AI music cost per track: Sume Music vs Lyria 3.5 and ACE Step

Sume Music is a flat $0.125 for each accepted generation, so 100 tracks are $12.50. Google's Gemini API page lists Lyria 3.5 at $0.08 per song (read 2026-10-07), or $8.00 per 100. The Sume docs say Music 1.0 runs on Google Lyria 3.5, so Sume's price is 56 percent above Google's listed per-song price ($0.125 / $0.08 = 1.5625), a larger gap than the 25 percent margin on most Sume models.

What is in the price

Sume's price does not vary with prompt length or optional image conditioning. The model makes full-length structured songs of up to a few minutes, steered by the prompt. For new integrations, Sume recommends the Music Router at POST /v1/music-router/generate, with sume/music-auto selecting the engine; the catalog lists it at the same $0.125. Music 1.0 is being retired and resolves through the router.

The ACE Step row needs a warning

fal's page lists ACE Step Prompt to Audio at $0.0002 per second (5,000 seconds per dollar). A 120-second track is 120 x $0.0002 = $0.024, and 100 are $2.40. That is a different, much cheaper class of model. Sume does not list ACE Step, so I make no quality or catalog claim. Whether a Google song and a 120-second ACE Step track are comparable depends on your use.

Where music sits in the budget

For a 100-ad batch, the music is $12.50 if every ad has its own track. Reusing one track across ten ads cuts that to $1.25. Compared with video, which is dollars per clip, music is a rounding error; the larger saving is nearly always on the video line.

Gaps to know about

I did not read how long a Lyria 3.5 song is on Google's page, so the per-song comparison is unit price only. Sume's docs do not state a fixed track length either. Generate a few and measure the duration you actually get before multiplying.

Ads that need a fresh track each

Suppose each of 200 ads gets its own track. Sume costs 200 x $0.125 = $25.00. Google's listed Lyria 3.5 price would be 200 x $0.08 = $16.00, and fal's ACE Step for 120 seconds would be 200 x $0.024 = $4.80. The Sume premium over Google on this batch is $9.00.

Set that against the video. 200 fifteen-second ads at Standard avatar rates are 200 x $2.76 = $552.00, so $25.00 of music is 4.5 percent of the bill. If budget is tight, the first saving is reusing one track across a campaign, not switching music vendors.