Skip to content
trending

Benchmarks

Eleven Music: music_v2_5 vs music_v1 API Default Test

We tested Eleven Music's music_v2_5 against the API default music_v1 with English lyrics. Both models sang clearly and hit the requested genres.

By 6 min read
Eleven Music: music_v2_5 vs music_v1 API Default Test

👁 117 reads

Eleven Music is ElevenLabs’ AI music generator: you give it a style description and optional lyrics, and it produces a complete track with or without vocals. We tested it with four samples: a lo-fi instrumental and an indie pop song with English lyrics, comparing the new music_v2_5 model against music_v1, which is still the API default during the transition period. Below you’ll find all four tracks, generation times, costs, and measurements of genre accuracy and vocal clarity.

Summary

All four tracks landed in the requested genre according to CLAP, and the English lyrics were sung clearly: both songs had just 3 differences in 54 words. Each track was ready in 6-10 seconds, and the four samples (2.5 minutes of music) used 5734 credits - approximately $0.15 per minute via API. The music_v2_5 model showed slightly stronger genre alignment for lo-fi. Best for: background music for videos, podcast intros, ads, jingles with vocals. Not for: tracks longer than 5 minutes or work requiring full DAW-level control.

3/54differences in English lyrics sung by AI
6-10 sgeneration time for 30-45 s tracks
5734credits for 4 tracks, 2.5 min total
$0.15price per minute of music via API

What Eleven Music Can Do

  • Models: Music v2.5 (newest), v2, and v1. Important: the API still defaults to v1 during the transition period, so you must specify the newer model manually (model_id: music_v2_5). The web app defaults to v2.
  • Length: according to the documentation, from 3 seconds to 5 minutes.
  • Vocal or instrumental: the force_instrumental parameter guarantees a version without singing.
  • Control: composition plan (sections with lengths and styles in JSON), inpainting to regenerate a single fragment, and Audio Reference - a ~30 s sample that guides the sound (v2 and v2.5, paid plans, each sample is checked for copyright).
  • License: ElevenLabs states that the music is approved for nearly all commercial uses, from videos to ads and games. Commercial use starts with the Starter plan.

Our Test: Four Tracks, music_v2_5 vs music_v1

We generated four tracks via the API on September 29, 2026, comparing music_v2_5 with music_v1 (still the API default). We didn’t judge genre by ear - we used the CLAP model, which compares audio against descriptions of six genres (lo-fi, indie pop, rock, classical, EDM, folk). We transcribed the vocals with Scribe and compared them to the input lyrics.

Lo-fi Instrumental: music_v2_5 vs music_v1

Both with force_instrumental, 30 seconds each. Transcription found no words, only an [outro jingle] marker.

Upbeat lo-fi hip hop beat for a YouTube vlog intro, warm Rhodes piano
chords, vinyl crackle, soft kick and snare, 85 BPM, instrumental
Lo-fi, music_v2_5 (30 s, generated in 6.3 s).
Lo-fi, music_v1 (API default) (30 s, 6.9 s).

Song with English Lyrics

We wrote the lyrics ourselves: a verse and chorus about a late bus in the rain. We tested the same prompt with both models.

Dreamy indie pop song with a soft female vocal singing in English,
clean electric guitar, warm bass, brushed drums, 100 BPM,
melancholic but hopeful. Lyrics:
[Verse]
The late bus splashes through the street,
The rain makes mirrors at my feet,
I hold my coffee in both hands,
And watch the morning shift its plans.
[Chorus]
Seattle wakes up slow and grey,
Nobody asks you where you'll stay,
Five minutes just for me to breathe,
Before the city shows its teeth.
Song, music_v2_5 (45 s, 7.0 s).
Song, music_v1 (API default) (45 s, 10.2 s).

Results

Track Genre Identified by CLAP Score for Requested Genre Lyrics: Differences / Words Detected Language
Lo-fi, music_v2_5 lo-fi 0.540 - -
Lo-fi, music_v1 (API default) lo-fi 0.448 - -
Song, music_v2_5 indie pop 0.386 3 / 54 English (96.4%)
Song, music_v1 (API default) indie pop 0.330 3 / 54 English (96.5%)

All four tracks landed in the requested genre - CLAP identified it as the best match out of six each time. The music_v2_5 model showed stronger genre alignment for lo-fi (0.54 vs 0.448) and indie pop (0.386 vs 0.330). The lyrics were sung clearly and intelligibly: Scribe transcribed 51 of 54 words correctly in both versions, and automatic language detection recognized English with ~96% probability. The three differences per song were minor: one added word (“singing” or “humming”), the grey/gray spelling variant, and one dropped contraction. This clarity contrasts with ElevenLabs’ sound effects generator, which in our sound effects test struggled with non-English prompts.

