Starting a code project to make my doll speak

[MannequinFan]

Unfortunately I don't have a cell phone or a tablet, so I can't try out the app.
(Yes I'm old school - land line and desktop/laptop puters)
Does it require pushing a button to start listening?

BTW, on the VA site it mentions now being able to use the new Microsoft Speech Platform 11, which is supposedly better than the standard windows VR. No training required. Looks like it works with Win7.
I'm not sure if it will work with Kari and others though.

VA page: http://voiceattack.com/alternate.aspx
Microsoft site: https://www.microsoft.com/en-us/downloa ... x?id=27225

Wow! That might do the do. Downloading now.
On Doll33, you do have to push a button to start it, but then it runs for a little bit. Still some bugs to work out.

Cool, let me know how it works!
Looks like you need to download both the runtime and the appropriate speech engine(s).
I think you can still use any SAPI5 voice like Ivona.
Here's the VA forum page discussing this...
https://voiceattack.com/SMF/index.php?topic=1375.0

[Begog]

You know what? I think i already had it.
Version:11
File Name:x86_SpeechPlatformRuntime\SpeechPlatformRuntime.msi

Speech Platform Runtime 11 Licence.rtf x64_SpeechPlatformRuntime\SpeechPlatformRuntime.msi
Date Published: 12/30/2011

Can't tell any difference.

[MannequinFan]

Yeah I didn't notice much difference either, and it kills the dictation feature in VA.
I also lost my training for all the "colorful" words.
Fuck you became Thank you, lol. So switching back to version 8.

[Begog]

Yeah I didn't notice much difference either, and it kills the dictation feature in VA.
I also lost my training for all the "colorful" words.
Fuck you became Thank you, lol. So switching back to version 8.

Yeah, that is my biggest complaint when using WinVR for doll speak. It wants to G-rate (pronounced convolute) the conversation. My Kari would kick ass if I could just TALK to her, but I keep having to say "delete message," then try again, and eventually get frustrated and go back to typing, not to mention having to say "enter" each time when it does get it right. Were it not for this problem, my Kari would rock, and I like being in Windows on a laptop.

On a side note, I do not lose the dictation feature on Kari. Could be a setting.

[MannequinFan]

Yeah, that is my biggest complaint when using WinVR for doll speak. It wants to G-rate (pronounced convolute) the conversation. My Kari would kick ass if I could just TALK to her, but I keep having to say "delete message," then try again, and eventually get frustrated and go back to typing, not to mention having to say "enter" each time when it does get it right. Were it not for this problem, my Kari would rock, and I like being in Windows on a laptop.

On a side note, I do not lose the dictation feature on Kari. Could be a setting.

You'll have to train the Win VR to recognize the dirty words.
You may need to go back to the standard SAPI VR - don't think there's a way to train platform 11.
11 doesn't even show up in my control panel, so I can't select it to train.
Here's a comparison of two systems...
https://msdn.microsoft.com/en-us/library/jj127858.aspx

Unfortunately the dictation issue is a known thing with VA - they have it noted on the website.
With VA, the recognition problem is not quite as bad because it only needs to match keywords instead of analyzing an entire sentence. You can use wildcards in spoken commands (*command=ends with, *command*=contains, command*=starts with).

[Begog]

I hear ya bro, but that brings us back to the original problem I had to start with - WinVR sucks! It's wrong half the time, with or without profanity. It just sucks, and the upgrade really didn't help matters, and the older version is no better. Also, the older version had some weird problem where words were popping up on the screen without me saying anything and the TV or Kari not on in a quiet house. Not sure where that was coming from, but I hated it. When I first installed 11, it fixed that, so I'm not going back. We have probably hijacked this thread enough. If you got any more good ideas, please PM them to me and let this thread get back to philpw99's Doll33 app. We had one somewhere on this very thing a while back.

[philpw99]

I know, I should probably stick to 3d printing, lol, but I noticed you were giving Dialogflow another look see?
Just thought these examples might give you some ideas.
...

Dialogflow and Google will be in the next version of doll33. I already reserve the action tags for them. It's just the current version is still full of bugs.

philpw99's app, while new, shows promise. The basic premise checks all the boxes. If he goes in a direction I don't care for, I may even develop my own. We shall see.

Well, Begog, if you want to develop a fork, I don't mind sharing all my source code with you. The reason I don't put it in a git, is that I don't want everyone from Internet have the code, but for a special few ones like you, I don't really mind, and my coding skill is kinda bad, so the code might be hard to read.

And of course I hope my direction can be align with your direction, I hope my app will have both the features you want, and the features I want.

Meanwhile I am a little lazy on the weekend, and much work to do this week. So the progress is slow now.

