Tiny Desk AI Prompts
Three starting points for three different jobs. Edit the wording to match your actual files before you generate.
Make your concert prompt
FREEChoose your approach. Copy the prompt into your video tool.
Recast the lead singer in an existing clip
Use with a tool that accepts a source video plus a reference image. Keep the reference order consistent: Video 1 is the performance; Image 1 is the new performer. If your tool labels inputs differently, update the wording.
Edit the supplied performance video. Replace only the lead singer closest to the camera with the consenting adult in Reference Image 1. Match their face, hair and natural proportions. Keep the original clothing, body position, gestures and expression timing as closely as possible. Preserve the other performers, instruments, microphones, lighting, background and camera movement. Keep the source framing; do not add cuts, people, text or logos. Keep hands and microphones distinct. Handle music separately using audio you have permission to publish.
Create a new desk-concert scene
Use with an image-reference video model. This asks for a new performance; it does not reproduce the original Outstanding camera shots or choreography.
Create a short, newly generated desk-concert scene in a warm record shop, in landscape 16:9 framing. Feature the consenting adult in Reference Image 1 as the lead singer. Use their face and hair as the visual reference. Add fictional adult backing musicians. Use subtle rhythmic gestures, warm window light and a steady camera. Keep faces, hands, instruments and microphones distinct. No existing logos or captions. Make the video silent and add permitted audio separately. This is a new performance, not a reconstruction of an existing recording.
Keep the next attempt focused
Describe one visible issue instead of asking the model to “make it better.” This example addresses a changed camera angle. It is a revision instruction for a tool that supports editing or reference video, not a guarantee that the model will preserve all details.
Keep the source camera angle and framing throughout the clip. Do not zoom out or add a cut. The lead performer should remain at the same scale and position as in the source. Preserve the other performers and background. Apply only the requested identity change using Reference Image 1. Pay particular attention to hands crossing the microphone.
Make the prompt fit your scene
- Replace “lead singer closest to the camera” if your target stands elsewhere.
- Say whether you want to preserve clothing or request a different outfit.
- Start with one identity before attempting a larger cast.
- Keep sound separate unless your chosen model explicitly supports your audio workflow.
These templates express an intended result. Identity, movement and framing can still drift. Check the full output and your provider’s current capabilities and pricing before running another take.