What we didn’t measure: how natural the phrasing sounds to native English speakers, and how Eleven Music compares to Suno on identical prompts. That’s a topic for a separate comparison. We’ve collected the current state of music tools in our AI music guide.

When We’d Use Eleven Music

This is our opinion after this test. The strongest point is speed and predictability for short forms: a 30-second background track for a YouTube video or a jingle with vocals appears in seconds and comes with a commercial license on paid plans. For longer songs, the composition plan will be more convenient - you set the length and style of verse, chorus, and ending separately, then use inpainting to fix only the weak fragment instead of re-rolling the entire track. If you’re already using ElevenLabs for voice-over or dubbing, music from the same account saves you one subscription.

How to Write Prompts for Eleven Music

  • Start with genre and mood, then instruments, tempo in BPM, and vocal type.
  • State the singing language explicitly (“female vocal singing in English”). Add lyrics after “Lyrics:” and mark sections [Verse], [Chorus].
  • If you don’t want vocals, in addition to the word “instrumental” set force_instrumental in the API, which according to the documentation guarantees a version without singing.
  • Set length with a parameter, not in the description.

The API command we used (tested September 29, 2026, also with 10 s length):

curl -X POST "https://api.elevenlabs.io/v1/music?output_format=mp3_44100_128" 
  -H "xi-api-key: $ELEVENLABS_API_KEY" -H "Content-Type: application/json" 
  -d '{"prompt":"Upbeat lo-fi hip hop beat for a YouTube vlog intro, warm Rhodes piano chords, vinyl crackle, 85 BPM, instrumental",
       "music_length_ms":30000,"model_id":"music_v2_5","force_instrumental":true}' -o music.mp3

How Much Does Eleven Music Cost

API pricing as of September 29, 2026: $0.15 per minute of music, excluding VAT. In our test, four tracks totaling 150 seconds used 5734 credits, approximately 38 credits per second, around 2300 per minute. At this rate, the free plan (10,000 credits, no commercial use) is enough for about 4 minutes of music per month, Starter at $6 for about 13 minutes, and Creator at $22 for about 52 minutes. According to ElevenLabs help, cost depends on length and number of variants, and the app shows it before you click “Generate”. We compare plans in our ElevenLabs pricing guide, and all the company’s tools in our ElevenLabs overview.

If you need voice-over alongside music, we described the newest voice model in our article about Eleven v4.

Frequently Asked Questions

Can Eleven Music sing in languages other than English?

Yes. In a separate test with Polish lyrics, two songs were sung completely: transcription recovered 37 out of 37 words, and the language was recognized as Polish with ~98% probability. You need to specify the language explicitly in the prompt.

How much does Eleven Music cost?

Via API, $0.15 per minute, excluding VAT. In paid plans you pay with credits: our 2.5 minutes of music used 5734 credits, approximately 2300 per minute.

Can you use Eleven Music commercially?

Yes, starting with the Starter plan. ElevenLabs states that Eleven Music is approved for nearly all commercial uses; details are in the music terms.

How long a track can Eleven Music make?

According to the documentation, from 3 seconds to 5 minutes. Our 30 and 45 second tracks were ready in 6-10 seconds.

Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →

Sources and Date Checked

Kinetyra

The link to ElevenLabs is a partner link: if you buy through it, QUANTSLANT earns a commission and you pay the same price. It has no influence on the content, including where the coverage is critical.

// read next

More on this from QUANTSLANT

Piotr Olszewski

Editor and publisher · Quantslant

I edit and publish Quantslant. AI tools help with research, drafts and translation; choosing topics, running the tests, checking facts and the decision to publish are mine, and so is the responsibility for what appears here.

read nextElevenLabs Voice Cloning: clones from 10 s, 30 s and 2 minai toolsElevenLabs Voice Cloning: clones from 10 s, 30 s and 2 minWe cloned a synthetic voice from 10 s, 30 s and 2 minutes of audio. The 10-second clone reached 0.992 similarity, the 2-minute clone 0.996.

Most read

  1. 1Best AI Video Generator Apps for Your Phone (2026)
  2. 2Claude Opus 5.5 Motion Graphics: 60s Showreel from One Prompt
  3. 3Wan 2.2 on RTX 4090: timings, VRAM and four test clips
// site map

This page in other languages9

Quantslant in numbers 1,427posts in archive26news from the last 7 days18video models in the observatory
🎬 AI Video Benchmarks: Higgsfield DoP vs OpenAI Sora • Higgsfield vs Runway Gen-3 • AI Video Credit Calculator
× ‹ enlarge ›