AI news, with the context that matters.

Products & Services · ·

Gemini 3.8 Flash TTS arrives: A comparison with GPT-Live

Compare Gemini 3.8 Flash TTS and GPT-Live with response speed as the priority. We examine first audio versus substantive answers, language support, speech evaluations across seven languages, API pricing, differing measurement conditions, and Google's 2027 price changes.

On September 23, Google announced Gemini 3.8 Flash TTS and Flash-Lite TTS. For conversation and read-aloud applications, the first question is how long users wait to hear speech. A comparison with GPT-Live needs to distinguish first audio, the start of a substantive answer, and interruption handling. We also compare language support, perceived speech quality, and pricing from published sources.

Key points

  1. Response measurements use different conditions. GPT-Live's measured median onset in an English app test was 1.099 seconds, but this does not establish response times in other languages, API performance, or a head-to-head result against Gemini.
  2. Google lists 130 languages for Flash and 101 for Lite. Flash leads their Japanese listening comparison, while Lite leads in English and some other languages.
  3. Flash TTS currently charges $0.0135 per generated audio minute; GPT-Live charges $0.05 per active session minute. These cover different usage, and Google's prices change in 2027.
Official announcement image for Gemini 3.8 Flash TTS and 3.8 Flash-Lite TTS
Source: Google

What happened

  • Google's September 23 announcement introduced Flash TTS for expressive speech and Flash-Lite TTS for high-volume, low-latency use.
  • Availability includes the Gemini API and Google AI Studio. This article compares the paid Standard tier of the Gemini Developer API with OpenAI API pricing.
  • This article compares published specifications and reported evaluations. LATENT has not conducted its own audio measurements or listening tests.

Primary sources: Google's announcement, GPT-Live specifications

Measure first audio separately from the start of an answer

When speed matters, first align what is being timed. TTS onset is the delay between sending text and receiving the first audio. Conversational response time runs from the end of the user's speech to audible output, and can include turn-end detection, recognition, response generation, networking, and playback delays.

A quick acknowledgment such as 'Let me check' may arrive before the answer to the question. Measuring time to first audio, time until a meaningful answer begins, and time until the answer finishes separately avoids confusing conversational pace with task completion speed. OpenAI also recommends comparing time to a useful spoken response and task success on the same scenarios when optimizing GPT-Live's backend. GPT-Live latency optimization

Published seconds do not form a speed ranking under matched conditions

Response-time figures have been published, but statistics from a Gemini TTS serving provider and measurements from a smartphone app using GPT-Live start and stop the clock at different points.

Published response measurements and their scope
TargetReported valueWhat was measured
Gemini 3.8 Flash TTSLatency P50
3.47 seconds
Google AI Studio through OpenRouter; displayed for all locations over one week
Gemini 3.8 Flash-Lite TTSLatency P50
2.66 seconds
Same display scope; not a test with matched scripts and languages
GPT-Live 1
ChatGPT app
Median 1.099 seconds
P90 1.203 seconds
End of English user speech to first audible output, measured by Agora

ConditionsOpenRouter's Flash and Lite values checked September 30. P50 is the median; P90 indicates a value at or below which approximately 90% of observations fall. GPT-Live was measured by Agora on July 9. These conditions cannot establish cross-model speed ratios or the fastest model.

GPT-Live's roughly one-second result was measured in an English app test

Agora tested an iPhone 13 running ChatGPT v1.2026.183 on a Plus account, in English, with 30 trials per condition. First audio includes acknowledgments, so the substantive answer need not begin at 1.099 seconds. Median time to stop speaking after an interruption was 1.399 seconds. This is a July English-language snapshot from one device, location, and account. Response times in other languages were not measured, and this result does not guarantee API performance in any language, including English. Agora also sells voice communication services. Methods and limitations

OpenRouter's figures aggregate observed latency on that routing path. They do not mean a direct API call from Japan will always take 2.66 or 3.47 seconds from text submission to audible playback. They are not a comparison with matched audio lengths and streaming conditions; relative first-audio performance in Japanese remains unverified.

If conversational responsiveness is the priority, we would first try a setup that uses GPT-Live for the full conversation and one that uses Flash-Lite TTS for speech output. This choice follows their product roles as full-duplex conversation and low-latency TTS, rather than a measured speed ranking. Google positions Lite for low latency and high throughput, and both TTS models support streaming. OpenAI's dedicated TTS also supports streaming and recommends WAV or PCM when response speed is a priority. Lite's positioning / Gemini streaming / OpenAI TTS output formats

Language coverage and natural speech quality are different claims

Google lists 130 languages for Flash and 101 for Lite. Both lists include Japanese, English, Chinese, Cantonese, Korean, French, German, Spanish, Portuguese, Arabic, Hindi, and Vietnamese. Language and script variants also appear in the lists, so counts alone do not establish a quality difference between providers.

