Gemini 3.8 Flash TTS Text to Speech API

google/gemini-3.8-flash/text-to-speech

Gemini 3.8 Flash TTS Text to Speech turns scripts and two-speaker dialogues into WAV speech across 130 languages with 30 preset voices. Give the whole request and each line its own style instructions, so delivery follows your direction while the transcript is read as written.

Input

Request form

Text

Voice

Style Instructions

266 billable characters

Enable Safety Checker

Output

Explore more speech models

Gemini 3.8 Flash Lite TTS Text to Speech Gemini 3.1 Flash TTS Text to Speech ElevenLabs V4 Text to Speech

Examples

Mandarin museum audio guide

Mandarin museum audio guide — A Mandarin guide script; the language is detected from the text.

Gemini 3.8 Flash TTS Text to Speech API

Generate speech with the API

Submit a script, keep the returned task_id, and retrieve the completed audio through the task status endpoint.

Authenticate your request

Send your Vidgo API key in the Authorization header using Bearer authentication.

Endpoint: POSThttps://api.vidgo.ai/api/generate/submit
Authentication: Authorization: Bearer YOUR_API_KEY

Submit the script or dialogue

Send model and input as JSON. The example below uses the first audio sample on this page.

Request example

REQUEST_BODY=$(cat <<'JSON'
{
  "model": "google/gemini-3.8-flash/text-to-speech",
  "input": {
    "text": "The snow had been falling for three days when the letter arrived. It carried no stamp, no return address, only my name, written in my grandmother's careful hand.",
    "voice": "Charon",
    "style_instructions": "Read like a seasoned audiobook narrator: unhurried, low and warm, with a short pause after each sentence.",
    "enable_safety_checker": true
  }
}
JSON
)

RESPONSE=$(curl --silent --show-error --fail-with-body \
  --request POST \
  --url "https://api.vidgo.ai/api/generate/submit" \
  --header "Authorization: Bearer $VIDGO_API_KEY" \
  --header "Content-Type: application/json" \
  --data "$REQUEST_BODY")

CODE=$(printf '%s' "$RESPONSE" | jq -r '.code // empty')
if [ "$CODE" != "0" ] && [ "$CODE" != "200" ]; then
  printf 'API error: %s
' "$RESPONSE" >&2
  exit 1
fi
printf '%s
' "$RESPONSE"

Follow the task

Query the returned task_id until status is finished or failed.

Status endpoint

GET https://api.vidgo.ai/api/generate/status/{task_id}

Keep the task_id from submission. Check its status and read the generated files when it finishes.

not_startedrunningfinishedfailed

Submission response

{
  "code": 200,
  "data": {
    "task_id": "3LCLZTC34PBO54EP",
    "status": "not_started",
    "created_time": "2026-11-05T07:14:43"
  }
}

Completed response

{
  "code": 200,
  "data": {
    "task_id": "3LCLZTC34PBO54EP",
    "status": "finished",
    "files": [
      {
        "file_url": "https://cdn.vidgo.ai/apis/models/google/gemini-3.8-flash/text-to-speech/v2/02/speech-01.wav",
        "file_type": "audio",
        "file_name": "speech-01.wav"
      }
    ],
    "created_time": "2026-11-05T07:14:43",
    "error_message": null,
    "progress": 100
  }
}

Read the audio files

Use the audio file_url to play or download the WAV file.

Read the audio files

set -euo pipefail
: "${VIDGO_API_KEY:?Set VIDGO_API_KEY in your environment}"

REQUEST_BODY=$(cat <<'JSON'
{
  "model": "google/gemini-3.8-flash/text-to-speech",
  "input": {
    "text": "The snow had been falling for three days when the letter arrived. It carried no stamp, no return address, only my name, written in my grandmother's careful hand.",
    "voice": "Charon",
    "style_instructions": "Read like a seasoned audiobook narrator: unhurried, low and warm, with a short pause after each sentence.",
    "enable_safety_checker": true
  }
}
JSON
)

SUBMIT_RESPONSE=$(curl --silent --show-error --fail-with-body \
  --request POST \
  --url "https://api.vidgo.ai/api/generate/submit" \
  --header "Authorization: Bearer $VIDGO_API_KEY" \
  --header "Content-Type: application/json" \
  --data "$REQUEST_BODY")

TASK_ID=$(printf '%s' "$SUBMIT_RESPONSE" | jq -r '.data.task_id // .task_id // empty')
BUSINESS_CODE=$(printf '%s' "$SUBMIT_RESPONSE" | jq -r '.code // empty')
if [ "$BUSINESS_CODE" != "0" ] && [ "$BUSINESS_CODE" != "200" ]; then
  printf 'Submit failed:
%s
' "$SUBMIT_RESPONSE" >&2
  exit 1
fi
if [ -z "$TASK_ID" ]; then
  printf 'Submit response did not include task_id:
%s
' "$SUBMIT_RESPONSE" >&2
  exit 1
fi

