IFPI AI chart rules: scoring a music video with AI audio

Can an AI-made song chart? IFPI's announcement (read 2026-10-04) says a recording developed with generative AI qualifies for official charts only if it meets three tests: the AI service used is authorised and lawful, the track is substantially human made, and it raises no stream or chart manipulation concerns. The principles were announced on July 30, 2026 and, per IFPI, apply first to official charts it runs directly across Latin America, the Middle East, Africa and Southeast Asia.

If you make music videos, this matters less for the picture and more for the audio. A video whose only soundtrack is one unedited AI generation is the case the principles are aimed at. A video that carries a human-made song, with AI used somewhere around it, is not. Sume's side is the tooling: generate, cut and lay audio, and caption.

What the three tests ask of a production

Read the principles on IFPI's page and the MBW report. Chart eligibility belongs to the chart body, so nothing Sume returns can certify a track either way.

Where Sume helps with the record

The Music Router echoes the requested id in job.model (sume/music-auto stays sume/music-auto) and names the engine that ran in job.request.routed_model, for example lyria-3.5. Saving that field next to each generated cue is a cheap way to answer where a sound came from later.

That is provenance of the generation, not a rights opinion. The docs do not say anything about charts or licensing, and this post does not either.

A production pattern that stays human-led

Use AI for what the principles most plausibly leave room for, the creative process around a human-made song: a scratch bed to cut picture against, a temp score for a pitch, a reference for a composer. Then let the final audio be the human record. In Timeline 1.0 the audio spine and the optional soundtrack bed are separate inputs, so swapping the temp bed for the finished audio is one URL change on the render request.

curl https://api.sume.com/v1/music-router/models \
  -H "Authorization: Bearer $SUME_API_KEY"

Short answer for your release plan

If a track might go to charts, ask the distributor and the chart body in writing before you spend on the video. If the audio only lives inside the video, none of this changes your render.