Flash and Lite also differ in language coverage
ModelCoverage confirmed in official documentationLimits when assessing naturalness
Gemini 3.8 Flash TTSGoogle lists 130 languages, including Japanese and English as well as Thai, Finnish, and SwedishListening preference results for seven languages appear below
Gemini 3.8 Flash-Lite TTSGoogle lists 101 languages; the three additional languages above were absent from the Lite list reviewedThe count does not guarantee naturalness across every supported language
gpt-4o-mini-ttsPublishes a language list including Japanese, English, Chinese, Korean, and ThaiVoices are optimized for English
GPT-Live 1Documents language prompting; no exhaustive language list or total was found in the API documentation reviewedThe dedicated TTS language list cannot be assumed to apply to GPT-Live

ConditionsFlash language list / Lite language list / OpenAI TTS languages / GPT-Live language prompting, checked September 30.

Voice design and line-by-line direction in one production workflow

Flash TTS can create a voice from a natural-language description, then change emotion and delivery speed for individual lines while retaining that voice. It supports two-speaker scripts, laughter, sighs, and backchannel responses. One example is a character explaining something calmly in one scene and reacting with surprise in the next. Google's announcement

The official specifications position Flash TTS for voice fidelity, acting, and regional accents, and Flash-Lite TTS for high-volume read-aloud use and the speech output stage of voice agents. They support 130 and 101 languages respectively, including Japanese. Both share the same API schema, making it easier to switch for different workloads. Flash TTS specifications / Flash-Lite TTS specifications

Voice replication is also available, with a process that verifies the speaker's recorded consent. The announcement also mentions Voice remixing for adjusting an existing voice's pitch and timbre, but describes it as coming soon. It should not be confused with features already available. Available and planned features

Japanese speech evaluations outperform OpenAI's dedicated TTS model

Google published evaluation results from Hume AI and Voice Arena. Hume AI assesses multiple dimensions of speech, including naturalness and expressiveness. Voice Arena calculates Elo ratings from blind pairwise listening preferences. Google's methodology says Gemini was tested using production checkpoints, default sampling settings, and single-attempt generation. Evaluation methodology and results

Seven-language listening evaluations show different strengths for Flash and Lite
Metric or languageGemini 3.8
Flash TTS
Gemini 3.8
Flash-Lite TTS
OpenAI
gpt-4o-mini-tts
Hume AI overall quality index0.9200.9140.740
Voice Arena Japanese12321152975
Voice Arena English10611087940
Brazilian Portuguese11041134946
Vietnamese11351156839
Modern Standard Arabic12041181911
Hindi11061076843
Mexican Spanish11521146880

ConditionsThird-party evaluations reproduced in Google's September 2026 materials. Language rows use Voice Arena Elo; higher scores indicate stronger listening preferences. These do not isolate naturalness alone and may reflect pronunciation and voice preferences. Hume AI's index uses a different scale, and scores should not be compared across languages. GPT-Live is not included.

These results do not establish an audio-quality winner against GPT-Live

Flash TTS scores above Flash-Lite TTS and gpt-4o-mini-tts in Japanese. In English, however, Lite scores above Flash, so the more expensive model does not always lead. Elo measures listening preferences: 1232 versus 975 does not mean Japanese capability is approximately 26% better.

Within these seven languages, Flash scores above Lite in Japanese, Modern Standard Arabic, Hindi, and Mexican Spanish. Lite scores above Flash in English, Brazilian Portuguese, and Vietnamese. Chinese, Korean, French, and other languages are absent from this published table: support can be confirmed without establishing a naturalness ranking.

The OpenAI model compared in this evaluation is gpt-4o-mini-tts, not GPT-Live 1. The materials reviewed contain neither scores for Gemini 3.8 TTS and GPT-Live under the same conditions nor side-by-side measurements of their response latency. The table cannot establish which is better than GPT-Live or how many times faster either is.

The published charts also do not specify per-model vote counts, confidence intervals, or detailed OpenAI model versions. Hume AI itself explains that naturalness, expressiveness, identity stability, and reliability should be assessed separately. These results can help select a Japanese narration model, but do not guarantee pronunciation of proper nouns or stability across long scripts. Hume AI's evaluation framework

GPT-Live keeps conversing while listening to the user

GPT-Live 1 is a voice conversation model that can listen and speak simultaneously. It handles conversations in which users add details or correct themselves, and can delegate reasoning and tool execution to another model or agent. Suitable uses include changing booking requirements during a conversation or continuing to talk while information is being retrieved.

