Simple, token-based pricing
Start free, buy a pack that never expires, or subscribe monthly. Transcription and answer generation are billed through one simple token model.
Token cost reference
ShortAnswer, LongAnswer, extension actions, and transcription are billed separately.
Common questions
Can the interviewer see the overlay?
No. It is a separate always-on-top window under your control. If you share only the browser or a specific app window, Sovia stays outside the shared content.
Does Sovia send audio to the cloud?
Audio goes through Sovia backend for transcription. After that, answer generation can use managed providers, your existing Cursor or Claude setup, or your own LMStudio model.
Which AI providers are supported?
Managed providers, Anthropic/Claude, Cursor-based setups, and local models via LMStudio. You can choose the answer-generation setup that fits your workflow.
How do tokens work?
Backend transcription costs 5 tokens per call. Managed AI answers cost 25 tokens per response. BYO Cursor, Claude/Anthropic, and LMStudio answers cost 3 tokens per response.
Does it work in Zoom, Meet, and Teams?
Yes. Sovia captures system audio, so it works with Zoom, Google Meet, Teams, and browser-based calls.
What macOS version do I need?
macOS Ventura 13 or later, or Windows 10+. Use the desktop app that matches your setup.
Can I use my own local model?
Yes. LMStudio stays supported for answer generation, but speech-to-text itself is handled through Sovia backend.
Is there a free plan?
Yes. You get a signup token bonus with no card required so you can test the workflow before paying.