Speech Recognition

Configure cloud speech-to-text and BYOK options

Overview

Yark uses multimodal AI models for speech recognition, delivering high accuracy with semantic understanding. All transcription is cloud-based, powered by Google's Gemini models via Vertex AI.

Yark Cloud

Yark Cloud is the default transcription engine, requiring no additional setup beyond signing in to your Yark account.

Modes

| Mode | Latency | Quality | Best For | |------|---------|---------|----------| | Normal | ~1-2s | High | Everyday transcription | | Thinking | ~3-5s | Very High | Complex content, technical jargon, nuanced context |

Normal mode — Fast, low-latency transcription suitable for most use cases.
Thinking mode — Deeper AI reasoning for significantly improved quality on complex or technical content.

BYOK (Bring Your Own Key)

Yark supports "Bring Your Own Key" mode for users who want full control over their API usage:

Google AI Studio — Use your Gemini API key from Google AI Studio
Vertex AI — Use your Vertex AI credentials for enterprise-grade data protection

Setting Up BYOK

Go to Settings > Transcription > Engine
Select BYOK as your engine
Choose your provider (Google AI or Vertex AI)
Enter your API key
Select your preferred model

Data Protection

Yark Cloud is powered by Vertex AI with enterprise-grade data protection:

Zero data retention — Audio is processed and immediately discarded
No model training — Your data is never used to train models
No human review — No human ever accesses your audio
Regional data residency — Data stays in your chosen region
Compliance — Vertex AI holds certifications including SOC 2, ISO 27001, FedRAMP, HIPAA, PCI-DSS

All transcription recordings are saved locally on your device only.

Choosing the Right Option

| Priority | Recommendation | |----------|----------------| | Ease of use | Yark Cloud (Normal mode) | | Highest quality | Yark Cloud (Thinking mode) | | Cost control | BYOK with Google AI Studio | | Enterprise compliance | BYOK with Vertex AI |