START_TIME=$(date +%s)
POLL_DELAY=2
while true; do
  if [ $(( $(date +%s) - START_TIME )) -ge 600 ]; then
    printf 'Timed out after 600 seconds
' >&2
    exit 1
  fi
  STATUS_RESPONSE=$(curl --silent --show-error --fail-with-body \
    --url "https://api.vidgo.ai/api/generate/status/$TASK_ID" \
    --header "Authorization: Bearer $VIDGO_API_KEY")
  STATUS=$(printf '%s' "$STATUS_RESPONSE" | jq -r '.data.status // .status // empty')
  BUSINESS_CODE=$(printf '%s' "$STATUS_RESPONSE" | jq -r '.code // empty')
  if [ "$BUSINESS_CODE" != "0" ] && [ "$BUSINESS_CODE" != "200" ]; then
    printf 'Status request failed:
%s
' "$STATUS_RESPONSE" >&2
    exit 1
  fi

  case "$STATUS" in
    finished)
      printf '%s' "$STATUS_RESPONSE" | jq -r '(.data.files // .files // [])[]?.file_url'
      break
      ;;
    failed)
      printf '%s' "$STATUS_RESPONSE" | jq -r '.data.error_message // .error_message // "Generation failed"' >&2
      exit 1
      ;;
    not_started|running)
      sleep "$POLL_DELAY"
      if [ "$POLL_DELAY" -lt 10 ]; then POLL_DELAY=$((POLL_DELAY + 1)); fi
      ;;
    *)
      printf 'Unexpected task status: %s
' "$STATUS" >&2
      exit 1
      ;;
  esac
done

Request parameters

Place speech controls inside input; keep model and callback_url at the top level.

Field	Type	Required	Default	Description
model	string	Yes	—	Use google/gemini-3.8-flash/text-to-speech to select this speech model.
input	object	Yes	—	Use text and voice for one speaker, or speakers and turns for a dialogue, plus the optional fields below.
input.text	string	Single speaker	—	Single speaker: the script to speak. Write the words as they should be heard; place acting directions in Style Instructions.
input.voice	string	Optional	Kore	Single speaker: one of the 30 preset voices. Default Kore. Voices: Achernar, Achird, Algenib, Algieba, Alnilam, Aoede, Autonoe, Callirrhoe, Charon, Despina, Enceladus, Erinome, Fenrir, Gacrux, Iapetus, Kore, Laomedeia, Leda, Orus, Pulcherrima, Puck, Rasalgethi, Sadachbia, Sadaltager, Schedar, Sulafat, Umbriel, Vindemiatrix, Zephyr, Zubenelgenubi.
input.speakers	array	Dialogue	—	Dialogue: exactly two speakers, each with a unique speaker_id and one of the 30 preset voices.
input.speakers[].speaker_id	string	Per item	—	A unique, non-empty name for the speaker, such as Host.
input.speakers[].voice	string	Per item	—	The speaker's preset voice. Voices: Achernar, Achird, Algenib, Algieba, Alnilam, Aoede, Autonoe, Callirrhoe, Charon, Despina, Enceladus, Erinome, Fenrir, Gacrux, Iapetus, Kore, Laomedeia, Leda, Orus, Pulcherrima, Puck, Rasalgethi, Sadachbia, Sadaltager, Schedar, Sulafat, Umbriel, Vindemiatrix, Zephyr, Zubenelgenubi.
input.turns	array	Dialogue	—	Dialogue: the lines in order. Each turn names a speaker_id from Speakers, its text, and optional style_instructions for that line.
input.turns[].speaker_id	string	Per item	—	The speaker_id of the speaker who says this line.
input.turns[].text	string	Per item	—	The words of this line, not blank.
input.turns[].style_instructions	string	Optional	—	Optional direction for this line only.
input.style_instructions	string	Optional	—	Optional delivery direction for the whole request, such as tone, pace or setting. Its characters count toward billing.
input.enable_safety_checker	boolean	Optional	true	Runs the upstream safety checker on the request. Default true.
callback_url	string	No	—	Optional HTTP or HTTPS URL for task notifications. Include it at the top level of the request.

Response fields

Submission identifies the task. Status responses report progress toward the completed audio.

