You upload one to five photos. The first person sings lead, the second picks up sax, the rest go to backing vocals, and any empty slots get filled with AI musicians. About a minute later you get a 15-second, 9:16, 720p vertical video with the live audio, plus a no-music MP4 and a GIF.

The site is https://tinydeskai.studio if you want to see it. No signup, $7.99 a video.

Three things I got wrong

Face consistency across a wide, moving stage shot cannot be done by swapping faces afterwards. I spent about two weeks on that approach before giving up on it. What actually works is generating a reference portrait per person first, then using those portraits to constrain the video generation. Trying to fix a bad face after the fact never lands.

Audio sync matters more than resolution. I kept pushing for sharpness and it made no difference to how convincing the result felt. A vocal that drifts 200 milliseconds off the beat is what gives it away.

A no-music export turned out to be essential rather than a nice extra. TikTok and Reels mute licensed audio, so without it the clip is silent on the platforms where people actually share things.

How it is put together

Two stages. The still image with the swapped-in people comes from Nano Banana 2, and the motion comes from Seedance 2.0. The app itself is Next.js.

Skipping the account step was deliberate. An earlier version had a registration wall and conversion fell off a cliff, so now it is guest checkout all the way: pay, and the file comes back.

The case I am least confident about is group photos with three or more people. If you try one and it breaks, I would genuinely like to know how.