Text-to-Speech Service

Deliver natural-sounding speech with configurable voices, streaming playback, and file-based output. All endpoints use a consistent JSON payload via POST requests, making integration simple and straightforward.

Need an API Key? If you don't have an API key yet, you can create one here: https://playground.induslabs.io/register

Screenshot: where to find your API key. Create one at playground.induslabs.io/register

Shared Request Payload

All text-to-speech endpoints use the same JSON schema sent via POST request. Simply adjust parameters like output_format or stream depending on your use case.

{
  "text": "Hello, this is a test request.",
  "voice": "Indus-hi-Inaya",
  "output_format": "wav",
  "model": "indus-tts-v1",
  "api_key": "YOUR_API_KEY",
  "normalize": true,
  "stream": true,
  "speed": 1,
  "pitch_shift": 0,
  "loudness_db": 0,
  "sample_rate": 24000
}

Payload Fields

Name	Type	Default	Description
`text`	string	required	The text to be synthesized into speech.
`voice`	string	required	The voice model to be used (e.g., "Indus-hi-Inaya", "Indus-multi-Aadhya").
`output_format`	string	pcm	Audio format for output: "pcm", "wav", or "mulaw" (8 kHz, for telephony). MP3 is not supported.
`model`	string	indus-tts-v1	The TTS model to use (e.g., "indus-tts-v1").
`api_key`	string	required	Authentication API key.
`normalize`	boolean	true	Whether to normalize text before synthesis (default: true).
`stream`	boolean	true	Whether to stream the output (default: true).
`speed`	number	1	Speed of speech synthesis (default: 1).
`pitch_shift`	number	0	Pitch shift adjustment (default: 0).
`loudness_db`	number	0	Loudness adjustment in decibels (default: 0).
`sample_rate`	number	24000	Audio sample rate in Hz. Accepted values: 24000 (recommended), 16000, 8000.

POST/v1/audio/speech

Synthesize Speech

This endpoint is used to synthesize speech (TTS - Text-to-Speech) and stream the audio data.

Functionality

Converts input text into speech audio.
Uses credit system authentication.
Returns audio data directly in the response body.
Supports streaming for real-time audio playback.

Inputs

Name	Type	Default	Description
`text`	string	required	The text to be synthesized into speech.
`voice`	string	required	The voice model to be used (e.g., "Indus-hi-Inaya", "Indus-multi-Aadhya").
`output_format`	string	pcm	Audio format for output: "pcm", "wav", or "mulaw" (8 kHz, for telephony). MP3 is not supported.
`model`	string	indus-tts-v1	The TTS model to use (e.g., "indus-tts-v1").
`api_key`	string	required	Authentication API key.
`normalize`	boolean	true	Whether to normalize text before synthesis (default: true).
`stream`	boolean	true	Whether to stream the output (default: true).
`speed`	number	1	Speed of speech synthesis (default: 1).
`pitch_shift`	number	0	Pitch shift adjustment (default: 0).
`loudness_db`	number	0	Loudness adjustment in decibels (default: 0).
`sample_rate`	number	24000	Audio sample rate in Hz. Accepted values: 24000 (recommended), 16000, 8000.

Outputs

Status	Type	Default	Description
`200 OK`	audio/wav	-	Returns synthesized speech audio as binary data.
`422 Validation Error`	application/json	-	Validation failure. Inspect detail array.

200 OK

Binary audio data (WAV format)

422 Validation Error

{
  "detail": [
    {
      "loc": ["string", 0],
      "msg": "string",
      "type": "string"
    }
  ]
}

POST/v1/audio/speech/file

Synthesize Speech File

This endpoint is used to synthesize speech (TTS - Text-to-Speech) and return the complete audio file as a downloadable file.

Functionality

Converts input text into speech audio.
Returns the synthesized audio as a complete file download.
Unlike /v1/audio/speech, this endpoint returns the full audio file at once.

