Desktop v1.5.891 — A more reliable Wispr Flow

A more reliable Wispr Flow

Addressed infrastructure strain that caused slower dictation, lower accuracy, and sign-in issues. Scaled up infrastructure and isolated demanding work onto dedicated systems.

Accuracy improvements:

• Better language routing for UK English and Swiss German
• Fixed audio compression handling for better transcription
• Stronger evaluation sets across all models to catch regressions

Updated data controls:

Introduced Cloud Sync, a new setting alongside Privacy Mode for granular data control:

• Privacy Mode: Controls whether data is used for model training (unchanged)
• Cloud Sync: Controls whether transcription data is stored on servers, unlocking cross-device sync for Wispr Scratchpad and future features

Existing users with Privacy Mode on see no automatic changes. Dictionary, snippets, and custom prompts continue to sync regardless of Cloud Sync setting.

Building toward personalized speech models that adapt to how you speak over time.

Data Controls documentation