ElevenLabs Music free tier: lossless downloads, 5 a day, and Sume

ElevenLabs lists lossless music downloads on its free tier, capped at 5 a day, according to a September 2026 changelog roundup. Sume's docs describe no per-day download cap: each generation is a paid job that returns a hosted audio file, which its docs describe as typically audio/mpeg. If you need a lossless master for a broadcast or a mastering chain, the file Sume returns is not lossless; if you need a track under a social video, it usually is enough.

The sections below separate the two cases so you do not pay for a format you will compress anyway.

What does the free tier actually include?

The roundup reports that Music v2.5 became the default on 2026-09-11, that API support followed on 2026-09-14, and that the free tier includes lossless downloads, five per day. It does not say whether free-tier tracks may be used commercially, so read the vendor's terms before using one in an ad; this page does not claim either way.

When is an MP3-type file enough?

For video, yes in most cases. A track that goes under a voiceover is mixed and then encoded again by the video export and by the platform, so a lossless source adds little. Where it matters is archival: a master you plan to remaster, stem-split or sell as audio. For that, ask whether the vendor can deliver a lossless or stem file, and keep the original download.

How do I get a Sume track into a video?

Submit the brief to the Music Router, wait for the job, and read the audio artifact from result.artifacts[]. Then join or cut it with Timeline audio, or lay it under a render as a soundtrack bed. Sume's own docs mark Music 1.0 as retiring gradually in favour of the router, so new code should call the router; it resolves to Lyria 3.5 today.

The Music Router page lists the request fields, and Timeline audio covers joining and slicing the result at a flat $0.01 per job.

What should I do?

Decide the deliverable first. If it is a lossless master, use a vendor that offers one and keep the download. If it is a video bed, generate on Sume, store the job id and artifact URL, and move on. Either way, record the model and date next to each track.