-
oh look, sama adds tts to chatgpt put it in a smart speaker and add a hotword, coward.
-
then let it connect to my security cameras to find where I am in the house, and get visual context
-
then add on-device TTS and on-device continuous ASR to refer to previous audio context where needed.
-
make the TTS streaming from tokens, make it as expressive as XTTS or 11, make the hotword optional, allowing for the assistant to chime in when people are having an argument, or when it deems it could answer a question you have before you can even start asking
-
theres a speaker only, and a screen-enabled version of the thing, it has a cool name that's short but unique and easy to say in most languages, speaker only is 149,99, with screen is 229,99, but includes 3 months free of chatgpt plus
-
create companion apps for Windows, Mac and Linux that allow it to take a look at your screen if needed, the companion app can also serve as a half-baked version of the screen model, its not perfect, and you pass it off as engineering issues when its just to sell more devices
-
fact being, the screen's UI is CEF, so porting it to other platforms wouldn't be technically challenging. its really just a cashgrab
AuroraNemoia’s Twitter Archive—№ 312