GoogleAudio modelActive
Lyria 3.5
PopularityRank 145 of 875
47out of 100same as last month
Summary
Lyria 3.5 is Google’s flagship model for generating full-length songs. Its defining strength is structural coherence across a longer piece, not merely producing a short loop from a prompt. Lyria 3.5 accepts text or images and returns music audio together with lyrics, not speech synthesis or voice-agent replies.
Google’s catalog separates Lyria 3 Clip for 30-second segments from Lyria 3.5. Lyria 3.5 targets complete songs with coherent structure, while the Clip variant targets short, fixed-length segments.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 13 published benchmark results
Page 1 of 2
Measurements without matching peer values
These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.
A Thousand Listeners scene-equalized full-song preference score (Show measurement, test conditions, and source)
- Source value
- 34.6
- Score
- 34.6
- Metric
- Scene-equalized model score from blind head-to-head listener preferences
- Unit
- %
- Benchmark version
- A Thousand Listeners v1.0; August 2026 collection window
- Category
- music generation
- Direction
- Higher is better
- Lower confidence bound
- 31.7
- Upper confidence bound
- 37.8
- Source type
- community-reported
- Evaluator
- AI Music Benchmark, A Thousand Listeners
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Community-reported result. Full-song generation preference evaluation; 1,000 listeners, 104 prompts across 13 scenarios, two candidate tracks per model and prompt, and 6,000 primary pairwise judgements. Pairwise preferences are aggregated within scenario and the 13 scenarios receive equal weight. The report labels the Lyria column "Evaluated 3.5"; uncertainty uses 300 judgement-level bootstrap resamples within scenario, seed 20260829. Exact generation route and prompts are not published.
- Context
- The evaluator identity/affiliation is not independently established; treat as a community-reported study. The score is a scenario-weighted preference statistic, not raw vote share; raw Lyria votes were 399 of 1,858 (21.5%). The report does not publish an exact model route or generation prompt pipeline.
Artificial Analysis AA-Music-Instrumental Elo (Show measurement, test conditions, and source)
- Source value
- 1078
- Score
- 1078
- Metric
- blind pairwise same-prompt music preference Elo
- Unit
- elo
- Benchmark version
- AA-Music-Instrumental v1.1
- Category
- audio
- Direction
- Higher is better
- Sample
- 2302
- Confidence interval
- 21
- Source type
- independent-evaluator
- Evaluator
- Artificial Analysis
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Independent evaluation. Blind pairwise comparisons generated from the same prompts, judged by validated evaluators; instrumental generation cohort; 2,302 appearances.
- Context
- Appearances are not unique listeners; vocal and instrumental cohorts measure separate tasks.
Artificial Analysis AA-Music-Vocal Elo (Show measurement, test conditions, and source)
- Source value
- 1052
- Score
- 1052
- Metric
- blind pairwise same-prompt music preference Elo
- Unit
- elo
- Benchmark version
- AA-Music-Vocal v1.1
- Category
- audio
- Direction
- Higher is better
- Sample
- 2548
- Confidence interval
- 19
- Source type
- independent-evaluator
- Evaluator
- Artificial Analysis
- Status
- active
- Retrieved at
- 2026-10-15
- Methodology
- Independent evaluation. Blind pairwise comparisons generated from the same prompts, judged by validated evaluators; vocal generation cohort; 2,548 appearances.
- Context
- Appearances are not unique listeners; vocal and instrumental cohorts measure separate tasks.
FreeSongMaker Lyria 3.5 generation time for a 60-second request (Show measurement, test conditions, and source)
- Source value
- 27.9
- Score
- 27.9
- Metric
- wall-clock generation latency for a 60-second requested song
- Unit
- seconds
- Benchmark version
- FreeSongMaker direct Gemini API measurements, 2026-09-05
- Category
- audio generation / generation latency
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- FreeSongMaker direct Gemini API observation
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. FreeSongMaker says it called the Gemini API directly with exact model ID lyria-3.5 on 2026-09-05, using its production generator's same prompt preamble and default settings. This table row is an individual 60-second request measured wall-clock from request to response on a machine in Asia; it returned 101.5 seconds of audio. Neither a repeated-run count for this request nor an immutable serving revision is published.
- Context
- This is one externally reported API observation, not a repeated benchmark distribution. The requested duration overshot substantially, and wall-clock timing varies with region and load; neither API serving revision nor immutable weights are pinned. No CI is published.
FreeSongMaker Lyria 3.5 generation time for a 120-second request (Show measurement, test conditions, and source)
- Source value
- 30.8
- Score
- 30.8
- Metric
- wall-clock generation latency for a 120-second requested song
- Unit
- seconds
- Benchmark version
- FreeSongMaker direct Gemini API measurements, 2026-09-05
- Category
- audio generation / generation latency
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- FreeSongMaker direct Gemini API observation
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Independent evaluation. FreeSongMaker says it called the Gemini API directly with exact model ID lyria-3.5 on 2026-09-05, using its production generator's same prompt preamble and default settings. This table row is an individual 120-second request measured wall-clock from request to response on a machine in Asia; it returned 121.0 seconds of audio. Neither a repeated-run count for this request nor an immutable serving revision is published.
