Music 1.0 at $0.125 a generation: budget five takes per final track
Music 1.0 has no seed or duration field, so plan retries. At a fixed $0.125 a generation, five takes cost $0.625 and 12 tracks at five takes cost $7.50.
Why budget more than one take
Music 1.0 bills a fixed $0.125 per generation. The request takes a prompt and an optional image, and it has no seed, temperature, guidance or duration parameter, so two runs of the same prompt are not guaranteed to match. You choose a track by listening, which means buying a few candidates.
The Sume docs say the prompt controls the result and recommend a scene-specific brief with emotion, genre, tempo as a number, key, instruments, an arc and an era.
Take budget table
The price is a fixed fee per generation, so cost is takes x $0.125.
| Final tracks | Takes per track | Generations | Cost |
|---|---|---|---|
| 1 | 3 | 3 | $0.375 |
| 1 | 5 | 5 | $0.625 |
| 12 | 5 | 60 | $7.50 |
| 24 | 4 | 96 | $12.00 |
Spending fewer takes
A specific brief is the cheapest lever. State tempo as a number such as 72 BPM, name two to four instruments with texture, and name one moment in the arc, for example a breakdown at 0:20. End with one clause such as instrumental, no vocals. Do not send a negative_prompt or a duration field: both are rejected, and exclusions go into the positive prompt.
If a take is close, keep the brief and change one axis rather than rewriting everything; that makes each retry informative.
What the fee does not cover
The $0.125 covers generation only. Placing the track under a video is a Timeline render at $0.10 per output minute, and ducking under narration is a render option. Full rules are in Music 1.0 and Timeline 1.0. Policy rejections are retried after changing the flagged content, within your own budget.
Keep the best take, record the prompt
Save the prompt that produced a track you liked, next to the artifact URL. Because there is no seed, the prompt is the only handle you have on the style. When you need a second track in the same family, change the tempo and key but keep the instrument list and production era, which the docs name as the continuity levers.
If a project needs contrast between scenes, the docs advise changing the genre family, the tempo by at least 12 BPM and the lead instrument. That reduces the number of near-duplicate takes you pay for.
When to stop retrying
Set a take cap per track in advance. Five takes is $0.625, and a sixth rarely beats the first five if the brief is the problem. Rewrite the brief after the second weak take instead of rolling the same prompt again.
Sources
More in Pricing
- Music for 20 video ads, three takes each: 60 generations, $7.50
Sume Music is a fixed $0.125 per accepted generation, whatever the prompt length. Three takes for each of 20 ads is 60 generations and $7.50. Arithmetic inside.
- Music Router: 24 takes cost $3.00, every model charges $0.125
Sume's Music Router charges the fixed Music price per audio generation whichever model it picks. Twenty-four takes cost 24 x $0.125 = $3.00.
- Music Router at 12.5 cents a song vs ElevenLabs at 15 cents a minute
Sume's Music Router charges a fixed $0.125 per generation. ElevenLabs lists Music at $0.15 a minute, and Gemini lists Lyria 3.5 at $0.08 a song. The sums.
- Nano Banana 2.1 for 400 product images: $30 at 0.5K up to $80 at 4K
Nano Banana 2.1 replaced Nano Banana 2 on Sume this week. 400 images cost $30 at 0.5K, $40 at 1K, $60 at 2K and $80 at 4K. Arithmetic and a mixed plan.
Written by Sume