Ultra realistic TTS at 10x cheaper than ElevenLabs

Our custom models are built on a transformer backbone paired with our own neural audio codec. This architecture delivers natural, expressive speech with low latency and high efficiency.

Yes. Our models are proprietary and closed source, which allows us to maintain quality control, security, and continuous improvements without exposing our core IP.

We support a wide range of voices. Our custom-trained models cover major accents, with new voices added regularly. Contact us for a full list of supported languages.

Yes. We offer custom voice cloning and training services for enterprise customers. You can create unique, branded voices that match your identity and use case.

Yes. Send a recording (WAV, MP3, FLAC or OGG, up to 10 minutes) to /v1/stt and get the transcript back, at $0.009 per hour of audio from the same credit balance. The endpoint is OpenAI-compatible, so existing transcription code works by changing the base URL. Recordings are transcribed in memory and never stored.

Yes. Enterprise plans can run on dedicated capacity reserved for your traffic, with priority support and SLAs.

We operate in data centers across multiple regions globally. This ensures low latency, high availability, and compliance with regional data regulations.

We offer enterprise-grade SLAs with 99.9%+ uptime. Our multi-region infrastructure provides redundancy and failover to ensure consistent service availability.

Not by default. Zero Data Retention (ZDR) is on for every account and API key: your text is processed in memory and discarded, and generated audio is streamed back without a copy being kept. We only log request metadata (character count, audio length, voice, latency) for billing and usage stats. ZDR stays on unless your account owner chooses to turn it off in the console.

Not unless you opt in. With ZDR on (the default) we never use your text or audio for training. If your account owner turns ZDR off, requests are kept and may be used to improve our models, and in return they cost 15% less. You can turn it back on and delete kept data at any time. Cloned voices are never used for training.

All traffic is encrypted in transit with TLS and stored data is encrypted at rest. API keys and session tokens are stored only as hashes. See our privacy policy for the full details, and contact us for specific enterprise compliance requirements.

Our optimized infrastructure delivers under 200ms latency for most requests. For real-time streaming use cases, we offer low-latency endpoints designed for conversational AI and live applications.

We offer pay-as-you-go pricing, volume discounts, and custom enterprise plans. If you choose to turn Zero Data Retention off and share your requests with us, they cost 15% less. Contact our sales team for a quote tailored to your usage and requirements.