Inputs

Name	Type	Default	Description
`text`	string	required	The text to be synthesized into speech.
`voice`	string	required	The voice model to be used (e.g., "Indus-hi-Inaya", "Indus-multi-Aadhya").
`output_format`	string	pcm	Audio format for output: "pcm", "wav", or "mulaw" (8 kHz, for telephony). MP3 is not supported.
`model`	string	indus-tts-v1	The TTS model to use (e.g., "indus-tts-v1").
`api_key`	string	required	Authentication API key.
`normalize`	boolean	true	Whether to normalize text before synthesis (default: true).
`stream`	boolean	true	Whether to stream the output (default: true).
`speed`	number	1	Speed of speech synthesis (default: 1).
`pitch_shift`	number	0	Pitch shift adjustment (default: 0).
`loudness_db`	number	0	Loudness adjustment in decibels (default: 0).
`sample_rate`	number	24000	Audio sample rate in Hz. Accepted values: 24000 (recommended), 16000, 8000.

Outputs

Status	Type	Default	Description
`200 OK`	audio/wav	-	Returns the synthesized speech audio as a downloadable file.
`422 Validation Error`	application/json	-	Validation failure. Inspect detail array.

200 OK

Binary audio data (WAV format)

422 Validation Error

{
  "detail": [
    {
      "loc": ["string", 0],
      "msg": "string",
      "type": "string"
    }
  ]
}

POST/v1/audio/speech/preview

Speech Preview

This endpoint provides a preview of how text will be processed for speech synthesis without actually generating audio.

Functionality

Accepts input text and parameters, then shows how the text would be processed by the TTS system.
Does not generate audio, only returns metadata and analysis of the input.
Useful for estimating credits, duration, and validating parameters before synthesis.

Inputs

Name	Type	Default	Description
`text`	string	required	The text to be synthesized into speech.
`voice`	string	required	The voice model to be used (e.g., "Indus-hi-Inaya", "Indus-multi-Aadhya").
`output_format`	string	pcm	Audio format for output: "pcm", "wav", or "mulaw" (8 kHz, for telephony). MP3 is not supported.
`model`	string	indus-tts-v1	The TTS model to use (e.g., "indus-tts-v1").
`api_key`	string	required	Authentication API key.
`normalize`	boolean	true	Whether to normalize text before synthesis (default: true).
`stream`	boolean	true	Whether to stream the output (default: true).
`speed`	number	1	Speed of speech synthesis (default: 1).
`pitch_shift`	number	0	Pitch shift adjustment (default: 0).
`loudness_db`	number	0	Loudness adjustment in decibels (default: 0).
`sample_rate`	number	24000	Audio sample rate in Hz. Accepted values: 24000 (recommended), 16000, 8000.

Outputs

Status	Type	Default	Description
`200 OK`	application/json	-	Returns detailed analysis including character count, word count, estimated duration, credit cost, and configuration details.
`422 Validation Error`	application/json	-	Validation failure. Inspect detail array.

200 OK

{
  "analysis": {
    "total_characters": 30,
    "total_words": 6,
    "estimated_duration_seconds": 2.4,
    "estimated_credits": 0.04,
    "chunking_strategy": {
      "total_chunks": 1,
      "max_words_per_chunk": 15,
      "overlap_words": 0,
      "chunks": [
        {
          "index": 0,
          "word_count": 6,
          "text_preview": "Hello, this is a test request.",
          "is_final": true
        }
      ]
    }
  },
  "configuration": {
    "voice": "Indus-hi-Inaya",
    "model": "indus-tts-v1",
    "output_format": "wav",
    "stream": true,
    "temperature": 0.6,
    "max_tokens": 1800,
    "top_p": 0.8,
    "repetition_penalty": 1.1,
    "bitrate": null
  },
  "user_info": {
    "user_id": "USR_A3E785AF",
    "credits_remaining": 399.58,
    "tts_unit_cost": 1,
    "sufficient_credits": true
  },
  "output_settings": {
    "format": "wav",
    "voice": "Indus-hi-Inaya",
    "model": "indus-tts-v1",
    "streaming": true,
    "sample_rate": 24000,
    "channels": 1,
    "bit_depth": 16
  },
  "text_processing": {
    "original_text": "Hello, this is a test request.",
    "processed_text": null,
    "normalization_applied": false,
    "normalize_setting": true,
    "character_change": 0
  },
  "size_estimates": {
    "pcm_bytes": 115200,
    "wav_bytes": 115244,
    "mp3_bytes": 38400,
    "target_format_bytes": 115244
  }
}

422 Validation Error

{
  "detail": [
    {
      "loc": ["string", 0],
      "msg": "string",
      "type": "string"
    }
  ]
}

WS/v1/audio/speech_ws

WebSocket Streaming Speech Synthesis

Real-time text-to-speech synthesis using WebSocket for bidirectional streaming. Perfect for live speech generation, voice assistants, and low-latency applications.

