Budget a 100-track music library: $0.08, $0.125 or $0.15 a minute
Lyria 3.5 lists $0.08 per song, Sume charges a fixed $0.125 per accepted music job, ElevenLabs music is $0.15 a minute. Library cost at 100 and 1,000 tracks.

A 100-track library costs $8.00 at Google's listed Lyria 3.5 price of $0.08 per song, $12.50 at Sume's fixed $0.125 per accepted music generation, and $15.00 to $45.00 at ElevenLabs' $0.15 per minute if tracks run one to three minutes. At 1,000 tracks, multiply by ten.
The comparison is not like for like: one price is per song, one is per job, one is per minute. The table below fixes the arithmetic, and the last section covers retries, which usually decide the real bill.
The three prices
Vendor figures are list prices as read on 2026-10-03. Sume's figure is from its Music 1.0 docs, which say the price does not vary by prompt length or optional image conditioning.
| Provider | Unit | Price |
|---|---|---|
| Google Lyria 3.5 | Per generated song | $0.08 |
| Sume Music | Per accepted generation | $0.125 |
| ElevenLabs music | Per minute | $0.15 |
Library cost, derived
The ElevenLabs rows depend on track length, which you choose, so three lengths are shown. Lyria 3.5 songs run a couple of minutes by prompt, so one to three minutes brackets the usual case. These are arithmetic on the prices above, not measurements.
| Tracks | Lyria 3.5 at $0.08 | Sume at $0.125 | ElevenLabs, 1 min | ElevenLabs, 2 min | ElevenLabs, 3 min |
|---|---|---|---|---|---|
| 100 | $8.00 | $12.50 | $15.00 | $30.00 | $45.00 |
| 1,000 | $80.00 | $125.00 | $150.00 | $300.00 | $450.00 |
Retries decide the bill
Google's docs describe Lyria 3.5 as single-turn only: there is no step that edits a track you almost like. Each variant is a new generation. If you keep one track in three, a 100-track library means about 300 generations, which is $24.00 at $0.08 and $37.50 at $0.125. Budget by generations, not by finished tracks.
Two cost behaviours are worth knowing on the Sume side. The price is fixed per accepted generation, so a long prompt costs the same as a short one. And a prompt that violates a documented rule is rejected before it becomes a job: non-empty negative_prompt and a duration field both return errors on Music 1.0, so a malformed batch fails fast rather than billing.
A budget formula you can reuse
Keep the formula in a sheet, with your own keep-rate.
- Generations = tracks needed divided by keep-rate. A keep-rate of one in three gives three generations per track.
- Cost = generations times the unit price, for per-song and per-job prices.
- For per-minute pricing, cost = generations times average minutes times the minute price. Set the average from tracks you have already made.
- Add the cost of any joining or trimming steps separately. Sume's timeline audio and audio detach jobs are listed at a flat $0.01 per job in its docs.
Sources
Related posts
More in Pricing
- Caption 25 shorts in 3 languages: 75 jobs at $0.20 is $15
The cost of burning captions into 25 shorts in three languages with Sume's fixed $0.20 caption job, why a restyle costs the same, and what to batch first.
- Caption price for a 61 to 180 second Short: confirm it in the catalog
The Sume docs state $0.20 per caption job for videos up to 60 seconds. They give no rate for 61 to 180 seconds, so read GET /v1/catalog before you budget.
- Captions for 100 ad variants: transcripts at one cent per audio minute
A 15-second ad transcript on Sume is billed per audio minute at $0.01, so 100 variants of 15 seconds plan out at $0.25. Arithmetic inside.
- ChatGPT Pro 500 vs a Sume wallet: what pays for a video job
OpenAI's ChatGPT Pro 500 plan raises ChatGPT usage limits. It does not fund Sume generation. Which bill pays for what when an agent renders video.
Written by Sume