AI Audio Generator — Gemini 3.8 TTS, Eleven v4 & 2 more

Back to pipelines

Create your first voiceover

Write what you want to say, or listen to an example.

Complete to generate

Text

Text to convert to speech.

Runs on

5 models

Google
- Gemini 3.8 Flash-Lite TTS
- Gemini 3.8 Flash TTS
MiniMax
- MiniMax Speech 2.8 Turbo
- MiniMax Speech 2.8 HD
ElevenLabs
- Eleven v4

Examples

Audio

Gemini 3.8 Flash TTS

Meet Pyra: voiceover

Gemini 3.8 Flash TTS. Pyra's own story, read in a warm storyteller voice: a voiceover you can drop straight into any video.

Reproduce

How to use Audio Generator

A compact decision guide for placing this pipeline in a workflow.

Best for

Narration and voiceover in 70+ languages
Text-to-speech with natural language style control (e.g. 'speak warmly', 'whisper', 'excited tone')
Multi-voice content with dozens of expressive voices
Multilingual scripts, the language is auto-detected from the text

Tips

Write narration in short, clear sentences for better pacing
Use punctuation to control pauses: periods for full stops, commas for brief pauses, ellipsis for long pauses
Leave the voice on Auto and the AI picks one that matches the style instructions, or pin a specific voice if you want a recurring character
Use natural language style hints: 'speak slowly and dramatically', 'cheerful and upbeat', 'calm and reassuring'

Audio Generator API

Call this pipeline from your own code. One request dispatches a run; the model is an input field, not a separate endpoint.

Get an API token

inputs: 5
required: 1
Pricing: 0.002 to 40.0 credits

Language

TypeScript Python Go CLI

bun add @pipe2-ai/sdk

Run a pipeline

import { createClient } from "@pipe2-ai/sdk";

const client = createClient(process.env.PIPE2_TOKEN!);

const { run_pipeline } = await client.RunPipeline({
  pipeline_slug: "audio-generator",
  input: {
    text: "Welcome back to the show. Today we're talking about something a little strange.",
    model: "gemini-3-8-flash-tts",
  },
});

pip install pipe2

Run a pipeline

import asyncio
import os

from pipe2 import Pipe2Client

client = Pipe2Client(token=os.environ["PIPE2_TOKEN"])

run = asyncio.run(client.run_pipeline(
    pipeline_slug="audio-generator",
    input={
        "text": "Welcome back to the show. Today we're talking about something a little strange.",
        "model": "gemini-3-8-flash-tts",
    },
))

go get github.com/pipe2-ai/sdk-go

Run a pipeline

client := pipe2.NewClient(os.Getenv("PIPE2_TOKEN"))

input := json.RawMessage(`{
  "text": "Welcome back to the show. Today we're talking about something a little strange.",
  "model": "gemini-3-8-flash-tts"
}`)

run, err := pipe2.RunPipeline(context.Background(), client, "audio-generator", input)
if err != nil {
    log.Fatal(err)
}

brew install pipe2-ai/tap/pipe2

Run a pipeline

pipe2 pipelines run \
  --pipeline audio-generator \
  --input-json '{"text": "Welcome back to the show. Today we're talking about something a little strange.", "model": "gemini-3-8-flash-tts"}' \
  --wait

Per page

Input	Type	default	Description
`text`required	string	default—	Text to convert to speech.
`enhance_prompt`	boolean	defaulttrue	Improve the prompt for the selected model before generation. Turn off to send your prompt unchanged.
`instructions`	string	default—	Describe tone, accent, pacing, emotion. The AI crafts a rich style prompt.
`model`	enum	defaultauto	Voice engine to use. auto · gemini-3-8-flash-tts · gemini-3-8-flash-lite-tts +3 `autogemini-3-8-flash-ttsgemini-3-8-flash-lite-ttseleven-v4minimax-speech-2-8-turbominimax-speech-2-8-hd`
`voice`	string	defaultauto	Voice name from dozens available, or auto to let the AI pick.

Point any MCP client at pipe2 and your agent gets these tools. It finds this pipeline, reads the same schema above, then runs it.

list_pipelines
get_pipeline_schema
run_pipeline
get_pipeline_run_status
request_upload

https://mcp.pipe2.ai/mcp

{
  "mcpServers": {
    "pipe2ai": {
      "url": "https://mcp.pipe2.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_TOKEN" }
    }
  }
}

MCP setup

Full API reference CLI reference

Recipes using this pipeline

Show all recipes

Audience question → animated visual answer

Turn your approved answer and visual beats into a narrated vertical explainer with stock footage, timed diagrams, and synchronized captions.

6 pipelines one CLI command

Open recipe
1. 1 Audio Generator
2. 2 Transcription
3. 3 Footage Search
4. 4 Video Reel
5. 5 Text Card
6. 6 Captions
Open recipe

Articles about this pipeline

Show all articles

Frequently Asked Questions

What voices are available?

Dozens of expressive voices spanning bright, deep, warm, firm, and dramatic characters. Each responds to plain-language style instructions for tone and delivery.

What languages are supported?

70+ languages with automatic detection. From major languages like English, Chinese, Japanese, Spanish to regional languages including Cebuano, Konkani, and Luxembourgish.

How do instructions work?

Describe how you want the voice to sound in plain English. For example: 'A 40-year-old British journalist, speaking slowly with dramatic pauses on key points.' AI interprets your direction and crafts the vocal performance.

What audio format is the output?

MP3, ready to download or chain into pipelines like Video Reel as a narration track.

Which model should I pick?

Gemini 3.8 Flash-Lite TTS is the fastest and cheapest, good for drafts and short clips. Gemini 3.8 Flash TTS gives studio-quality narration in more than 130 languages. Eleven v4 is the most expressive: it laughs, sighs, or whispers where you mark it. MiniMax Speech 2.8 Turbo and HD are a fast and a higher-fidelity alternative. Each returns an MP3 in seconds.

How many credits does it cost?

0.002 to 40.0 credits, depending on the model and the length of your text. Gemini models charge only for the audio they actually produce. Results are ready in seconds.