ElevenLabs agent hold audio: 180 s, 40 MB; make it with Sume

For ElevenLabs agent hold audio, the file must be MP3 or WAV, up to 40 MB and 180 seconds. With Sume, generate the track with a length written into the prompt (there is no duration field), cut it to range with Timeline audio split if it runs long, and export mp3 or wav.

The upload limits are from the ElevenLabs changelog entry of 2026-09-14, read 2026-10-01. Sume facts are from Music Router and Timeline audio.

What did ElevenLabs add?

The 2026-09-14 changelog adds call queueing for agents at their concurrency limit. Queued callers hear hold audio and receive queue_status events until admitted or timed out. New API methods upload or delete custom hold audio: POST /v1/convai/agents/{agent_id}/hold-audio takes a hold_audio_file as multipart form data.

How do I get a track of the right length from Sume?

Sume music is prompt-driven. The Music Router docs say to steer length in the prompt, for example "a 2-minute track" or time-marked sections, and that duration and duration_seconds are rejected. Ask for an instrumental of about two minutes, then check the actual length of the result.

What if the result is over 180 seconds?

Slice it. Timeline audio split takes ranges[] (1 to 20), each { start, end? }, and returns segments with their own audio_url. Keep the first range that fits, then choose the export format: output.format is wav (default) or mp3 (smaller). A 180-second WAV at 16-bit stereo 44.1 kHz is roughly 30 MB by arithmetic, so mp3 is the safer choice if you are near the 40 MB limit; check the file size either way.

Is it suitable to loop?

A shorter loop-friendly track is also fine; ask for a seamless-sounding bed in the prompt. For music for phone systems in general, see on-hold music for business.