the separate soundtrack step work together to suggest the reference performance, but precise lip timing and every hand shape can vary. Review the full exchange rather than judging whether the opening still looks like your photo.
What the browser workflow creates
This is one adult identity in an existing interview-style composition. Your photo guides the interviewee's face, hair and clothing; it does not cast a second person as the interviewer or replace every bystander in the background. The recipe contains the phone glance, microphone grab and performance gestures, so there is no separate prompt-writing step.
The current catalog offers Seedance 2.5 at 183 credits for a 12-second 720p video in a landscape 16:9 frame. Landscape is intentional: a square or vertical crop may remove the gesture that makes the microphone handover readable. If you later reframe for a social feed, preview that crop before posting.
The original reference soundtrack, including the interview setup and musical response, is added after picture generation. The file is saved with that audio before completion, and Library uses the saved result. This does not clone your real voice, create a new interview recording or grant rights to the song. It is a generated performance using a configured audio reference.
Checking identity, hands and mouth timing
The downward glance can make an identity look less stable than it does facing the camera. A sharp, evenly lit upload with the entire face visible gives the model the best chance to preserve recognizable features through that movement. If a reference-associated moustache or clothing color appears, compare your source image: a tight selfie with little outfit information leaves more to invention.
The microphone grab is another useful checkpoint. The hands should read as one person's deliberate action rather than a blurred exchange with duplicated fingers. Watch the moment the mic reaches the mouth and the final calming gestures too. Small motion artifacts can be more noticeable here because the frame centers on conversation rather than spectacle.
The picture is generated and the soundtrack is placed afterward, so exact mouth alignment is not guaranteed. A different take may vary, but another paid run is not a certain fix for every mismatch. Start with better source information when possible, then decide whether the result works as a humorous AI reaction before you publish it.
Use the reaction without misrepresenting the person
The strongest captions supply a new situation: being asked for an update when you have none, facing an unexpected quiz, or trying to answer a complicated meeting question. Those contexts explain why the character suddenly changes the subject. Write the setup in your own words rather than copying a lyric block or presenting the clip as a real interview.
Credit YJ when discussing the song and original performance. The finished file's soundtrack remains subject to the relevant music rights, and platform rules differ for personal posts, business accounts and advertisements. Access to a template is not a universal publishing license. The creator links above are a better route to the source than a transcription of the lyrics.
Use the available AI disclosure, and post only a consenting adult's likeness. If the joke is about a friend, check that they are comfortable being shown singing and handling the microphone. The result is intended as a playful fictional reaction, not evidence of something that person said, sang or was unable to answer.
