FAQ
Frequently Asked Questions
What kinds of speech data do you offer?
We offer four tiers of pre-labeled speech datasets — from untranscribed audio with metadata tags (Tier I) to fully annotated conversational data with speaker turns, diarization and custom labels (Tier IV). Languages, accents and domains cover 50+ languages.
Can I license data for commercial use?
Yes. Our datasets are fully licensed for commercial training. The open-sourced sample (OleSpeech-IV) is non-commercial research-only; the full tiers are available under commercial agreements. Commercial licenses are available to U.S. entities only.
Do you build custom voice AI models?
Yes. We build custom TTS, STT, voice cloning, speaker embedding models and so on, tuned to your domain and deployed on your infrastructure — on-prem or in your cloud account. You keep 100% of the IP on the code we deliver.
What is your voice cloning quality like?
Our proprietary fine-tuned voice cloning produces a nearly identical voice to the original speaker. Listen to samples, or contact us to arrange a private demo.
How do you handle data privacy and NDAs?
Every engagement is covered by a strict NDA. We minimize human involvement in your data pipeline to reduce breach risk, maintain end-to-end audit trails, and comply with AB 2013 and other applicable regulations.
Can you consult on our voice AI strategy?
Yes. We offer technology and market evaluation, strategic business planning, and build-vs-buy assessments under NDA. See Tech Consulting for details.
How do I get started?
A typical engagement begins with a short email exchange — no confidential information needed — followed by an NDA or mNDA. Consulting engagements include a complimentary 30-minute kickoff session.
Our datasets and services are available to organizations only, not individual consumers.