Inworld

Inworld AI

Text-to-Speech platform with voice cloning, audio markups, and timestamp alignment.

Quick Navigation

When to Use

• Text-to-speech audio generation
• Voice cloning from 5-15 seconds of audio
• Emotion-controlled speech ([happy], [sad], etc.)
• Word/phoneme timestamps for lip sync
• Custom pronunciation with IPA

Models

Minimal Example

import requests, base64, os

response = requests.post(
    "https://api.inworld.ai/tts/v1/voice",
    headers={"Authorization": f"Basic {os.getenv('INWORLD_API_KEY')}"},
    json={"text": "Hello!", "voiceId": "Ashley", "modelId": "inworld-tts-1.5-max"}
)
audio = base64.b64decode(response.json()['audioContent'])

Key Features

• 15 languages — en, zh, ja, ko, ru, it, es, pt, fr, de, pl, nl, hi, he, ar
• Instant cloning — 5-15 seconds audio, no training
• Audio markups — [happy], [laughing], [sigh] (English only)
• Timestamps — word, phoneme, viseme timing for lip sync
• Streaming — /voice:stream endpoint
• TTS-2 steering — natural-language bracketed directions such as [say excitedly] or [whisper in a hushed style]
• Delivery mode — STABLE, BALANCED, CREATIVE trade consistency for emotional range
• Cross-lingual synthesis — reuse one voice across multiple languages; voice localization improves native-sounding output

Release Highlights (TTS-2)

• Realtime TTS-2 becomes the new primary model line via modelId="inworld-tts-2".
• Steering moves beyond the older fixed emotion tags: free-form bracketed directions can control style, pitch, speed, intensity, and non-verbals.
• Multilingual coverage expands with production quality across 15 languages and broader experimental coverage beyond that.
• deliveryMode adds a stability-vs-creativity knob, and specifying language matters more for cross-lingual output quality.

Prohibitions

• Audio markups work only in English
• Use ONE emotion markup at text beginning
• Match voice language to text language
• Instant cloning may not work for children's voices or unique accents

Links

• Documentation
• Changelog
• Platform

API Reference

Base URL

https://api.inworld.ai/tts/v1

Authentication

Authorization: Basic {INWORLD_API_KEY}

Endpoints

Synthesize Speech

POST /voice

{
  "text": "Hello, world!",
  "voiceId": "Ashley",
  "modelId": "inworld-tts-2"
}

Response:

{
  "audioContent": "base64-encoded-audio",
  "timestampInfo": {...}
}

Streaming

POST /voice:stream

Same body. Returns chunked audio + alignment per chunk.

Audio Formats

{ "audio_config": { "audio_encoding": "LINEAR16", "sample_rate_hertz": 48000 } }

Request Parameters

Models

Languages: en, zh, ja, ko, ru, it, es, pt, fr, de, pl, nl, hi, he, ar

TTS-2 notes:

• deliveryMode lets you choose consistency vs expressiveness.
• Set language explicitly when reusing a voice across multiple languages or when localizing a voice for a specific target language.

Integrations

Audio Generation

Supported Formats

Audio Config

{
  "audio_config": {
    "audio_encoding": "LINEAR16",
    "sample_rate_hertz": 48000
  }
}

Voice Parameters

Emphasis

Use asterisks to emphasize words:

{ "text": "I *really* need this done today" }

Languages

en, ar, zh, nl, fr, de, he, hi, it, ja, ko, pl, pt, ru, es

Best Practice: Match voice language to text language for best quality.

TTS-2 notes:

• Cross-lingual synthesis lets you reuse one voice across multiple languages, but quality is best when you also provide the language hint explicitly.
• Voice localization is the preferred path when you want the same voice identity to sound native in a target language instead of merely reading cross-lingually.

Built-in Voices

Available in TTS Playground. For custom voices, use voice cloning (min 5 seconds audio).

Voice Cloning

Cloning Types

Instant Cloning via Portal

1. Go to TTS Playground → Clone Voice 2. Choose Upload audio or Record audio 3. Name voice, select matching language 4. Upload/record up to 3 samples (5-15 seconds each) 5. Enable "Remove background noise" if needed 6. Click Continue → wait for validation 7. Test and copy voiceId for API use

Audio Requirements

• Formats: WAV, MP3, WebM
• Max total size: 16MB
• Samples > 15 seconds are auto-trimmed
• Match language to text you'll synthesize

Recording Tips

• Find quiet environment
• Keep reasonable mic distance
• Speak expressively (varied emotions)
• Use suggested scripts (Math Tutor, AI companion, etc.)

