GoogleAudio modelPreview
Lyria 3 Clip
26out of 100same as last month
Summary
Lyria 3 Clip is the short-form member of Google’s music-generation family. It turns text or images into compact music segments and loops with a fixed length of 30 seconds, and it also returns lyrics. That fixed output limits the checkpoint to short segments rather than a full song that needs sustained structure.
Google’s catalog identifies Lyria 3.5 as the full-song variant with complex structural coherence. Lyria 3 Clip remains a preview endpoint with 30-second output and therefore a separate clip checkpoint, not a shortened output variant of the full-song model.
All benchmark results
Every published result with its source, test setup and comparison models.
Show all 5 published benchmark results
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.
OnMusic: completed generation rate (Show measurement, test conditions, and source)
- Source value
- 84.1
- Score
- 84.1
- Metric
- Completed means the song finished and was playable
- Unit
- %
- Direction
- Higher is better
- Sample
- 1023
- Source type
- independent-evaluator
- Evaluator
- OnMusic AI Song Generation Study 2026
- Status
- active
- Retrieved at
- 2026-10-21
- Methodology
- Independent evaluation. OnMusic-recorded Lyria 3 Clip (Google) generations, 1,023 events in the reported June 13–October 5, 2026 period; the comparison includes only models with more than 300 generations.
- Context
- Observational platform data, not a controlled laboratory quality benchmark. The exact Google serving revision and separate latency denominator are not reported. Render time includes queue time; the study says 91 system-wide file-retrieval failures were excluded but does not give a model-specific effect.
OnMusic: provider-rejected generation rate (Show measurement, test conditions, and source)
- Source value
- 10.9
- Score
- 10.9
- Metric
- Rejected means the model provider declined the prompt
- Unit
- %
- Direction
- Lower is better
- Sample
- 1023
- Source type
- independent-evaluator
- Evaluator
- OnMusic AI Song Generation Study 2026
- Status
- active
- Retrieved at
- 2026-10-21
- Methodology
- Independent evaluation. OnMusic-recorded Lyria 3 Clip (Google) generations, 1,023 events in the reported June 13–October 5, 2026 period; the comparison includes only models with more than 300 generations.
- Context
- Observational platform data, not a controlled laboratory quality benchmark. The exact Google serving revision and separate latency denominator are not reported. Render time includes queue time; the study says 91 system-wide file-retrieval failures were excluded but does not give a model-specific effect.
CharGen Blind Arena overall audio rating (Show measurement, test conditions, and source)
- Source value
- 1045
- Score
- 1045
- Metric
- Bradley-Terry preference rating, overall audio board
- Unit
- rating
- Benchmark version
- Live CharGen overall audio board snapshot captured 2026-10-17; embedded board updatedAt 2026-09-26T13:21:03.484203
- Category
- music-generation
- Direction
- Higher is better
- Sample
- 10
- Lower confidence bound
- 932
- Upper confidence bound
- 1142
- Source type
- official-leaderboard
- Evaluator
- CharGen Blind Arena audio leaderboard
- Status
- active
- Retrieved at
- 2026-10-17
- Methodology
- Official benchmark leaderboard. Overall AUDIO board. Each pair receives the same fantasy or tabletop prompt through the normal generation paths, model names are hidden while signed-in users vote, and ties count as half a win. The source describes a Bradley-Terry fit over counted board votes on a scale centered at 1000, with a pull toward 1000 equivalent to three wins and three losses against an average model. The row reports 10 votes, but does not report a unique-prompt count for this model. The captured row fields ratingLower 932 and ratingUpper 1142 are the source's 95% range for this score, per its methodology's 200-resample rule.
- Context
- The public evaluator label maps unambiguously through Google's official product documentation to Lyria 3 Clip Preview and lyria-3-clip-preview. CharGen does not publish its internal API route or checkpoint revision. The dynamic Bradley-Terry preference rating is not an objective acoustic-quality measure. Ten votes or battles do not establish ten independent tasks. The September 14 article snapshot at 1021 with the same 6-4 record from 10 battles is not duplicated as another row. The lower and upper bounds are copied directly from the published rating fields and the documented 95% range from 200 resamples.
OnMusic: median render time (Show measurement, test conditions, and source)
- Source value
- 19
- Score
- 19
- Metric
- Median render time from request to finished song
- Unit
- seconds
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- OnMusic AI Song Generation Study 2026
- Status
- active
- Retrieved at
- 2026-10-21
- Methodology
- Independent evaluation. OnMusic-recorded Lyria 3 Clip (Google) generations, 1,023 events in the reported June 13–October 5, 2026 period; the comparison includes only models with more than 300 generations.
