What is the cheapest AI music generation API in 2026?

Google's Lyria 3.5 at $0.08 per full song, available in the Gemini API since September 4, 2026. Because it prices per song rather than per minute, a two-minute track costs the same as a forty-second one. ElevenLabs' Eleven Music charges $0.15 per minute, about $0.30 for a two-minute track — roughly four times more for the same length.

Which AI music API gives exact track lengths?

Eleven Music. It sets duration in milliseconds through music_length_ms, from about 3 seconds to 10 minutes, and supports per-section durations through a composition plan. Lyria 3.5 has no duration parameter at all — Google documents output as 'a couple of minutes' guided by the prompt, and says exact duration can be influenced rather than set.

Which AI music API supports WAV output?

Eleven Music returns MP3 or WAV. Lyria 3.5 returns MP3 only, at 44.1 kHz stereo. If the audio has to be mastered, layered under other stems, or handed to an editor, MP3-only output is a hard constraint rather than a preference.

Do AI music APIs watermark their output?

Google embeds a SynthID audio watermark in every Lyria 3.5 track. Watermarking supports provenance and disclosure compliance, which is increasingly relevant under 2026 AI content-marking rules, but it is a consideration if your workflow assumes unmarked audio. Check each vendor's current documentation, since watermarking policies have been changing across the industry.

Quick Answer

Best AI Music Generation APIs 2026: Ranked by Cost

Published: September 7, 2026

The Short Answer

Two APIs are worth calling programmatically in 2026, and they are not substitutes for each other.

#	API	Price	Two-min track	Best for
1	Eleven Music (ElevenLabs)	$0.15/minute	$0.30	Anything with a deadline on the timeline
2	Lyria 3.5 (Google)	$0.08/song	$0.08	Volume background scoring
3	Lyria 3 Clip (Google, legacy)	$0.04/30s clip	$0.16	Fixed-length stings
—	Suno	$10–$30/month	subscription	Humans, not pipelines

The ranking flips depending on one question: does the length matter? If yes, Eleven Music, and the 4x price difference is irrelevant. If no, Lyria 3.5 and it is not close.

1. Eleven Music — The Only One That Fits a Timeline

$0.15 per minute · MP3 or WAV · section-level editing

This is first not because it is cheapest — it is nearly 4x Lyria per two-minute track — but because it is the only option that behaves like a production tool rather than a slot machine.

What you get for the premium:

Exact duration in milliseconds via music_length_ms, roughly 3 seconds to 10 minutes. You ask for 8,500 ms and you get 8,500 ms.
Per-section durations through a composition plan — intro, verse, chorus, each with its own length.
Section-level editing after generation. Change the sound or the lyrics of one part without regenerating the whole track.
~30-second audio reference to steer style toward an existing cue — the closest thing to “make it sound like our brand.”
WAV output, which anyone mastering or layering audio will need.
Multilingual generation.
Explicit commercial clearance on paid plans — ElevenLabs states its paid tiers are cleared for nearly all commercial uses.

Choose it when: the music sits under a video with hard cuts, when a client will ask for a revision, when the deliverable needs mastering, or when you need a licence position you can point a lawyer at.

⚠️ The cost is real at volume. A hundred two-minute tracks a month is $30 against $8 for Lyria. If nothing above is a requirement, you are paying for control you will not use.

2. Lyria 3.5 — Cheapest Full Songs, Fewest Guarantees

$0.08 per song · MP3 44.1 kHz stereo · public preview

Google made Lyria 3.5 callable through the Gemini API on September 4, 2026 under the model code lyria-3.5 (the model itself launched inside Flow Music on July 29, 2026). It is a latent-diffusion model over temporal audio latents that takes text and images and returns a full song — verses, choruses, bridges, vocals or instrumental — plus timed lyrics.

Per-song pricing is the whole pitch. $0.08 for a complete track means length is free. At scale — a bed for every reel, every episode, every product demo — that is the cheapest full-song generation available.

Image input is the underrated feature. Conditioning on a keyframe or brand board is a genuinely different steering mechanism from a text-plus-audio-reference approach, and no competitor here offers it.

⚠️ The constraints, all from Google’s own docs:

Limit	Detail
Duration	No parameter. “A couple of minutes,” prompt-guided; no published maximum
Editing	Single-turn. No revision — re-prompt and regenerate
Format	MP3 only. No WAV, no stems
Free tier	None
Watermark	SynthID on every track
Safety	Blocks artist-voice and copyrighted-lyric prompts
Not supported	Function calling, structured outputs, caching, Live API, batch

That last row matters more than people expect. No batch and no caching means Lyria 3.5 is a plain request/response endpoint, not something you compose into a larger Gemini pipeline.

Also worth knowing: Google publishes no numeric benchmarks. The model card reports qualitative gains over Lyria 2 on audio fidelity and lyric prompt adherence — no MOS score, no head-to-head. You have to run your own listening test. Budget for it.

3. Lyria 3 Clip — Still Useful for Fixed-Length Stings

$0.04 per 30-second clip · legacy

The previous-generation clip model remains listed. It is worth knowing about for exactly one reason: it has a fixed 30-second length, which is a crude form of the duration control Lyria 3.5 lacks.

For stings, transitions and short loops, four predictable 30-second clips at $0.16 may beat one unpredictable two-minute track at $0.08. Lyria 3 Pro is also still listed at $0.08 per full song.

Suno — Great Product, Wrong Interface

Free / $10 / $30 per month · v5.5 flagship

Suno’s stable flagship is v5.5 (released March 26, 2026), with v4, v4.5, v4.5+ and v5 selectable on paid plans and v4.5-all on the free tier. Pricing: free with 50 daily credits, Pro $10/month ($8 billed annually), Premier $30/month ($24 billed annually).

It is the best of these for a human making music, and the free tier is genuinely generous for experimentation. But subscription credits and a web app are the wrong shape for a pipeline that calls an endpoint a thousand times a month. Licensing note: paid plans grant output ownership; free-tier output is non-commercial.

Cost at Volume

100 two-minute tracks per month:

Option	Monthly
Lyria 3.5	$8.00
Lyria 3 Pro	$8.00
Lyria 3 Clip (stitched)	$16.00
Eleven Music	$30.00
Suno Premier	$24.00 (credit-limited)

A $22/month difference between the cheapest and the most controllable. For almost any business use, that gap is smaller than the cost of one round of manual re-editing — which is the honest argument for paying it.

The Licensing Reality Check

Every vendor here offers some reassurance about rights. None can deliver what people assume they are buying.

The US Copyright Office position is that works generated without meaningful human authorship are not copyrightable. A vendor granting you ownership is a contract between you and that vendor — not an enforceable copyright against a third party who copies your track.

For background beds and stings, this rarely matters. For anything that is the product — a released single, a licensed sync — get real legal advice and expect meaningful human authorship to be part of the answer.

How to Choose

If you need…	Pick
Cheapest full tracks at volume	Lyria 3.5
Exact timing to a video cut	Eleven Music
WAV or masterable output	Eleven Music
To revise a section after generating	Eleven Music
Image-conditioned generation	Lyria 3.5
Style transfer from a reference track	Eleven Music
Fixed 30-second stings	Lyria 3 Clip
A human-in-the-loop creative tool	Suno v5.5

Default recommendation: prototype on Lyria 3.5 because it is cheap enough to iterate freely, and move to Eleven Music the moment duration, format or revision control becomes a requirement rather than a nice-to-have.

Last verified: September 7, 2026.