Voice Recognition That Gets It Right

Yak Lifetime offers a 14-day free trial with no credit card required, plus a 30-day money-back guarantee after purchase. Yak Pro does not include a free trial - it's a subscription that you can cancel anytime, with access continuing until your subscription period ends.

Yak currently supports macOS 12 (Monterey) and later, with native support for both Apple Silicon (M1/M2/M3/M4) and Intel Macs. Windows and Linux versions are in development.

Yak supports 100+ languages with automatic detection, and this number is continuously growing. The multimodal AI engine handles various accents, dialects, and speaking speeds with ease.

Yes, data protection is a core principle. Yak inherits Vertex AI's zero data retention policy — your audio is processed and immediately discarded, never stored, never used for model training, and never reviewed by humans. All transcription data is saved locally on your device only. Vertex AI holds certifications including SOC 2, ISO 27001, HIPAA, and more.

Yes, Yak requires an internet connection for transcription as it uses cloud-based multimodal AI models. You can use Yak Cloud or BYOK (Bring Your Own Key) mode with your own API key.

Both plans can be activated on up to 3 devices. You can manage your devices from your account dashboard.

Yak Lifetime (lifetime license) comes with a 30-day money-back guarantee. For Yak Pro subscriptions, you can cancel anytime and keep access until the end of your billing period.

Yak Lifetime is a one-time $59 payment for lifetime access with 20 min/week cloud AI transcription and 1 year of updates — ideal for users who prefer to bring their own API key (BYOK) for unlimited use. Yak Pro is a subscription ($10/mo annual or $12/mo monthly) with unlimited cloud AI transcription and always-updated features.

The first few months of Yak's development revolved around a local-model architecture — until we realized that local models simply cannot deliver high-quality transcription in a single pass. Multimodal large language models are currently the best choice for voice-to-text applications, while local models have yet to reach the required level of capability. Our goal has always been to provide the best possible user experience, so when local models become powerful enough, we will absolutely add support for them.