Copy page

Voice (TTS)

Phantom Speech

Phantom Speech is Neuratel's proprietary text-to-speech engine running on dedicated H100 hardware. It delivers 63ms average latency with 10 natural voices across English and Arabic. No external API key required.

No API key required — uses Neuratel's own hardware.

Specifications

Property	Value
English model	"phantom-english-speech-preview"
Arabic model	"phantom-arabic-speech-preview"
Average latency	63ms
Voices	10
Languages	English, Arabic

Configuration

Code
{
  "voice": {
    "provider": "phantom",
    "voice": "aria",
    "model": "phantom-english-speech-preview"
  }
}

Important: Phantom Speech uses the voice field (not voice_id like ElevenLabs/Cartesia).

Fields

Field	Type	Default	Description
voice	string	"aria"	Voice name (see table below)
model	string	"phantom-english-speech-preview"	phantom-english-speech-preview or phantom-arabic-speech-preview

Available Voices

English

Voice	Gender
aria	Female
bella	Female
claire	Female
alex	Male
david	Male
marcus	Male

Arabic

Voice	Gender
layla	Female
nour	Female
omar	Male
tariq	Male

Arabic Support

Phantom Speech has native Arabic support with natural intonation across dialects:

Code
{
  "voice": {
    "provider": "phantom",
    "voice": "omar",
    "model": "phantom-arabic-speech-preview"
  },
  "transcriber": {
    "provider": "deepgram",
    "model": "nova-3",
    "language": "ar"
  }
}

Related

• Phantom Brain — proprietary LLM
• Deepgram — recommended STT provider