Test log
Eleven Music test data
Raw results for the Eleven Music review. The verdict is in the review.
- Tested
- 7 November 2026
- Tester
- Miles Foster
- Account
- Free demo account, signed up by RankDude, nothing paid to the brand
- Tests
- 10
- Screenshots
- 6
- Method
- Published methodology
- Environment
- Google Chrome on desktop
Results
| ID | Test | Key measurement | Score |
|---|---|---|---|
| T01 | Format, Download & Export | – | n/a |
| T02 | Audio Quality & Fidelity | Stream bit rate: 127 to 133 kbps | 4/5 |
| T03 | Prompt Adherence | Prompts fully met: 9 of 12 | 4/5 |
| T04 | Vocals & Lyrics | Written words heard, Spanish: 94 percent | 3/5 |
| T05 | Musical Coherence & Structure | Tracks that stop dead: 4 of 12 | 4/5 |
| T06 | Length & Customisation Control | Control points that worked: 7 of 10 | 3/5 |
| T07 | Editing, Stems & Regeneration | Edit points that worked: 5 of 10 | 2/5 |
| T08 | Commercial-Use Rights & Licensing | Commercial use on the free plan: No | 4/5 |
| T09 | Generation Speed | Median per 30-second track: 4.2 s | 5/5 |
| T10 | Free-Tier Limits & Pricing | Free credits a month: 10,000 | 4/5 |
| Overall (in the review) | 4.5/5 | ||
Format, Download & Export
n/aNot Tested- Method
- The Download menu on a finished track was opened and each option pressed.
- Remark
- Left out of the average: the plan I used did not allow the test. No default score is given.
Raw notes
The Download menu on a finished track was opened and each option pressed.
Not graded: the free plan lets no file leave the product. The Full song download button did nothing when pressed (no file after five seconds, pressed twice). The Stems button opened "Upgrade to download stems". The page footer says "For downloading and commercial use, upgrade now."
From the public pricing table, read on the test date: Starter ($6 a month) adds 2 and 4 stems and API access but not Lossless WAV or High-quality download; Creator ($22) adds Lossless WAV and release to Spotify and Apple; Pro ($99) adds High-quality download and up to 6 stems.
- MP3, WAV, MIDI, bit rate choice, watermark, batch download, share link and cloud push were not run.
Audio Quality & Fidelity
4/5Tested- Method
- All 12 tracks were saved from the product's own player (the free plan could not download a file) and measured: codec, bit rate, sample rate, loudness, peak level, share of samples at full scale, the highest frequency with content, and how the ending behaves.
- Stream bit rate
- 127 to 133 kbps
- Tracks with a peak above full scale
- 4 of 12
- Remark
- Scale 4: clean with occasional minor artefacts. No track showed clipping at a level that counts, no track had silence or dropouts, and loudness was steady; four hot peaks and a 128 kbps stream keep it from a 5.
Raw notes
All 12 tracks were saved from the product's own player (the free plan could not download a file) and measured: codec, bit rate, sample rate, loudness, peak level, share of samples at full scale, the highest frequency with content, and how the ending behaves. This is a measurement of the audio, not a listening opinion.
Every track is stereo AAC at 44.1 kHz and 127 to 133 kbps, 30.0 seconds long. The highest frequency with content is about 17.2 kHz on every track, the usual ceiling of a stream at this bit rate.
Loudness runs from -11.6 LUFS (road-trip pop, the loudest) to -21.7 LUFS (ambient drone, the quietest).
Four tracks decode to a peak above full scale: the corporate loop +3.1 dB, the EDM track +1.5 dB, the jazz trio +1.3 dB and the reggaeton +0.6 dB. No track had more than 0.004 percent of its samples at full scale.
The two weakest tracks by measurement are the corporate loop (hottest peak, stops dead at 30 seconds) and the EDM track (peak above full scale, stops dead at 30 seconds).
- I did not listen on headphones and speakers; the grade rests on measurement. The free plan gives no download, so the file measured is the stream the player loads, not an export.
Prompt Adherence
4/5Tested- Method
- Each of the 12 frozen prompts names a genre, a mood, a tempo and instruments.
- Prompts fully met
- 9 of 12
- Tempo within 10 BPM
- 10 of 11
- Remark
- Scale 4: all four points on 7 to 9 prompts. 9 of 12 prompts met every point I could check.
Raw notes
Each of the 12 frozen prompts names a genre, a mood, a tempo and instruments. Genre was read from the genre tags the app put on the track, mood and instruments from the written plan the app shows for the song (Edit Song), tempo by measuring the saved audio (octave errors, half or double time, accepted) and sung parts by transcribing the audio.
Fully met on 9 of 12 prompts: road-trip pop, lo-fi beat, chase cue, corporate loop, EDM, jazz trio, chiptune, trailer build and nursery tune.
