GoogleAudio modelPreview

Gemini 3.1 Flash TTS

PopularityRank 209 of 875

44out of 100same as last month

Summary

Gemini 3.1 Flash TTS Preview turns text into audio and supports expressive audio tags as well as steerable prompts. The API profile allows up to 8,192 input tokens and 16,384 audio output tokens.

Google’s current model catalog labels the endpoint a legacy TTS preview. For new production workloads, Google recommends Gemini 3.8 Flash TTS or Gemini 3.8 Flash-Lite TTS.

All benchmark results

Every published result with its source, test setup and comparison models.

Show all 48 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.

Hume Voice Instruction tone-following accuracy (Show measurement, test conditions, and source)

Source value: 90.2
Score: 90.2
Metric: Accuracy selecting the take that follows a tone instruction
Unit: %
Benchmark version: Hume Voice Controllability leaderboard, tone table, captured 2026-10-17
Category: audio
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Hume AI Real World VoiceEQ
Status: active
Retrieved at: 2026-10-17
Methodology: Independent evaluation. Hume exact table label google/gemini-3.1-flash. Paid blind raters choose which of two takes of the same voice better follows the tone instruction; the source gives 50% as chance accuracy.
Context: The evaluator label omits the API suffix preview and the immutable route version is undisclosed. Hume does not publish per-model N or a confidence interval.

Source Hume AI Real World VoiceEQ

Hume Voice Instruction emotion-identification accuracy (Show measurement, test conditions, and source)

Source value: 67
Score: 67
Metric: Accuracy identifying the intended emotion from six options
Unit: %
Benchmark version: Hume Voice Controllability leaderboard, emotion table, captured 2026-10-17
Category: audio
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Hume AI Real World VoiceEQ
Status: active
Retrieved at: 2026-10-17
Methodology: Independent evaluation. Hume exact table label google/gemini-3.1-flash. Raters identify the emotion from six options; the source gives 16.7% as chance accuracy.
Context: The evaluator label omits the API suffix preview and the immutable route version is undisclosed. Hume does not publish per-model N or a confidence interval.

Source Hume AI Real World VoiceEQ

Hume Voice Controllability role-fit win rate (Show measurement, test conditions, and source)

Source value: 66.9
Score: 66.9
Metric: Blind pairwise listener win rate over decided voice-role-fit votes
Unit: %
Benchmark version: Hume Voice Controllability leaderboard, overall role-fit table, captured 2026-10-17
Category: audio
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Hume AI Real World VoiceEQ
Status: active
Retrieved at: 2026-10-17
Methodology: Independent evaluation. Hume exact table label google/gemini-3.1-flash. Paid blind raters compare two voices made for the same production brief. Score is the share of decided votes for role fit; 50% is even.
Context: The evaluator label omits the API suffix preview. This is a pairwise role-fit result; Hume does not publish per-model votes or a confidence interval.

Source Hume AI Real World VoiceEQ

Hume Voice Controllability naturalness win rate (Show measurement, test conditions, and source)

Source value: 56.1
Score: 56.1
Metric: Blind pairwise listener win rate over decided naturalness votes
Unit: %
Benchmark version: Hume Voice Controllability leaderboard, overall role-fit table, captured 2026-10-17
Category: audio
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Hume AI Real World VoiceEQ
Status: active
Retrieved at: 2026-10-17
Methodology: Independent evaluation. Hume exact table label google/gemini-3.1-flash. Paid blind raters separately judge which voice sounds more natural for the same production brief. Score is the share of decided votes; 50% is even.
Context: The evaluator label omits the API suffix preview. This is a pairwise naturalness result, distinct from role fit; Hume does not publish per-model votes or a confidence interval.

