Infrastructure scaled; UK English and Swiss German routing fixed; Cloud Sync data control added

A more reliable Wispr Flow

Rapid growth in our user base strained our infrastructure recently, and some users saw slower dictation, lower accuracy, or trouble signing in and using polish.

We've scaled up our infrastructure to keep pace with growth and isolated the most demanding work onto a dedicated system, so heavy usage in one area can't disrupt the rest. We've also tightened monitoring to catch and resolve issues faster. This work is ongoing.

On accuracy, we fixed several specific issues:

• Better language routing: UK English and Swiss German weren't always being sent to the right place, which hurt accuracy for those users. Both are now routed correctly.
• Cleaner audio handling: Audio compression wasn't fully working as intended, which could affect how well your speech was transcribed. It's now working the way it should.
• A higher bar for every model: We built a much stronger set of evaluations across all of our models, so we can catch accuracy regressions before they reach you.

We're also testing a new round of model improvements to push accuracy further, and building toward personalized speech models that adapt to how you speak over time.

Updated data controls — Cloud Sync

We're introducing Cloud Sync, a new setting alongside Privacy Mode that gives you more granular control over how Flow handles your data.

• Privacy Mode controls whether your data can be used for model training. This hasn't changed.
• Cloud Sync controls whether your transcription data (transcripts, audio, and dictation history) is stored on Wispr's servers. Turning Cloud Sync on unlocks cross-device sync for Wispr Scratchpad and enables future features.

Existing users with Privacy Mode on: nothing changes automatically. Privacy Mode remains on, and Cloud Sync stays off. Your dictionary, snippets, and custom prompts continue to sync across devices regardless of your Cloud Sync setting.