Snapchat now turns your chats into AI songs, and won't say what model made them

Snapchat can now turn a text message into a song. Snap announced Chat to Song on September 21, 2026, one of five AI features going to Lens+ subscribers. You press and hold a message in Chat, hit Create Song, pick a genre, and the app writes a short original track out of whatever you typed.

The example in Snap’s own press asset is “I’ll pick you up at 10 for the movies”. The genres on the picker are Hip Hop, Rock, Country, EDM and Pop.

What Snap does not say anywhere in the announcement is what makes the music.

What Snap disclosed about Chat to Song, and what it left out

The announcement is a product post. It describes the flow, shows the screens, and moves on to the other four features: AI Photoshoot for fixing a badly timed photo, AI Remix for restyling a friend’s Snap, sticker restyling, and AI-enhanced caption fonts.

There is no model name. No partner. No mention of what the song generator was trained on. The only acknowledgement that generative AI is involved at all sits in small grey type at the bottom of the Create Song sheet.

That line reads “This feature uses generative AI”. It is the full extent of the disclosure, and you only see it if you are already inside the sheet with your finger on the button.

Why Snap’s music licensing deals don’t answer the question

Snap is not new to music rights. The company has licensing agreements with UMG, Sony Music, Warner Music, Warner Chappell, Kobalt, BMG, MERLIN, the NMPA and DistroKid, built up since 2020 to power Sounds, the feature that lets you drop a commercial track onto a Snap.

Those deals cover playing existing recordings. Generating new music is a different permission, and it is the one the whole industry has spent two years fighting about in court.

So there are two possibilities here. Either Snap built or licensed a model trained on catalogue it has cleared for this, which would make it one of the quieter licensed music AI launches of the year. Or it is using a general-purpose model whose training data nobody has accounted for, inside one of the biggest messaging apps in the world. Snap has said nothing that separates the two, and Digital Music News reported the same gap.

“

This feature uses generative AI

What Chat to Song means for your own songs

Honestly, nothing immediate. Nobody is releasing a Chat to Song track, and the output is a novelty attached to a message thread.

But the pattern matters, because this is how AI music reaches normal people now. Most of them will meet a music generator as a button inside an app they already pay for, long before they ever open Suno or a DAW. Apple has committed to Made with AI labels, Spotify started licensing fan-made covers, and Deezer now reports more than half its daily uploads as AI. Snap just added a generator to a chat window.

The question I would want answered before that becomes normal is the one Snap skipped: whose music taught it.

Snap says the Lens+ features are live now for paying subscribers. If the company names the model later, or one of its label partners confirms the deal covers generation, that changes the story. Until then this is a music product with no stated rights position, shipping at consumer scale.