MobMate – GPU Acceleration Pack · build 23483403 · published
MobMate WhispTalk V2 has been released.
In this V2 update, we have significantly reviewed the internal processing, UI, and initial setup experience, with a particular focus on speech recognition.
Rather than directly porting every feature that had been built up in V1, we prioritized stabilizing the core STT → TTS flow first.
Because of this, V2 is being released with a more focused feature set.
If you still need features that were available in V1, you can continue using the V1 version for the time being by launching the included MobMateWhispV1.exe.
We plan to gradually bring V1 features into V2 through future updates.
Improved speech recognition accuracy
Reworked the real-time STT processing flow
Faster audio processing during speech
Added a mechanism to reduce unwanted reactions to surrounding voices and sounds
Added noise floor measurement
Simplified the initial setup, with more settings handled automatically
Refined the UI for V2
Reorganized the flow from STT to TTS
In V1, some parts of the setup and speech recognition behavior were difficult to configure.
In V2, many of these processes have been redesigned to work in the background, so users can get started with less manual adjustment.
During development benchmarks, we also saw a significant improvement in recognition accuracy compared to V1.
In particular, we aimed to improve stability for harder-to-recognize voices and audio closer to real-world usage environments.
V2 is a version where we have significantly reworked the internal structure to make future improvements easier.
At the same time, not every feature from V1 is included in V2 from the beginning.
If you need V1 features, please launch the included MobMateWhispV1.exe and continue using the V1 version.
We plan to keep including the V1 version for a while.
As V2 continues to receive updates, we will gradually move necessary features over to the V2 side.
MobMate WhispTalk V2 is not simply a direct replacement for V1.
It is a rebuilt foundation for future improvements.
For this release, we focused first on making the core flow for joining voice chat without speaking — speech recognition → text → TTS output — more stable and easier to use.
There are still areas we want to improve, and features we plan to bring back, but with V2 we aimed to make MobMate WhispTalk feel more natural and easier to try.
Please give it a try in your own environment.
Feedback is always welcome.