Field	Type	Description
code	integer	Response code. A successful response uses 200 or 0.
data.task_id	string	Task identifier used to query status.
data.status	string	Task state: not_started, running, finished, or failed.
data.created_time	string	Task creation time returned by the service.
data.files	array	Generated files; the audio entry holds the WAV speech.
data.files[].file_url	string	URL used to play or download a generated file.
data.files[].file_type	string	audio for the generated speech.
data.files[].file_name	string	File name, when returned.
data.error_message	string \| null	Error details when a task fails.

Task lifecycle

Use the task state to decide when to check again or read the result.

not_started
The task has been accepted and is waiting to start.
running
Speech generation is in progress. Continue checking the same task.
finished
Generation is complete. Read files for the WAV audio.
failed
Generation failed. Read error_message before deciding whether to submit a new request.

Handle results and interruptions

Keep the current task identifierIf a status request is interrupted, query the same task_id again to recover its latest state.
Match each file to its purposeRead the audio entry in files and use its file_url and file_name for download.
Inspect a failed taskWhen status is failed, read error_message and review the script and settings before submitting another generation.

Speech specifications

Item	Value	Description
Task	Text to Speech	Single-speaker speech or a two-speaker dialogue.
Voices	30 presets	Kore by default; one voice per speaker in dialogue.
Speakers	1 or 2	Dialogue uses exactly two speakers and one or more turns.
Languages	130	Detected automatically from the transcript.
Output	WAV	One audio file per task.
Price	$0.045 / 1K characters	9 credits per 1,000 billable characters.

View the Gemini 3.8 Flash TTS Text to Speech API docs

Gemini 3.8 Flash TTS Text to Speech

Gemini 3.8 Flash TTS Text to Speech is a Google model from September 2026 that Google describes as its flagship creative speech tier, aimed at voice fidelity, acting nuance and dialect coverage. It detects the language of each script across 130 languages and voices it with one of 30 presets, or performs a two-speaker conversation turn by turn. Style instructions carry tone, pace and setting, which suits audiobooks, podcasts and narrated video.

Why choose Gemini 3.8 Flash TTS?

Two voices in one requestSend two speakers and a list of turns to produce a podcast segment or interview as a single audio file.
Direction kept apart from the transcriptGlobal and per-turn style instructions set tone and pace, while the text itself is read exactly as written.
130 languages detected automaticallyWrite the script in the target language; no language code is needed.
30 preset voicesChoose voices such as Kore, Puck, Charon or Aoede for narrators and characters.

Speech parameters

Parameter	Requirement	What it controls
text	Single speaker	Single speaker: the script to speak. Write the words as they should be heard; place acting directions in Style Instructions.
voice	Optional	Single speaker: one of the 30 preset voices. Default Kore. Default`KoreAchernarAchirdAlgenibAlgiebaAlnilamAoedeAutonoeCallirrhoeCharonDespinaEnceladusErinomeFenrirGacruxIapetusLaomedeiaLedaOrusPulcherrimaPuckRasalgethiSadachbiaSadaltagerSchedarSulafatUmbrielVindemiatrixZephyrZubenelgenubi`
speakers	Dialogue	Dialogue: exactly two speakers, each with a unique speaker_id and one of the 30 preset voices.
speakers[].speaker_id	Per item	A unique, non-empty name for the speaker, such as Host.
speakers[].voice	Per item	The speaker's preset voice. `AchernarAchirdAlgenibAlgiebaAlnilamAoedeAutonoeCallirrhoeCharonDespinaEnceladusErinomeFenrirGacruxIapetusKoreLaomedeiaLedaOrusPulcherrimaPuckRasalgethiSadachbiaSadaltagerSchedarSulafatUmbrielVindemiatrixZephyrZubenelgenubi`
turns	Dialogue	Dialogue: the lines in order. Each turn names a speaker_id from Speakers, its text, and optional style_instructions for that line.
turns[].speaker_id	Per item	The speaker_id of the speaker who says this line.
turns[].text	Per item	The words of this line, not blank.
turns[].style_instructions	Optional	Optional direction for this line only.
style_instructions	Optional	Optional delivery direction for the whole request, such as tone, pace or setting. Its characters count toward billing.
enable_safety_checker	Optional	Runs the upstream safety checker on the request. Default true. Default`truefalse`

Create speech in four steps

Choose the request formSelect Single speaker for one voice, or Dialogue for two speakers and their turns.
Write text or turnsEnter the transcript as it should be heard. For dialogue, assign each turn to a speaker.
Add style instructionsDescribe tone, pace or setting for the whole request, and add a short direction to individual turns if needed.
Run and download WAVCheck the cost on the Run button, select Run and download the WAV file when the task finishes.