Key Features

🔌 Persistent Connection: Maintains a WebSocket connection for real-time audio streaming.
⚡ Real-time Generation: Receive audio chunks as they are synthesized—no waiting for complete generation.
🎯 Low Latency: Optimized for live speech generation and real-time voice applications.
📊 Streaming Output: Get audio data progressively as binary chunks for immediate playback.
🔄 Bidirectional: Send synthesis request and receive audio chunks continuously.
🎵 Multiple Formats: Supports raw PCM, WAV, and 8 kHz mu-law audio with configurable parameters.

🔌 WebSocket Connection

Connect to: wss://voice.induslabs.io/v1/audio/speech_ws

Unlike REST endpoints, WebSocket maintains a persistent bidirectional connection for real-time streaming synthesis.

🤖 Available Models

indus-tts-v1: Standard TTS model with broad language support.
indus-tts-v3: Latest generation model with improved quality and naturalness.

WebSocket Message Flow

Message	Type	Order	Description
`Initial JSON Request`	JSON	first	Send complete request payload as JSON string including text, voice, model, and all parameters.
`Headers Message`	JSON	first reply	JSON message with type "headers" describing the audio (content type, sample rate, encoding).
`Audio Chunks`	binary	continuous	Raw audio data received as binary WebSocket frames (synthesized speech audio).
`Status Messages`	JSON	intermittent	JSON messages with type "done" (synthesis complete) or "error" (processing failed).

Request Parameters (JSON)

Name	Type	Default	Description
`text`	string	required	The text to be synthesized into speech.
`voice`	string	required	The voice model to use (e.g., "Indus-multi-Aadhya", "Indus-hi-Inaya").
`output_format`	string	pcm	Audio format for output: "pcm", "wav", or "mulaw" (8 kHz, for telephony). MP3 is not supported.
`stream`	boolean	true	Whether to stream the output (default: true).
`model`	string	indus-tts-v3	The TTS model to use (e.g., "indus-tts-v1", "indus-tts-v3").
`api_key`	string	required	Authentication API key.
`normalize`	boolean	true	Whether to normalize text before synthesis (optional).
`speed`	number	1	Speed of speech synthesis (optional).
`pitch_shift`	number	0	Pitch shift adjustment (optional).
`loudness_db`	number	0	Loudness adjustment in decibels (optional).
`sample_rate`	number	24000	Audio sample rate in Hz (optional).

Outputs

Status	Type	Default	Description
`headers`	JSON	-	Audio format details, sent before the first audio chunk.
`audio chunks`	binary	-	Raw synthesized audio data in the requested format (pcm, wav or mulaw).
`done`	JSON	-	Completion message indicating synthesis is finished.
`error`	JSON	-	Error message if synthesis fails.

Headers Message

{
  "type": "headers",
  "headers": {
    "Content-Type": "audio/wav",
    "X-Sample-Rate": "24000",
    "X-Audio-Encoding": "pcm_s16le",
    "X-Channels": "1",
    "X-Bit-Depth": "16",
    "X-Voice": "Indus-multi-Aadhya",
    "X-Request-ID": "f3b1c2d4-5e6f-4a7b-8c9d-0e1f2a3b4c5d"
  }
}

Audio Chunk (Binary)

Binary audio data (PCM/WAV/mu-law format)
Example: 4096 bytes of audio data per chunk

Done Message

{
  "type": "done",
  "request_id": "f3b1c2d4-5e6f-4a7b-8c9d-0e1f2a3b4c5d"
}

Error Message

{
  "type": "error",
  "message": "Invalid API key",
  "code": "AUTH_ERROR"
}

Quick Integration

Reference snippets for generating speech instantly.

API EndpointLanguage

import requests

url = "https://voice.induslabs.io/v1/audio/speech"
payload = {
    "text": "Hello from IndusLabs Voice API",
    "voice": "Indus-hi-Inaya",
    "output_format": "wav",
    "stream": False,
    "model": "indus-tts-v1",
    "api_key": "YOUR_API_KEY"
}

response = requests.post(url, headers={"Content-Type": "application/json"}, json=payload, timeout=30)
response.raise_for_status()

with open("output.wav", "wb") as f:
    f.write(response.content)
print("Audio saved to output.wav")