Changelog
What is new in Dialect.
We ship improvements regularly. Here is what has changed recently, newest first.
Updates
Recent releases.
July 24, 2026v0.9New
Japanese, Korean, and Mandarin added
Dialect now supports Japanese, Korean, and Mandarin Chinese, bringing the total to ten languages. Each new language uses a native-speaker voice profile tuned for conversational pace and natural rhythm.
- Japanese conversations use polite register by default and adapt to a casual register as you advance.
- Korean gently guides you on the honorifics that everyday speech depends on.
- Mandarin keeps beginner turns short and clear so tones and phrasing have room to sink in.
June 11, 2026v0.8Improved
Faster voice responses and lower latency
The gap between your turn and the AI partner's spoken reply has been cut by roughly half. Conversations now feel closer to a real back-and-forth exchange in the language you are learning.
- ElevenLabs voice streaming now begins while the language model is still generating the response, so you hear the first words sooner.
- Average time-to-first-audio dropped from 2.3 seconds to under 1.1 seconds.
- Latency improvements apply to all plans, no setting change needed.
May 20, 2026v0.7New
Practice points over time, level picker, and typed fallback
Your dashboard now shows how your practice scores change across conversations, so you can see yourself improving. A new level picker lets you move between beginner, intermediate, and advanced at any time. And if your microphone is unavailable, you can now type instead.
- The progress view shows a sparkline of your practice scores per language, so you can spot where you are getting stronger.
- The level picker appears at the start of each conversation. Switching levels mid-journey takes one tap and applies to the next session.
- The typed fallback gives you the same patient AI partner and the same practice report, no microphone required.