🎤Micdrop

Gradium TTS in Node.js

GradiumTTS reads the answers of your agent with Gradium’s voices over its WebSocket streaming API. It connects to the EU endpoint by default.

Installation

Terminal window
npm install @micdrop/gradium

The same package also provides a speech to text. The providers page shows what every provider covers.

Usage with MicdropServer

import { GradiumTTS } from '@micdrop/gradium'
import { MicdropServer } from '@micdrop/server'
const tts = new GradiumTTS({
apiKey: process.env.GRADIUM_API_KEY || '',
voiceId: 'Bla6SbVMczYnOhfK', // Gradium voice ID
modelName: 'default', // Optional: model name
outputFormat: 'pcm_16000', // Optional: audio format
region: 'eu', // Optional: 'eu' or 'us'
})
// Use with MicdropServer
new MicdropServer(socket, {
tts,
// ... other options
})

Usage without MicdropServer

import { GradiumTTS } from '@micdrop/gradium'
import { Readable } from 'stream'
const tts = new GradiumTTS({
apiKey: process.env.GRADIUM_API_KEY || '',
voiceId: 'Bla6SbVMczYnOhfK',
})
// Audio is raw PCM, 16 bits, 16 kHz, mono
tts.on('Audio', (chunk) => console.log('Audio:', chunk.length, 'bytes'))
tts.on('Failed', (texts) => console.error('Failed:', texts))
tts.speak(Readable.from(['Hello! ', 'What can I do for you?']))

Events

EventPayloadDescription
AudioBufferA chunk of audio, PCM 16 bits, 16 kHz, mono, ready to be played.
Failedstring[]Gradium refused the request, or every reconnection attempt failed, with the text that stayed unspoken. The next utterance connects again.

See the TTS interface for the full contract.

Options

OptionTypeDefaultDescription
apiKeystringRequiredYour Gradium API key
voiceIdstringRequiredGradium voice ID
modelNamestring'default'Model name to use for speech synthesis
outputFormatGradiumOutputFormat'pcm_16000'Audio output format
region'eu' | 'us''eu'API region (EU or US endpoint)
jsonConfigGradiumJsonConfigOptionalAdvanced voice configuration
connectionTimeoutnumber5000Timeout in milliseconds for WebSocket connection
retryDelaynumber1000Delay in milliseconds between reconnection attempts
maxRetrynumber3Number of failed connection attempts in a row before emitting Failed

Interruptions and reconnection

Gradium stops a synthesis when its connection closes, so an interruption closes the connection and opens a new one right away, without counting as a failed attempt. A connection that drops on its own, a graceful close included, opens again and receives the sentence under way.

Output Formats

FormatDescription
pcmPCM 48kHz, 16-bit signed integer mono
pcm_16000PCM 16kHz, 16-bit signed integer mono
pcm_24000PCM 24kHz, 16-bit signed integer mono
pcm_8000PCM 8kHz, 16-bit signed integer mono
wavWAV format
opusOpus codec wrapped in an Ogg container
ulaw_8000u-law 8kHz
alaw_8000A-law 8kHz

Advanced Voice Configuration

The jsonConfig option allows you to fine-tune voice characteristics:

const tts = new GradiumTTS({
apiKey: 'your-api-key',
voiceId: 'your-voice-id',
jsonConfig: {
temp: 0.7, // Temperature: 0 to 1.4 (default: 0.7)
cfg_coef: 2.0, // Voice similarity: 1.0 to 4.0 (default: 2.0)
padding_bonus: 0, // Speed control: -4.0 to 4.0
},
})
ParameterTypeRangeDefaultDescription
tempnumber0 - 1.40.7Controls randomness in speech generation
cfg_coefnumber1.0 - 4.02.0Controls similarity to the original voice
padding_bonusnumber-4.0 - 4.00Controls speech speed

On-device voices with Phonon

Gradium also publishes Phonon, a model of 31 to 40 million parameters that runs on the CPU, with no API key and no audio leaving the machine. @micdrop/phonon runs it inside your Node server, in English, French, German, Spanish and Portuguese. It suits a call that has to stay on the machine, and the hosted API suits a call that needs more voices or voice cloning.

Getting Started

  1. Sign up for a Gradium account and get your API key
  2. Choose a voice ID from the Gradium voice library
  3. Install the package and configure with your credentials
import { GradiumTTS } from '@micdrop/gradium'
const tts = new GradiumTTS({
apiKey: 'your-gradium-api-key',
voiceId: 'your-voice-id',
region: 'eu', // Use 'us' for US endpoint
outputFormat: 'pcm_16000',
})
// Use with MicdropServer
new MicdropServer(socket, {
tts,
// ... other options
})