- Context
- This is one externally reported API observation, not a repeated benchmark distribution. Wall-clock timing varies with region and load; neither API serving revision nor immutable weights are pinned. No CI is published.
Fuser 30-second instrumental output duration, run 1 (Show measurement, test conditions, and source)
- Source value
- 60.7
- Score
- 60.7
- Metric
- Duration of generated track, measured with ffprobe
- Unit
- seconds
- Benchmark version
- Fuser comparison, published 2026-10-02 and updated 2026-09-28
- Category
- audio
- Direction
- No direction published
- Source type
- independent-evaluator
- Evaluator
- Fuser, Lyria 3.5 vs MiniMax Music 3
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. Instrumental lo-fi hip-hop brief: instrumental only, no vocals; 82 BPM in F major; prompt ends “A 30-second track”. Two runs per model; duration measured with ffprobe.
- Context
- Fuser states that nobody judged the music by ear; these are two individual output-duration measurements, not a perceptual quality score.
Fuser 30-second instrumental output duration, run 2 (Show measurement, test conditions, and source)
- Source value
- 61.1
- Score
- 61.1
- Metric
- Duration of generated track, measured with ffprobe
- Unit
- seconds
- Benchmark version
- Fuser comparison, published 2026-10-02 and updated 2026-09-28
- Category
- audio
- Direction
- No direction published
- Source type
- independent-evaluator
- Evaluator
- Fuser, Lyria 3.5 vs MiniMax Music 3
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Independent evaluation. Instrumental lo-fi hip-hop brief: instrumental only, no vocals; 82 BPM in F major; prompt ends “A 30-second track”. Two runs per model; duration measured with ffprobe.
- Context
- Fuser states that nobody judged the music by ear; these are two individual output-duration measurements, not a perceptual quality score.
FreeSongMaker Lyria 3.5 180-second-request output duration, reported minimum (Show measurement, test conditions, and source)
- Source value
- 178.8
- Score
- 178.8
- Metric
- Generated audio duration, lower endpoint of the range reported across three 180-second requests
- Unit
- seconds
- Benchmark version
- FreeSongMaker direct Gemini API measurements, 2026-09-05
- Category
- audio generation / output duration
- Direction
- No direction published
- Sample
- 3
- Source type
- community-reported
- Evaluator
- FreeSongMaker direct Gemini API observation
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Community-reported result. FreeSongMaker reports three direct Gemini API requests using exact model ID lyria-3.5 with a 180-second requested length, the same prompt preamble as its production generator, and default settings. It reports a 178.8-181.1 second audio-duration range and a separate 37-52 second request-to-response generation-time range; the range endpoints are not paired across metrics. Wall-clock time was measured on a machine in Asia and may vary with region and load.
- Context
- The numeric value is the reported lower endpoint of a three-run range, not an individual run or a mean. No confidence interval, per-run records, exact prompt, or immutable serving revision is published. Output duration is not a perceptual audio-quality score. Source role: observation by the product/API operator, not an independent quality replication.
FreeSongMaker Lyria 3.5 180-second-request output duration, reported maximum (Show measurement, test conditions, and source)
- Source value
- 181.1
- Score
- 181.1
- Metric
- Generated audio duration, upper endpoint of the range reported across three 180-second requests
- Unit
- seconds
- Benchmark version
- FreeSongMaker direct Gemini API measurements, 2026-09-05
- Category
- audio generation / output duration
- Direction
- No direction published
- Sample
- 3
- Source type
- community-reported
- Evaluator
- FreeSongMaker direct Gemini API observation
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Community-reported result. FreeSongMaker reports three direct Gemini API requests using exact model ID lyria-3.5 with a 180-second requested length, the same prompt preamble as its production generator, and default settings. It reports a 178.8-181.1 second audio-duration range and a separate 37-52 second request-to-response generation-time range; the range endpoints are not paired across metrics. Wall-clock time was measured on a machine in Asia and may vary with region and load.
- Context
- The numeric value is the reported upper endpoint of a three-run range, not an individual run or a mean. No confidence interval, per-run records, exact prompt, or immutable serving revision is published. Output duration is not a perceptual audio-quality score. Source role: observation by the product/API operator, not an independent quality replication.
FreeSongMaker Lyria 3.5 180-second-request generation latency, reported minimum (Show measurement, test conditions, and source)
- Source value
- 37
- Score
- 37
- Metric
- Request-to-response wall-clock generation latency, lower endpoint of the range reported across three 180-second requests
- Unit
- seconds
- Benchmark version
- FreeSongMaker direct Gemini API measurements, 2026-09-05
- Category
- audio generation / request-to-response latency
- Direction
- No direction published
- Sample
- 3
- Source type
- community-reported
- Evaluator
- FreeSongMaker direct Gemini API observation
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Community-reported result. FreeSongMaker reports three direct Gemini API requests using exact model ID lyria-3.5 with a 180-second requested length, the same prompt preamble as its production generator, and default settings. It reports a 178.8-181.1 second audio-duration range and a separate 37-52 second request-to-response generation-time range; the range endpoints are not paired across metrics. Wall-clock time was measured on a machine in Asia and may vary with region and load.
