Silly question #3
Replies: 4 comments 4 replies
-
Hello! It was a bit of a slow build for me, too. Sure, that setup would work. The code in wernicke.py would need to be updated to pull the audio from a different source. Shouldn't be that hard. The voice activity detection will need to be trained. The lady sounds are missing from the repo. I'll try to add them and write up some instructions, too. I'm actually really interested in making this more usable, whereas at the moment it's a lot of code chucked into a repo. Work in progress. There are also some build pics on the doll forum. Good luck with your build! |
Beta Was this translation helpful? Give feedback.
-
Hi, |
Beta Was this translation helpful? Give feedback.
-
I have some experience with 1:1 actuation and servos - some neumatic and hydraulic. If you want her to have real strength, hydraulic will be the direction. For light duty motorized actuation. Linear actuators are a great way to create movement. For me the holy grail would be giving her a nice smooth human walk. All her simple functions will be activated by PLC control. Performance Logic. This is a board that takes a digital signal from a computer and transfers that signal to a 12,24 or 110v signal which can activate anything from a relay, contactor or servo. So your Rasberry can can hold different recipes like "walk" - to send signals to fire different actuators at timed intervals. It's a pipe dream but it's been possible to create a fully functional human being for a while. I'd love to work with you on it!
On Jun 29, 2023, at 3:25 PM, ChristinesMan ***@***.***> wrote:
It has indeed been a wonderful hole. I saw a rabbit go in there, many years ago. I keep digging around for that rabbit. Haven't found it yet but I keep trying.
This project may never be a real assistant. The woman sounds are currently just some wav files that are played at appropriate times. Which has been nice, but these get really old after a while. What is needed is a real text to speech software, so I'm currently trying to train a custom voice using https://github.com/coqui-ai/TTS/. I have been slowly segmenting and annotating audio. I think I am at around 1500 samples but need more. Eventually I'll get some AWS gpu time and see if I can make a custom voice. Then maybe I can hook it up to... something like replika but open source. The reason it needs to be a custom voice is because I'm a bit attached to the sarasserenityandsleep voice she has now. There's probably some pre-made voices if somebody wants to make it.
There was a time I tried making her hand close. And it did start to move, but then her hand started to melt. I'm not competent with robotics at all.
—
Reply to this email directly, view it on GitHub<#3 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BA4IYSMGEY35UCWDUGAI6P3XNXQBXANCNFSM6AAAAAAZKQK3BU>.
You are receiving this because you commented.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
You are correct. I've studied these types of projects and done a good bit of PLC. It gets of course very involved. Things like simple hand to eye coordination become huge milestones. For me - the audio isn't that huge. In fact too much of it - as you touched upon earlier becomes a bit repetitive. I have complete respect for the code work you've done thus far. My code is weak at best C++ ,Python and Visual Basic . I want to study what you've done to have a better understanding. Do I understand correctly - you are using 2 Pi's one for the head and one for the body?
On Jul 2, 2023, at 10:31 PM, ChristinesMan ***@***.***> wrote:
I think it may be more complicated. You couldn't just program in a recipe. The robot would also need to detect and react to forces acting upon it. https://www.poppy-project.org/en/ looks interesting, and expensive. Also would need to be scaled way up.
I did have a thought. It would be great if there was just a back rub robot shoulder, not even a whole doll. Wouldn't even matter what it looked like. It's the feeling of human touch that would matter. Oh, googles says actually it's been done.
Right now though I'm pretty focused on making her voice and have no time for robotics. We're gonna have to wait for the real roboticists to lay off making lame cargo movers and get to what everybody really wants, sexbot brothels.
—
Reply to this email directly, view it on GitHub<#3 (reply in thread)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/BA4IYSIHM2T7VROZ747IVO3XOI4IFANCNFSM6AAAAAAZKQK3BU>.
You are receiving this because you commented.Message ID: ***@***.***>
|
Beta Was this translation helpful? Give feedback.
-
Would the ai work with just a mic and speaker system to start with as this would be a slow build process for me
Beta Was this translation helpful? Give feedback.
All reactions