Character-based pricing

Gemini 3.8 Flash TTS Text to Speech costs $0.045, or 9 credits, per 1,000 billable characters. Billable characters are Text plus Style Instructions, and in dialogue every turn's text and style instructions; speaker IDs are not counted.

Text length	Cost	Calculation
500 characters	$0.0225 · 4.5 credits	500 × 9 / 1000 = 4.5 credits
1,000 characters	$0.045 · 9 credits	1000 × 9 / 1000 = 9 credits
10,000 characters	$0.45 · 90 credits	10000 × 9 / 1000 = 90 credits

Where it fits

Two-host podcast segmentsScript a conversation between two hosts and render it as one WAV file.
Audiobook narrationUse a single narrator voice with a measured style direction for chapter openings and long passages.
Multilingual course voiceoversNarrate lessons in any of 130 languages with the language detected from the text.
Directed product videosDescribe the mood once in style instructions and keep it consistent across scenes.

Pro tips

Keep acting directions in Style Instructions; Google notes that Gemini 3.8 TTS reads the transcript literally.
Give dialogue speakers short, distinct IDs such as Host and Guest, and match each turn to one of them.
Use a per-turn style instruction only where a line needs a different emotion from the global direction.
Pick contrasting voices for two speakers, for example Kore with Puck, so listeners can follow who is talking.
Move a script to Gemini 3.8 Flash Lite TTS for high-volume runs by changing only model.

Before you publish

Audio is returned as a WAV file. Convert it in your audio tool if your player needs MP3 or Opus.
Google sets an input limit of 8,192 tokens for this model; split very long scripts into several requests.

Explore more speech models

Gemini 3.8 Flash Lite TTS Text to Speech preview

Gemini 3.8 Flash Lite TTS Text to SpeechLower-cost speech with the same voices and dialogue format.

Gemini 3.1 Flash TTS Text to Speech preview

Gemini 3.1 Flash TTS Text to SpeechThe previous Gemini speech model with output format and language options.

ElevenLabs V4 Text to SpeechExpressive speech with audio tags, 19 formats and timestamps.

Gemini 3.8 Flash TTS Text to Speech API frequently asked questions

What is the Gemini 3.8 Flash TTS Text to Speech API?

Gemini 3.8 Flash TTS Text to Speech is a Google model for generating natural speech from text. It produces single-voice narration or two-speaker dialogue in 130 languages with 30 preset voices and style instructions for tone and pace. Google built the 3.8 Flash tier for voice fidelity and acting nuance, and it reads the transcript as written while following your direction. You can call it programmatically or try it from the playground above.

How do I create a two-speaker dialogue with Gemini 3.8 Flash TTS?

Choose Dialogue and send exactly two speakers, each with a speaker_id and a preset voice, plus a turns list. Every turn names one speaker_id, its text and an optional style_instructions. Top-level text and voice are left out in this form.

Where should acting directions go in Gemini 3.8 Flash TTS?

Put them in style_instructions, either for the whole request or for a single turn. Google notes that Gemini 3.8 TTS reads the transcript literally, so directions written inside the text may be spoken aloud.

Which voices does Gemini 3.8 Flash TTS offer?

There are 30 presets: Achernar, Achird, Algenib, Algieba, Alnilam, Aoede, Autonoe, Callirrhoe, Charon, Despina, Enceladus, Erinome, Fenrir, Gacrux, Iapetus, Kore, Laomedeia, Leda, Orus, Pulcherrima, Puck, Rasalgethi, Sadachbia, Sadaltager, Schedar, Sulafat, Umbriel, Vindemiatrix, Zephyr, Zubenelgenubi. Kore is the default for a single speaker.

How does Gemini 3.8 Flash TTS handle languages?

Gemini 3.8 Flash TTS detects the language from the transcript across 130 languages. Write each script in its target language; there is no separate language setting.

Should I choose Gemini 3.8 Flash TTS or Gemini 3.8 Flash Lite TTS?

Choose Gemini 3.8 Flash TTS when acting nuance, dialects and complex dialogue matter most; Google positions it as the flagship creative tier with 130 languages. Gemini 3.8 Flash Lite TTS covers 101 languages at 6 credits per 1,000 characters for high-volume work. Both accept the same fields.

What audio format does Gemini 3.8 Flash TTS return?

Each finished task returns a WAV file in files. Convert it to MP3 or Opus in your audio tool if your player or platform needs a compressed format.

How are Gemini 3.8 Flash TTS characters counted for billing?

Billing counts Text, Style Instructions, and in dialogue every turn's text and style instructions, at 9 credits ($0.045) per 1,000 characters. Speaker IDs are not counted.