Source Hume AI Real World VoiceEQ

Addis AI Amharic TTS Benchmark (Show measurement, test conditions, and source)

Source value: 11.02
Score: 11.02
Metric: Word error rate (WER), Amharic, primary 87-prompt set, Chirp 3 ASR judge
Unit: %
Category: Text-to-speech
Direction: Lower is better
Sample: 87
Lower confidence bound: 8.94
Upper confidence bound: 13.27
Source type: community-reported
Evaluator: Addis AI Amharic TTS Benchmark
Status: active
Retrieved at: 2026-10-20
Methodology: Community-reported result. Gemini public preview model gemini-3.1-flash-tts-preview, voice Kore, Amharic. Run date 2026-08-12. Prompts were public test-split rows at pinned revisions: 34 FLEURS, 33 Waxal, and 33 Horn-ASR. Primary CER/WER uses 87 non-numeric prompts with a Chirp 3 ASR judge. Blind listening ratings use 100 clips. Latency uses 100 raw public-API requests and is not adjusted for generated clip length.
Context: This is one Amharic-language run over public test data, not a general multilingual TTS ranking. Addis AI publishes its own voice system in the comparison, so these results are publisher-reported and carry a commercial interest; they are not Google-reported or an independent Google evaluation. The page identifies the public-preview model ID and voice, but no immutable serving build.

Source Addis AI Amharic TTS Benchmark

Addis AI Amharic TTS Benchmark (Show measurement, test conditions, and source)

Source value: 3.9
Score: 3.9
Metric: Character error rate (CER), Amharic, strict full 100-prompt set
Unit: %
Category: Text-to-speech
Direction: Lower is better
Sample: 100
Source type: community-reported
Evaluator: Addis AI Amharic TTS Benchmark
Status: active
Retrieved at: 2026-10-20
Methodology: Community-reported result. Gemini public preview model gemini-3.1-flash-tts-preview, voice Kore, Amharic. Run date 2026-08-12. This strict full-set CER includes all 100 prompts, unlike the primary 87 non-numeric prompts. The source reports no confidence interval for this full-set value.
Context: This is one Amharic-language run over public test data, not a general multilingual TTS ranking. Addis AI publishes its own voice system in the comparison, so these results are publisher-reported and carry a commercial interest; they are not Google-reported or an independent Google evaluation. The page identifies the public-preview model ID and voice, but no immutable serving build.

Source Addis AI Amharic TTS Benchmark

Addis AI Amharic TTS Benchmark (Show measurement, test conditions, and source)

Source value: 12.48
Score: 12.48
Metric: Word error rate (WER), Amharic, strict full 100-prompt set
Unit: %
Category: Text-to-speech
Direction: Lower is better
Sample: 100
Source type: community-reported
Evaluator: Addis AI Amharic TTS Benchmark
Status: active
Retrieved at: 2026-10-20
Methodology: Community-reported result. Gemini public preview model gemini-3.1-flash-tts-preview, voice Kore, Amharic. Run date 2026-08-12. This strict full-set WER includes all 100 prompts, unlike the primary 87 non-numeric prompts. The source reports no confidence interval for this full-set value.
Context: This is one Amharic-language run over public test data, not a general multilingual TTS ranking. Addis AI publishes its own voice system in the comparison, so these results are publisher-reported and carry a commercial interest; they are not Google-reported or an independent Google evaluation. The page identifies the public-preview model ID and voice, but no immutable serving build.

Source Addis AI Amharic TTS Benchmark

Addis Voice 2 Amharic TTS, character error rate (Show measurement, test conditions, and source)

Source value: 2.89
Score: 2.89
Metric: Amharic character error rate (CER), primary table excluding prompts with digits; 95% interval 2.17–3.73%
Unit: %
Category: text-to-speech
Direction: Lower is better
Sample: 87
Lower confidence bound: 2.17
Upper confidence bound: 3.73
Source type: community-reported
Evaluator: Addis AI, Amharic TTS Benchmark
Status: disputed
Retrieved at: 2026-10-21
Methodology: Community-reported result. Gemini 3.1 Flash TTS public preview, exact model ID gemini-3.1-flash-tts-preview, voice Kore. Addis AI reports one Amharic run dated 12 August 2026. Its primary CER table uses 87 of 100 public prompts because prompts containing digits are excluded; Chirp 3 is the ASR judge. The page reports 2.89% CER, while the revision-pinned dataset README for the same run reports 2.90%.
Context: Addis AI publishes this evaluation itself. The revision-pinned dataset README for the same run reports 2.90% rather than 2.89%, so the CER value remains disputed. This is an ASR-based intelligibility proxy, not a direct measure of prosody or listening quality.