Tempo: 10 of the 11 prompts that name a tempo measured within 10 BPM of the number asked (pop 124 for 120, lo-fi 145 as double time of 72 for 70, chase 145 for 140, corporate 102 for 100, folk 82 for 80, EDM 65 as half time of 130 for 128, jazz 113 for 110, chiptune 76 as half time of 152 for 150, trailer 180 as double time of 90, nursery 102 for 100). The reggaeton measured 64 and could not be confirmed against 95.
Missed: the folk ballad asked for a male vocal and the transcriber heard no words (only "[instrumental music]"), and the reggaeton tempo was not confirmed. The ambient drone names no tempo, so it could not be scored on all four points and is not in the count of 9; on genre, mood and instruments it matched. Its plan lists percussion, drums and rhythmic hits as excluded.
- Instruments and mood are read from the app's own written plan, which says what it intended; only tempo and sung words were checked in the audio.
Vocals & Lyrics
3/5Tested- Method
- The four prompts that ask for singing (road-trip pop, folk ballad, Spanish reggaeton, nursery tune) were transcribed with ElevenLabs' own Speech to Text, 417 credits each, and compared word by word with the lyrics the app wrote for the song.
- Written words heard, Spanish
- 94 percent
- Vocal tracks with no words heard
- 1 of 4
- Remark
- Scale 3: understandable but lyric control limited, with one track with no usable vocal. Two tracks were clear (94 and 81 percent), one dropped a third of its words and one had no singing.
Raw notes
The four prompts that ask for singing (road-trip pop, folk ballad, Spanish reggaeton, nursery tune) were transcribed with ElevenLabs' own Speech to Text, 417 credits each, and compared word by word with the lyrics the app wrote for the song.
Spanish reggaeton: 76 of 81 written words heard, 94 percent, with the Spanish sung and spelled correctly. The written lyric itself has a slip: "Esta noche es para tú y yo".
Road-trip pop: 80 of 99 words, 81 percent. Two lines near the end of the chorus came out as different words.
Nursery tune: 37 of 55 words, 67 percent. The "Shh, little bear" lines and the "Sleepy bear, sleep tight" lines were not sung as written.
Folk ballad: no words heard. The transcriber returned only "[instrumental music]" for a track whose plan holds two verses of lyrics and a male vocal.
- A transcriber hearing a word does not prove it sounds natural or in tune; I did not rate the voice itself.
Musical Coherence & Structure
4/5Tested- Method
- Each track's section plan was read (names, seconds, BPM), and the saved audio was measured at the end: the time of the last audible sound and how long the level takes to fall from 6 dB to 40 dB below the body of the track.
- Tracks that stop dead
- 4 of 12
- Tracks that fade over a second or more
- 6 of 12
- Remark
- Scale 4: mostly coherent, the odd weak ending. 8 of 12 tracks fade or fall away rather than stop.
Raw notes
Each track's section plan was read (names, seconds, BPM), and the saved audio was measured at the end: the time of the last audible sound and how long the level takes to fall from 6 dB to 40 dB below the body of the track.
Every plan has two or three named sections that fill the 30 seconds, for example Verse 1 and Chorus, Intro Groove and Main Loop, Build and Drop, Verse and Outro.
Endings: 6 of 12 tracks fade over a second or more (lo-fi, chase, folk, jazz, ambient, nursery). 2 fall away in 0.4 to 0.6 seconds (pop, trailer). 4 stop dead at the 30-second mark within 0.2 seconds: the corporate loop, the EDM track, the chiptune theme and the reggaeton.
The most common fault is the abrupt stop at the length cap. I did not listen for a repeated loop, so that is not checked.
- Key and tempo drift were not measured beyond one tempo reading per track.
Length & Customisation Control
3/5Tested- Method
- Each of the ten points was tried on the test date, and a point counts only when it worked on this account.
- Control points that worked
- 7 of 10
- Remark
- Scale 3: 6 or 7 points.
Raw notes
Each of the ten points was tried on the test date, and a point counts only when it worked on this account.
Worked (7): tempo, set in the prompt and per section as a BPM tag; named instruments, written as tags in each section's plan; mood and energy; song structure, as named sections with a length in seconds that can be added; intro and outro, as sections of those names that came back in three plans; an instrumental track, by asking for no vocals or switching the Lyrics control; and exclusions, a second row of tags in each section that holds styles to leave out (the ambient drone listed percussion, drums and rhythmic hits there).
Not counted (3): length past the default, because the length menu offers 30 seconds, 1, 2, 4 and 6 minutes and custom but I only ran 30 seconds; a reference track, because the box has a "+ Reference" button I did not run; and key or scale, for which there is no control.
- Length past the default and the reference box need more credits than the free plan allowed once the 12 tracks were made.
Editing, Stems & Regeneration
2/5Tested- Method
- Three edits were run on the road-trip pop track in the Edit Song view and five of the ten points were checked against them; the stems and download menu was opened.
- Edit points that worked
- 5 of 10
- Stems on the free plan
- Paid only
- Remark
- Scale 2: 4 or 5 points.
Raw notes
Three edits were run on the road-trip pop track in the Edit Song view and five of the ten points were checked against them; the stems and download menu was opened.