Gemini TTS handles the stage that turns supplied text into audio. Streaming lets playback begin before generation finishes, but it does not include understanding incoming speech and deciding how to respond. A conversational application needs separate speech recognition, response generation, and handling such as stopping playback when the user interrupts. Gemini speech generation guide

Reading a script and holding a live conversation require different features
ComparisonGemini 3.8 TTSGPT-Live 1
Core taskGenerate speech from textTwo-way voice conversation
Voice controlVoice design and script directionInstructions for conversational style
When the user starts speakingThe app combines recognition and playback interruption handlingListening and speaking are handled simultaneously
Reasoning and external toolsCombine with another model or application logicDelegate to a backend agent

ConditionsFeature comparison based on Google's specifications and OpenAI's GPT-Live guide. This is not a measured ranking of audio quality or latency.

For reading a script aloud, include gpt-4o-mini-tts in the comparison

OpenAI, the company behind ChatGPT, also offers a dedicated text-to-speech API separate from GPT-Live. Its Text to speech guide describes gpt-4o-mini-tts, which generates audio from text and a selected voice, with instructions for speed, emotion, intonation, and whispering. It also supports streaming. Its voices are optimized for English, so Japanese users should evaluate it with their own scripts.

Per-request input limits also differ: 2,000 tokens for gpt-4o-mini-tts and 8,192 for Gemini 3.8 TTS. Google's developer specifications list an audio output serving limit of 16,384 tokens, equivalent to about 10.9 minutes at 25 tokens per second. Producing a long audiobook therefore requires splitting the script and checking for changes in voice or volume at the boundaries. OpenAI model specifications / Gemini model specifications / Audio token conversion

Gemini Flash TTS audio output currently costs $0.0135 per minute

In the Gemini Developer API's paid Standard tier, Flash TTS audio output costs $9 per million tokens and Flash-Lite TTS costs $6. Audio is billed at 25 tokens per second, or 1,500 tokens per minute. The following calculations cover audio output only; the input script and instructions cost extra. Official pricing

Gemini TTS prices change in January 2027
Model and periodInput text
per million tokens
Audio output
per million tokens
Audio output
calculated per minute
Flash TTS
Through December 31, 2026
$0.50$9$0.0135
Flash-Lite TTS
Through December 31, 2026
$0.50$6$0.009
Flash TTS
From January 1, 2027
$1$18$0.027
Flash-Lite TTS
From January 1, 2027
$1$12$0.018

ConditionsGemini Developer API prices, checked September 30. Standard pricing in US dollars; excludes free allowances, Batch, Flex, Priority, and caching. Per-minute calculation: audio token price x 25 x 60 / 1,000,000.

OpenAI's dedicated TTS costs $0.60 for input and $12 for audio output

The official gpt-4o-mini-tts specifications list $0.60 per million input text tokens and $12 per million audio output tokens. For example, 100,000 input tokens plus one million audio output tokens would cost $0.06 + $12 = $12.06.

This calculation does not compare equal lengths of generated audio. Providers tokenize audio differently, so Google's 25 tokens per second cannot be applied to OpenAI. The OpenAI pricing and model pages reviewed here do not provide a per-minute conversion under equivalent conditions, so we have not converted gpt-4o-mini-tts into a uniform per-minute rate. A precise speech-generation cost comparison requires measuring the length and actual usage of audio generated from the same script.

A GPT-Live minute measures session time, not generated audio length

GPT-Live 1 costs $0.05 per voice session minute, billed per second. According to OpenAI's cost guide, this includes user speech, assistant speech, silence, and time spent waiting for the backend. Muting the microphone does not end the session. Backend reasoning models and tools are billed separately.

Suppose a 10-minute conversation includes three minutes of assistant speech. GPT-Live's session fee would be $0.50. Generating three minutes of audio with Gemini TTS at current prices would cost $0.0405 for Flash or $0.027 for Lite in audio output fees. The latter figures exclude speech recognition, response generation, input text, and conversation management. The difference cannot simply be treated as savings for an entire voice application.

The more audio generated, the larger the difference caused by the price change. Output charges for 1,000 minutes of generated audio are $13.50 for Flash and $9 for Lite during 2026. From 2027, they rise to $27 and $18 respectively. GPT-Live costs $50 for 1,000 session minutes, plus backend charges. Budgeting requires separate totals for minutes of generated audio and minutes of connected conversation.

Measure conversational delay first, then compare quality and cost

For Japanese narration or character dialogue, the published results and direction features give a reason to try Flash TTS first. For large volumes of standard messages or announcements, compare the same script with the lower-output-cost Flash-Lite TTS and decide whether the quality difference matters. When adding read-aloud features to an OpenAI-based application, include gpt-4o-mini-tts as well.

For responsiveness, test the same short and long Japanese utterances and time first audio, substantive-answer onset, and stopping after an interruption. Check P90 as well as the median to identify occasional long waits. Once candidates meet the required speed on the intended device and network, compare pronunciation errors, naturalness, and costs including retries.

Sources and references

Prices and specifications checked September 30, 2026. Evaluation scores come from Google's September 2026 materials.