I want to compare my experience with you guys' and get your input and feedback on the following in hopes of getting a more complete picture on the subject.
I wanted to install and test the different google STT options. So in a new profile I installed all google play components first and initiated them by opening play store and going to the signup page. I then installed Gboard, Recorder, and Speech Services By Google one by one and tested each, I installed all of them using Aurora and not from play store. Didn't want to use a google account.
First I installed Gboard, with network, and set the "Faster Voice Typing" with English-US. After turning network off It worked offline nicely. It lacked automatic punctuation, however if you say "dot" or "period" it will place a ".". If you say comma, hyphen, underscore, colon, semicolon, question mark, exclamation, it will place all the appropriate symbols, Usually. On some occasion it missed and wrote the word, but I suppose that's expected. When you say "enter", it also usually inputs an "enter" and jumps to next line. It didn't require anything else to be installed.
Next I installed the Recorder. With or without network, Recorder would record voice but going to the transcribe page would give me a message that to get transcription I need voice installed. I realized it probably means "Speech Services". Clicking "download" didn't work but I later found out this was because I had blocked either the network access to one of the google components or perhaps I hadn't given the "install from unknown sources" permission to the play store. I can't recall which it was. But regardless, at the time I simply went to Aurora Store and installed Speech Services from there manually. Once done, Speech Services needed to be selected as the system voice engine, and I also checked its settings for languages and downloaded both US and UK english for it. US was set as default.
Now recorder worked both online and offline. Recorder would do auto punctuation and I didn't need to verbally announce punctuation. Although about a 3rd of the time it didn't get things right, but also I didn't spend too much time adjusting my talking in a way to help it. I was mostly speaking non-sense. But Recorder was the most complete auto writing of the bunch. Replaying the audio actually highlights each written word as its being played back in correct timing.
Next I tried using the Speech services by itself using the "Voice Keyboard". This one had no punctuation at all. So even when I said "dot", "period", "comma", etc, it only wrote the words. I felt it was rather impractical for actually writing a message.
I would have liked to test each of the 3 apps without the presence of the other two just to ensure they are not pulling from each other's libraries, and to see exactly how independently they work and if they require other component installations. But in the end it seemed like a lot of work and I got the feeling there wasn't too much more to it.
One other thing to note is that in the end I also tested all apps with "ALL components" (Play Store, Play Services, Services Framework, Gboard, Speech Services, Recorder) network blocked. And then I also tested them again network blocked and with Play Store, Play Services, Services Framework fully disabled. Gboard and Speech services worked just file and all the same. Recorder however would repeatedly throw up a notification warning about Play Services not working or something. It was annoying so I had to disable that notification for the app, and now it just works with no warnings.
Now to the points of discussion:
Reading many of the other posts on the subject, I had heard the Recorder didn't need any other component to be installed. But in my experience it needed the Speech Services. Has anybody experienced otherwise?
I am wondering if I did anything incompletely and whether the voice typing experience on any of these apps could have been further improved that I may have missed out on because of the way I went about it. Has anybody experienced better results from Speech Services than I did? Are there any hints/tips that I missed? Would other methods of installation (such as installing from google play store) make any difference?
One thing I didn't properly test here is the actual MINIMUM dependency required for each app. As mentioned I initially installed all 3 google components. So I'm wondering if any of you guys can recount the minimum required for each of those apps if you have used them alone. I've heard for Gboard its only GSF, and GSF can even be disabled after as I experienced. But I don't know for the other two.
One thing I would like to know is which one of these three is taking advantage of the "hardware" acceleration or so called Neural AI of the CPU for speech, and which one is not. Anybody know this?
Have I missed any other google Voice options? I know I didn't even mention SayBoard or RH Voice, but I wanted to focus on the strongest contenders in the market that can be placed on GOS.
Would love to get your comments about your experiences and knowledge base that could fill some of the holes in my observations.