Source Addis AI, Amharic TTS Benchmark

Cekura TTS benchmark, medical-terms score (Show measurement, test conditions, and source)

Source value: 100
Score: 100
Metric: Cekura label: “Medical terms”; percentage-score definition and denominator are not specified in the captured excerpt
Unit: %
Benchmark version: Cekura TTS benchmark page, last updated 2026-10-19; 26 voices compared
Category: audio
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Cekura Bench
Status: active
Retrieved at: 2026-10-24
Methodology: Independent evaluation. Cekura's text-to-speech page labels the item “Gemini 3.1 Flash” and compares 26 voices. Google's API catalog separately maps “Gemini 3.1 Flash TTS” to model ID gemini-3.1-flash-tts-preview. Cekura does not disclose the request model ID/route, serving revision, language, exact prompt set, metric denominator, sample N, or confidence interval in the captured excerpt.
Context: Google's API catalog lists Gemini 3.1 Flash TTS as the legacy preview model gemini-3.1-flash-tts-preview and recommends newer 3.8 TTS models; this is not a 3.8 result. Cekura does not identify its exact request route or serving revision, nor the score formula, denominator, task sample N, language, prompt set, scorer, or confidence interval in the captured excerpt. The 26 is a voice-comparison cohort, not N. These are Cekura's task-labelled reading percentages, not subjective audio-quality or latency metrics.

Source Cekura Bench

Cekura TTS benchmark, hard-parts-right score (Show measurement, test conditions, and source)

Source value: 87.7
Score: 87.7
Metric: Cekura label: “Reads hard parts right”; percentage-score definition and denominator are not specified in the captured excerpt
Unit: %
Benchmark version: Cekura TTS benchmark page, last updated 2026-10-19; 26 voices compared
Category: audio
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Cekura Bench
Status: active
Retrieved at: 2026-10-24
Methodology: Independent evaluation. Cekura's text-to-speech page labels the item “Gemini 3.1 Flash” and compares 26 voices. Google's API catalog separately maps “Gemini 3.1 Flash TTS” to model ID gemini-3.1-flash-tts-preview. Cekura does not disclose the request model ID/route, serving revision, language, exact prompt set, metric denominator, sample N, or confidence interval in the captured excerpt.
Context: Google's API catalog lists Gemini 3.1 Flash TTS as the legacy preview model gemini-3.1-flash-tts-preview and recommends newer 3.8 TTS models; this is not a 3.8 result. Cekura does not identify its exact request route or serving revision, nor the score formula, denominator, task sample N, language, prompt set, scorer, or confidence interval in the captured excerpt. The 26 is a voice-comparison cohort, not N. These are Cekura's task-labelled reading percentages, not subjective audio-quality or latency metrics.

Source Cekura Bench

Artificial Analysis Provider Voice Arena preference (Show measurement, test conditions, and source)

Source value: 1212
Score: 1212
Metric: blind pairwise native-voice preference Elo
Unit: elo
Benchmark version: Current Provider Voice Arena board retrieved 2026-10-15
Category: audio
Direction: Higher is better
Sample: 4065
Confidence interval: 11
Lower confidence bound: 1201
Upper confidence bound: 1223
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: superseded
Retrieved at: 2026-10-15
Methodology: Independent evaluation. Blind pairwise preference on provider-native voice set (7 voices); 4,065 board appearances; model release shown as Apr 2026.
Context: Appearances are not unique listeners. The current page does not state a per-model evaluation date range.

Source Artificial Analysis

Artificial Analysis Provider Voice Arena preference (Show measurement, test conditions, and source)

Source value: 1213
Score: 1213
Metric: blind pairwise preference Elo for provider-native voices
Unit: elo
Benchmark version: Current Provider Voice Arena board retrieved 2026-10-16
Category: audio
Direction: Higher is better
Sample: 4107
Confidence interval: 11
Lower confidence bound: 1202
Upper confidence bound: 1224
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: superseded
Retrieved at: 2026-10-16
Methodology: Independent evaluation. Exact board row Gemini 3.1 Flash TTS; blind pairwise preference across seven provider-native voices; 4,107 board appearances. The displayed April 2026 release date is retained from the board.
Context: Appearances are not unique listeners. The board does not state a per-model evaluation date range.