- Context
- Observational platform data, not a controlled laboratory quality benchmark. The exact Google serving revision and separate latency denominator are not reported. Render time includes queue time; the study says 91 system-wide file-retrieval failures were excluded but does not give a model-specific effect.
OnMusic: 90th-percentile render time (Show measurement, test conditions, and source)
- Source value
- 28
- Score
- 28
- Metric
- 90th-percentile render time from request to finished song
- Unit
- seconds
- Direction
- Lower is better
- Source type
- independent-evaluator
- Evaluator
- OnMusic AI Song Generation Study 2026
- Status
- active
- Retrieved at
- 2026-10-21
- Methodology
- Independent evaluation. OnMusic-recorded Lyria 3 Clip (Google) generations, 1,023 events in the reported June 13–October 5, 2026 period; the comparison includes only models with more than 300 generations.
- Context
- Observational platform data, not a controlled laboratory quality benchmark. The exact Google serving revision and separate latency denominator are not reported. Render time includes queue time; the study says 91 system-wide file-retrieval failures were excluded but does not give a model-specific effect.
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
As of November 2026 · Source: gradually.ai popularity index
Over time
Weekly values from October 18, 2026 to November 6, 2026
Lyria 3 ClipScribe v2 Realtime 29Voxtral Mini Transcribe 2 26Index-Echo-S2TT-2B 26Index-Echo-S2TT-9B 24
Method changes in the timeline, a jump at these points is no market event:
- 1November 1, 2026: Value only with search or usage data, 30-day attention
What is gaining popularity right now, across models, vendors, harnesses and AI subscriptions, is collected in All AI trends.
Similar models
More models of the same kind as Lyria 3 Clip, chosen by model line, developer, size, price and release date.
- Gemini 3.5 TranscribeGoogle · same developer
- Lyria 3.5Google · same developer
- Gemini 3.5 Live TranslateGoogle · same developer
- Gemini 3.1 Flash TTSGoogle · same developer
- Gemini 3.5 Transcribe LiveGoogle · same developer
- Lyria RealtimeGoogle · same developer
- Gemini 3.8 LiveGoogle · same developer
- Gemini 3.8 Live Extended ThinkingGoogle · same developer
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.
| Type | Evidence and data date |
|---|---|
| Research status | Research date October 15, 2026. 9 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication. |
| Unresolved | quality-benchmarks Note CharGen's Blind Arena reports a Bradley-Terry overall-audio preference rating of 1,045, with a 95% interval from 932 to 1,142 based on 10 votes. This is a small-sample preference signal, not an objective acoustic-quality measure; the votes do not establish 10 independent tasks. CharGen does not disclose its internal API route or checkpoint revision. Official Google documentation publishes no numeric quality result, Artificial Analysis has no exact row, and Lyria 3.5 and Lyria 3 Pro remain distinct models. |
| Not found in the inspected sources | quality-benchmarks Note The exact model page provides generation capabilities but no quality score. Current AA music boards have no exact row; Lyria 3.5 and Lyria 3 Pro results are separate checkpoints. |
| Not published | quality-benchmarks Note Google's Lyria 3 model card describes evaluation dimensions and qualitative comparison with Lyria 2, but publishes no numeric quality result or per-metric denominator. It covers Lyria 3 and subsequent versions, not a separately scored lyria-3-clip-preview checkpoint. |
| Unresolved | evaluator-route-and-checkpoint-revision Note The evaluator's public model label maps to Google's exact Lyria 3 Clip Preview product and model code. The evaluator's internal API route and checkpoint revision remain unpublished. |
| Additional source | Google Gemini API model catalog (retrieved September 22, 2026) · Google Gemini API model catalog · Editorial description reviewed October 2, 2026 |
| Additional source | CharGen Blind Arena audio leaderboard (retrieved October 17, 2026) |
| Additional source | OnMusic AI Song Generation Study 2026 (retrieved October 21, 2026) |
| Additional source | Model metadata (retrieved October 15, 2026; October 16, 2026; October 17, 2026) |
| Additional source | Model metadata (retrieved October 16, 2026; October 17, 2026) |
| Additional source | Model metadata (retrieved October 17, 2026)123 |
| Additional source | Gemini API Lyria music generation (retrieved November 5, 2026) |
| Additional source | OpenRouter: Popularity index (derived), CC BY 4.0 |
| Additional source | Vercel AI Gateway: Popularity index (derived), CC BY 4.0 |
| Additional source | Ollama: cumulative downloads for the popularity index |
| Additional source | Replicate: cumulative runs for the popularity index |
| Additional source | Google Trending Now: limited feed selection for Germany and the US |
| Additional source | Creative Commons Attribution 4.0 |