Replies: 1 comment 3 replies
-
|
You click the microphone on the keyboard and voice to text works great already, if you have it configured on the receiving end it works correctly as well. This functionality is easily provided already by modern phones. |
Beta Was this translation helpful? Give feedback.
-
|
Ok. So can you explain how to configure the phone to have Meshtastic automatically speak the currently receiving message with the TTS model assigned to the sending node? |
Beta Was this translation helpful? Give feedback.
-
This is something I agree with Garth on. What you are proposing is useless bloat. These features are accessibility features that have been available on modern operating systems for years. Just because these features aren't a 1-click solution doesn't mean they don't exist. If you want custom TTS models you should check with your operating aystem on how to install them. If the operating system supports different voice models per notification then this might be something that the Apps could utilize. |
Beta Was this translation helpful? Give feedback.
-
You use the microphone next to the keyboard when sending a message instead of the keyboard, on iOS you set Settings > Siri > Announce Notifications, then turn on Announce Notifications |
Beta Was this translation helpful? Give feedback.
-
I would like you to give this a consideration for future app support:
An implementation of a voice-to-text-to-air-to-text-to-voice flow that equals the quality of a classic walkie-talkie with a fraction of the bandwidth.
Let's discuss the single steps:
We know the text-to-air-to-text part, thats Meshtastic. The voice-to-text part is something every device can do today in some form via the internal keyboards microphone symbol. However there is no real integration into the app, like by just tapping one button and talk (it's always a series of taps to select the input field, to open the keyboard and then to activate the microphone) or preselecting the speech2text engine you'd prefer. So thats that.
On the text-to-voice part there is ai software now that can generate personalized tts models. So what we still need now is the app integration that takes one of these tts models and uses it to read the received message either automatically on receiving or by tap on a button. Also we need the possibility to assign different tts models to different nodes because the idea is, that every node generates a tts model with the node users personal voice so that the synthesized voice is 'real'.
A bit of an open topic to perfect quality is still remaining on the voice-to-text part in that there seems to be no speech2text engine that not only cleans the text and adds proper punktuation but also interprets the input to generate a text with fitting ssml annotations of how the output should be spoken and emphasized so that the quality of a real voice channel can be reached.
Beta Was this translation helpful? Give feedback.
All reactions