API Usage

response = requests.post(
    "https://api.inworld.ai/tts/v1/voice",
    headers={"Authorization": f"Basic {os.getenv('INWORLD_API_KEY')}"},
    json={
        "text": "Hello!",
        "voiceId": "your-cloned-voice-id",  # From TTS Playground
        "modelId": "inworld-tts-1.5-max"
    }
)

Limitations

• Instant cloning may not perform well for:
• Children's voices
• Unique accents
• Use professional cloning for these cases

Installation & Quickstart

API Key

1. Sign up at https://platform.inworld.ai/ 2. Go to Settings → API Keys 3. Create and copy Base64 credentials

export INWORLD_API_KEY='your-base64-api-key-here'

Basic Request (Python)

import requests
import base64
import os

response = requests.post(
    "https://api.inworld.ai/tts/v1/voice",
    headers={
        "Authorization": f"Basic {os.getenv('INWORLD_API_KEY')}",
        "Content-Type": "application/json"
    },
    json={
        "text": "Hello, world!",
        "voiceId": "Ashley",
        "modelId": "inworld-tts-1.5-max"
    }
)

audio_content = base64.b64decode(response.json()['audioContent'])
with open("output.mp3", "wb") as f:
    f.write(audio_content)

Streaming Request

import requests
import base64
import os
import json

response = requests.post(
    "https://api.inworld.ai/tts/v1/voice:stream",
    headers={
        "Authorization": f"Basic {os.getenv('INWORLD_API_KEY')}",
        "Content-Type": "application/json"
    },
    json={
        "text": "Your text here...",
        "voiceId": "Ashley",
        "modelId": "inworld-tts-1.5-max",
        "audio_config": {
            "audio_encoding": "LINEAR16",
            "sample_rate_hertz": 48000,
        }
    },
    stream=True
)

for line in response.iter_lines():
    chunk = json.loads(line)
    audio_chunk = base64.b64decode(chunk["result"]["audioContent"])
    # Process audio_chunk (skip first 44 bytes WAV header)

Voice Control

Audio Markups (English only, experimental)

Emotion & Delivery Style

Place at beginning of text. Use only ONE per request:

[happy] I can't believe this is happening!
[sad] I really miss those days.
[angry] This is unacceptable!
[surprised] You did what?!
[fearful] I heard something outside...
[disgusted] That's revolting.
[laughing] Oh that's hilarious!
[whispering] Don't let them hear us.

Split multiple emotions into separate API calls.

TTS-2 Natural Language Steering

TTS-2 expands voice control beyond the older fixed markup set. You can use bracketed natural-language instructions such as:

[say excitedly] We finally shipped it.
[whisper in a hushed style] Keep your voice down.
[speak as if barely holding back rage] I said stop.

Use this when you need broader control over articulation, volume, pitch, pacing, vocal texture, or non-verbal delivery than the old fixed emotion tags provide.

deliveryMode guidance:

• STABLE for consistent production narration.
• BALANCED for a middle ground.
• CREATIVE when you want more expressive range and can tolerate more variation.

Non-Verbal Sounds

Can be used anywhere in text:

[clear_throat] Did you hear me? [sigh] You never listen!
[breathe] Let me think... [cough] Excuse me.
[laugh] That's so funny! [yawn] I'm tired.

Custom Pronunciation (IPA)

Use /phonemes/ for uncommon words:

{ "text": "Your adventure in /kri:t/ begins today." }

Finding IPA:

• Ask LLM: "What are the IPA phonemes for Crete, pronounced like kreet?"
• Use Vocabulary.com IPA guide

Timestamps (English only)

Enable with timestampType:

Note: Increases latency (especially non-streaming).

Request

{
  "text": "Hello, world!",
  "voiceId": "Ashley",
  "modelId": "inworld-tts-1.5-max",
  "timestampType": "WORD"
}

Response (TTS 1.5)

{
  "timestampInfo": {
    "wordAlignment": {
      "words": ["Hello,", "world!"],
      "wordStartTimeSeconds": [0, 0.28],
      "wordEndTimeSeconds": [0.28, 0.8],
      "phoneticDetails": [
        {
          "wordIndex": 0,
          "phones": [
            { "phoneSymbol": "h", "startTimeSeconds": 0, "durationSeconds": 0.07, "visemeSymbol": "aei" },
            { "phoneSymbol": "l", "startTimeSeconds": 0.1, "durationSeconds": 0.09, "visemeSymbol": "l" }
          ],
          "isPartial": false
        }
      ]
    }
  }
}

Viseme Symbols