tts-rt-v1
AI model · Text to speech · Text → Audio / speech by Soniox
Real-time text-to-speech model with instant voice cloning and expressive emotional control.
tts-rt-v1 is Soniox's real-time text-to-speech model designed for lifelike speech generation in 60+ languages. It is built for voice agents and live conversations, offering low-latency streaming that allows audio to start before the full text response is available. The model features instant voice cloning and precise control over emotions and delivery through audio tags. It accurately handles complex alphanumeric data and multilingual code-switching, ensuring names and numbers are spoken correctly.
- Languages
- English, Spanish, French, German, Japanese, Korean, Chinese, Arabic, Portuguese, Italian
- Modality
- Text to Audio
- Supported languages
- 60+
- Voice options
- 200+ studio-quality voices
- Access method
- Streaming API / REST API
Overview
tts-rt-v1 is Soniox's real-time text-to-speech model designed for lifelike speech generation in 60+ languages. It is built for voice agents and live conversations, offering low-latency streaming that allows audio to start before the full text response is available. The model features instant voice cloning and precise control over emotions and delivery through audio tags. It accurately handles complex alphanumeric data and multilingual code-switching, ensuring names and numbers are spoken correctly.
Speech Synthesis
- Streaming generation: Generates audio from text in real-time, starting playback before the full sentence is generated.
- Instant voice cloning: Creates a high-fidelity voice clone from just seconds of audio, preserving identity and rhythm.
- Expressive control: Uses audio tags to direct emotion, delivery, and vocal reactions like whispers or laughter.
- Character timestamps: Provides character-level timestamps to synchronize text and audio or handle interruptions.
Voice and Delivery
- Alphanumeric accuracy: Pronounces phone numbers, email addresses, and verification codes exactly as written.
- Multilingual precision: Handles foreign names and technical terminology naturally within a single utterance.
- Voice library: Offers 200+ studio-quality voices across various accents, ages, and styles.
Pricing
- Input text tokens
- $4.00 per 1M tokens
- Output audio tokens
- $21.50 per 1M tokens
- Effective hourly rate
- ~$0.70 per hour of generated speech
Specifications
- Modality
- Text to Audio
- Supported languages
- 60+
- Voice options
- 200+ studio-quality voices
- Access method
- Streaming API / REST API
Added by AntSpace from public sources (soniox.com).