Source Artificial Analysis

Page 1 of 4

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-173

Nemotron-3-Diarization57

Gemini 3.5 Transcribe52

Confucius4-R2T252

Granite Speech 5.0 470M TurboCTC52

Kokoro-82M51

Nova-351

Fish Audio S2 Pro51

NVIDIA Nemotron 3.5 Transcribe50

Granite Speech 5.0 470M TurboCTC NC50

Orukeet49

Lyria 3.547

Gemini 3.5 Live Translate46

Phonon-244

Aura-244

Gemini 3.1 Flash TTS44

MiniMax-Music341

Audio8-ASR-Infinite41

GPT-Realtime 2.141

Eleven v339

pyannote Community-1

Nemotron-3-Diarization

Gemini 3.5 Transcribe

Confucius4-R2T2

Granite Speech 5.0 470M TurboCTC

Kokoro-82M

Nova-3

Fish Audio S2 Pro

NVIDIA Nemotron 3.5 Transcribe

Granite Speech 5.0 470M TurboCTC NC

Orukeet

Lyria 3.5

Gemini 3.5 Live Translate

Gemini 3.1 Flash TTS

As of November 2026 · Source: gradually.ai popularity index

Over time

Weekly values from October 18, 2026 to November 5, 2026

Gemini 3.1 Flash TTSAura-2Phonon-2Gemini 3.5 Live TranslateMiniMax-Music3

All specifications

Architecture

Model class: Text-to-speech; Source: Model class

Input and output

Input: Text; Source: Input
Output: Audio; Source: Output
Capabilities: Expressive audio tags, steerable prompts, and multilingual speech generation; Source: Capabilities

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 marks generated audio with the inaudible SynthID watermark.; Source: AI watermarking
Watermarking technique: SynthID; Source: Watermarking technique
Labelling reviewed: Nov 5, 2026; Source: Labelling reviewed

Licence and access

Access: API; Source: Access

API

API model ID: gemini-3.1-flash-tts-preview; Source: API model ID

Further details

Task: Speech generation; Source: Task
Supported details: Steerable prompts and tags; Source: Supported details
Text input: Up to 8,192 tokens; Source: Text input
Audio output limit: Up to 16,384 tokens; Source: Audio output limit

Sources and data date

Show all 36 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. 32 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not published	evaluator-route-revision-denominator-and-confidence-interval Note Hume publishes scores under google/gemini-3.1-flash, not the exact Google API model code. Google's exact page confirms Gemini 3.1 Flash TTS preview as a text-to-speech product, but Hume does not disclose an immutable route revision, per-model denominator or confidence intervals.
Not published	tts-serving-revision-and-metric-denominator Note Serving revision, complete generation controls and metric-specific sample count are not published.
Additional source	Google Gemini API model catalog (retrieved September 24, 2026) · Google Gemini API model catalog · Editorial description reviewed October 2, 2026
Additional source	Artificial Analysis (retrieved October 15, 2026; October 16, 2026)
Additional source	Hume AI Real World VoiceEQ (retrieved October 17, 2026)1 2
Additional source	Speko independent voice benchmark (retrieved October 17, 2026)
Additional source	VoiceArena TTS Leaderboard (retrieved October 17, 2026)
Additional source	Speko TTS benchmark (retrieved October 17, 2026)
Additional source	Addis AI Amharic TTS Benchmark (retrieved October 20, 2026; October 21, 2026)
Additional source	Cekura Bench (retrieved October 24, 2026)
Additional source	Artificial Analysis (retrieved October 28, 2026; October 29, 2026; November 3, 2026)
Additional source	DesignArena (retrieved October 28, 2026)1 2
Additional source	VoiceArena primary public leaderboard (retrieved October 28, 2026)
Additional source	Model metadata (retrieved October 15, 2026; October 16, 2026; October 17, 2026) · Google Gemini 3.1 Flash TTS model documentation · Editorial description reviewed October 2, 2026
Additional source	Model metadata (retrieved October 17, 2026)1 2 34 (unavailable on November 2, 2026) https://voicearena.com/assets/Methodology-CJGHT0Ur.js
Additional source	Model metadata (retrieved October 20, 2026)
Additional source	Model metadata (retrieved October 28, 2026)1 2 3 4 5 6 7 8 9 10
Additional source	Google Gemini 3.1 Flash TTS announcement (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