- Context
- The numeric value is the reported lower endpoint of a three-run range, not an individual run or a mean. No confidence interval, per-run records, exact prompt, or immutable serving revision is published. The source does not pair this latency endpoint with a specific output-duration endpoint. Source role: observation by the product/API operator, not an independent quality replication.
FreeSongMaker Lyria 3.5 180-second-request generation latency, reported maximum (Show measurement, test conditions, and source)
- Source value
- 52
- Score
- 52
- Metric
- Request-to-response wall-clock generation latency, upper endpoint of the range reported across three 180-second requests
- Unit
- seconds
- Benchmark version
- FreeSongMaker direct Gemini API measurements, 2026-09-05
- Category
- audio generation / request-to-response latency
- Direction
- No direction published
- Sample
- 3
- Source type
- community-reported
- Evaluator
- FreeSongMaker direct Gemini API observation
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Community-reported result. FreeSongMaker reports three direct Gemini API requests using exact model ID lyria-3.5 with a 180-second requested length, the same prompt preamble as its production generator, and default settings. It reports a 178.8-181.1 second audio-duration range and a separate 37-52 second request-to-response generation-time range; the range endpoints are not paired across metrics. Wall-clock time was measured on a machine in Asia and may vary with region and load.
- Context
- The numeric value is the reported upper endpoint of a three-run range, not an individual run or a mean. No confidence interval, per-run records, exact prompt, or immutable serving revision is published. The source does not pair this latency endpoint with a specific output-duration endpoint. Source role: observation by the product/API operator, not an independent quality replication.
FreeSongMaker Lyria 3.5 240-second-request successful output duration (Show measurement, test conditions, and source)
- Source value
- 181.7
- Score
- 181.7
- Metric
- Duration of the one successful audio output among two 240-second requests
- Unit
- seconds
- Benchmark version
- FreeSongMaker direct Gemini API measurements, 2026-09-05
- Category
- audio generation / output duration
- Direction
- No direction published
- Sample
- 1
- Source type
- community-reported
- Evaluator
- FreeSongMaker direct Gemini API observation
- Status
- active
- Retrieved at
- 2026-10-19
- Methodology
- Community-reported result. FreeSongMaker reports two direct Gemini API requests using exact model ID lyria-3.5 with a 240-second requested length, the same prompt preamble as its production generator, and default settings. One request returned 181.7 seconds of audio; the other returned no audio. The source separately lists 37.9 seconds of generation time but does not identify which attempt it describes. Wall-clock time was measured on a machine in Asia and may vary with region and load.
- Context
- N=1 refers only to the one numerical successful output, not all requests. The second attempt is reported as no audio and is not converted to a zero score or a success-rate estimate. The 37.9-second latency is excluded because its run attribution is unspecified. Output duration is not a perceptual audio-quality score; no confidence interval, per-run record, exact prompt, or immutable serving revision is published. Source role: observation by the product/API operator, not an independent quality replication.
Page 1 of 2
Popularity compared
How often the model is searched for, used and mentioned, compared with every model of its category.
Popularity index
0 to 100, higher is more popular
20 of 83 models
pyannote Community-1, pyannote: 73
Nemotron-3-Diarization, NVIDIA: 57
Gemini 3.5 Transcribe, Google: 52
Confucius4-R2T2, NetEase Youdao: 52
Granite Speech 5.0 470M TurboCTC, IBM: 52
Kokoro-82M, hexgrad: 51
Nova-3, Deepgram: 51
Fish Audio S2 Pro, Fish Audio: 51
NVIDIA Nemotron 3.5 Transcribe, NVIDIA: 50
Granite Speech 5.0 470M TurboCTC NC, IBM: 50
Orukeet, Oruk: 49
Lyria 3.5, Google: 47
Gemini 3.5 Live Translate, Google: 46
Phonon-2, Fermion Research: 44
Aura-2, Deepgram: 44
Gemini 3.1 Flash TTS, Google: 44
MiniMax-Music3, MiniMax: 41
Audio8-ASR-Infinite, Edge0 AI: 41
GPT-Realtime 2.1, OpenAI: 41
Eleven v3, ElevenLabs: 39
Lyria 3.5
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 5, 2026
Lyria 3.5Gemini 3.5 Live TranslateOrukeetGranite Speech 5.0 470M TurboCTC NCPhonon-2
All specifications
Architecture
Input and output
Watermarking and provenance
AI watermarking refers to invisible signals embedded in the output. Visible marks and provenance metadata are listed separately. Unknown means there is insufficient evidence for that particular field.
- AI watermarking
- Yes. Google says the Gemini API marks generated audio with the inaudible SynthID watermark.
- Source: AI watermarking
Licence and access
API
Further details
Sources and data date
Show all 15 sources
Benchmark sources at a glance
Which independent evaluators tested this model. This is coverage, not a quality ranking.
No active result from the reviewed external evaluators can currently be attributed to this model. Other source evidence remains available below.
Without vendor claims and historical or disputed results; those stay labelled in the single values.