Worked (5): regenerate, where each edit made a fresh full take of the song from the edited plan, kept in History; re-roll one section (Verse 1 regenerated in 4.7 seconds); an instrument added by the written edit (the "gentle piano outro" tag and a new take in 8.8 seconds); extend (a 10-second section added, the timeline went from 0:30 to 0:40, new take in 5.7 seconds); and history.
Not counted (5): stems, where the app opened "Upgrade to download stems" and said downloading stems requires a paid subscription; several variations at once (the menu offers 1 to 4, I ran 1); trimming; editing lyrics in place; and mix levels, which are a playback volume only. The vocal itself was not swapped.
The three edits used 780 credits together.
- Whether each edit changed the audible result was read from the app's history and plan, not by ear.
- Several variations at once and editing lyrics in place were not run to save credits.
Commercial-Use Rights & Licensing
4/5Tested- Method
- The terms of use and the Music page were read on 7 November 2026 and checked against the plan I was on (Free).
- Commercial use on the free plan
- No
- Cheapest plan with commercial use
- Starter
- Remark
- Scale 4: commercial rights clear but gated behind a reasonable paid plan.
Raw notes
The terms of use and the Music page were read on 7 November 2026 and checked against the plan I was on (Free).
Free use is non-commercial. The terms say: "if you access or use our Services free of charge (such a user, a “Free User”), you may only use the Services for non-commercial purposes". The Music pricing table says "Personal use only" for Free.
Paid: Starter is "Commercial, no streaming", Creator and Pro are "Commercial, not enterprise", and release to Spotify and Apple is ticked from Creator.
Ownership: "as between you and ElevenLabs, you retain all rights in and to your Output". ElevenLabs also takes a perpetual, non-exclusive licence to use your content to provide and improve the services and to develop new services and products.
The Music page says the tracks "are cleared for broad commercial use. Commercial rights vary by subscription tier." It says the music is built in partnership with artists, labels and publishers.
- Terms read by me, not a lawyer. No paid plan was bought, so the paid rights are from the terms and the pricing table, not from use.
Generation Speed
5/5Tested- Method
- Each generation was timed from pressing Generate to the finished track on the Generations list, on the same connection.
- Median per 30-second track
- 4.2 s
- Slowest track
- 7.0 s
- Remark
- Scale 5: median under 30 seconds.
Raw notes
Each generation was timed from pressing Generate to the finished track on the Generations list, on the same connection.
Median 4.2 seconds for a 30-second track, average 4.3, fastest 3.6, slowest 7.0. No generation failed and there was no queue on the free plan.
11 of the 12 were timed; the first track was made while I was still setting up the test. The three edits took 4.7, 8.8 and 5.7 seconds.
- 30-second tracks only; one-minute and longer tracks were not timed.
Free-Tier Limits & Pricing
4/5Tested- Method
- The free plan was used until I stopped at about 79 percent of its credits, and the public pricing page and the Music pricing table were read on 7 November 2026.
- Free credits a month
- 10,000
- Cheapest paid plan with commercial use
- Starter
- Remark
- Scale 4: a usable free tier to make tracks, and a paid plan under $12 that adds commercial rights and stems; clean export (Lossless WAV) needs a plan over $20, and the free tier cannot download.
Raw notes
The free plan was used until I stopped at about 79 percent of its credits, and the public pricing page and the Music pricing table were read on 7 November 2026.
Free: 10,000 credits a month, no card. A 30-second track costs 450 credits (900 a minute), so the credits pay for 22 thirty-second tracks. I made 12 tracks (5,400 credits), transcribed 4 of them (1,668) and ran three edits (780): 7,848 credits in all.
Paid: Starter $6 a month (30,000 credits, about 33 minutes of music), Creator $22 (121,000 credits, about 134 minutes) and Pro $99 (600,000 credits). Extra music costs $0.18, $0.16 and $0.15 a minute on those plans and $0.33 on Free.
The page runs a first-month offer on Starter and Creator. The cheapest plan that adds commercial rights and stems is Starter; Lossless WAV and Spotify release need Creator.
- No plan was bought, so the cost of a downloaded track was worked from the credit price, not from a download.
Coverage
Tested
- The 12 frozen prompts, 30 seconds each, first track kept: timing, audio measurements, tempo, genre tags and plans
- Speech to Text on the four vocal tracks to check the sung words
- A re-rolled section, a written edit and an added section on one track
- The Download menu, the stems paywall and the free plan's credit use
- The terms of use, the Music page and the pricing pages
Not tested or limited
- No file could be downloaded on the free plan, so audio was measured from the stream the player loads, and not by listening
- Tracks were 30 seconds, the shortest length, so the 12 prompts fit in the free credits
- Length past the default, the reference box, stems and format export were not run
- No plan was bought: paid rights, WAV and stems come from the terms and the pricing table
- Instruments and mood come from the app's written plan; only tempo and sung words were checked in the audio
Still to test
- Downloads, Lossless WAV and stems on a paid plan
- Tracks of one minute and longer, and a reference track
- Listening tests on headphones and speakers
- Several variations at once and lyric edits