We have probably hijacked this thread enough. If you got any more good ideas, please PM them to me and let this thread get back to philpw99's Doll33 app. We had one somewhere on this very thing a while back.

Not minding a bit. Hijack this thread away! I want to see the different implementation as well.

[Begog]

Obviously, philpw99 is a team player, and me and Mannequinfan are long time team mates.

Google Access: For a personal assistant, this is a good thing. For doll chat, bad. Here's why:
Case in point: the Harmony app.
As a chatbot, Harmony is extremely limited. Anything not in her database (99% of everything) gets Googled. The impending answer is usually 10-25 LINES long, giving you a very lengthy definition/explanation/encyclopedia answer to a simple question. Your girlfriend doesn't sound like that, and neither should your doll. I (and many others) complained about that, and it has been said that the next version will have the capability to turn that off. Googling everything also severely drains the battery, as does having to access the cloud for each response, making Harmony an absolute battery HOG. Needing the cloud just to function (you have to sign into their server just to run the app) is a severe drawback and privacy liability. I guarantee you Abyss could bring up everything you said to Harmony, and so could a determined hacker. Damn that. Kari may be dumber, but she's free of all that, and her answers don't sound like Hal9000. So, if you add Google access, I suggest you include the ability to turn it off when not in "personal assistant" mode. I realize the conversation choices are currently quite limited in Doll33, but that is where I can really help you, because I am a chatbot programming BEAST. You can see evidence of that by having my PF chatbot Kwench chat with other chatbots, and almost always destroys them, even when they have a much higher rating. All I need is a good set of chatbot tools and decent VR.

Well, Begog, if you want to develop a fork, I don't mind sharing all my source code with you. The reason I don't put it in a git, is that I don't want everyone from Internet have the code, but for a special few ones like you, I don't really mind, and my coding skill is kinda bad, so the code might be hard to read.

That is mighty nice of you. Not sure I can do much with that, but willing to try. At least you have coding skills. While I program high-tech machines at work, I do it in ladder logic, unrelated to Android App-building. While I can build computers, and make them run windows or Linux, the only Android experience I got is my phone, tablet, and Kodi box. But, I have programmed a shitload of chatbots, and know exactly what features they need to seem intelligent. Deep learning? Hell no, but you need extreme processing power for that, something a doll won't have for a long time.

[philpw99]

Obviously, philpw99 is a team player, and me and Mannequinfan are long time team mates.

Google Access: For a personal assistant, this is a good thing. For doll chat, bad. Here's why:
Case in point: the Harmony app.
As a chatbot, Harmony is extremely limited. Anything not in her database (99% of everything) gets Googled. The impending answer is usually 10-25 LINES long, giving you a very lengthy definition/explanation/encyclopedia answer to a simple question. Your girlfriend doesn't sound like that, and neither should your doll. I (and many others) complained about that, and it has been said that the next version will have the capability to turn that off. Googling everything also severely drains the battery, as does having to access the cloud for each response, making Harmony an absolute battery HOG. Needing the cloud just to function (you have to sign into their server just to run the app) is a severe drawback and privacy liability. I guarantee you Abyss could bring up everything you said to Harmony, and so could a determined hacker. Damn that. Kari may be dumber, but she's free of all that, and her answers don't sound like Hal9000. So, if you add Google access, I suggest you include the ability to turn it off when not in "personal assistant" mode. I realize the conversation choices are currently quite limited in Doll33, but that is where I can really help you, because I am a chatbot programming BEAST. You can see evidence of that by having my PF chatbot Kwench chat with other chatbots, and almost always destroys them, even when they have a much higher rating. All I need is a good set of chatbot tools and decent VR.

While I can build computers, and make them run windows or Linux, the only Android experience I got is my phone, tablet, and Kodi box. But, I have programmed a shitload of chatbots, and know exactly what features they need to seem intelligent. Deep learning? Hell no, but you need extreme processing power for that, something a doll won't have for a long time.

Alright then, I believe I can just keep sharing you with the Filemaker tool in the future, and it should be a much better tool in creating chatbot lines.

What you said about Google cloud is something new to me. I thought it will be great to have DialogFlow and CloudTextToSpeech running. Anyway, I think I will still try to program those, and see how it goes. The "personal assistant mode" will be done by voice, like "can you ask google ...", otherwise it will not access the cloud for answer.

I believe Doll33 will be greatly enhanced by your chatbot programming skills. I just need to do more programming and fix the bugs myself. I have literately zero experience in making chatbots.

One more thing, it's about the Google Cloud TextToSpeed API, seems it can speak like a real person in different voices. What do you think about this? It's free for 1 million characters per month, so it's basically free for my app. What do you think about using this instead of the android robotic voice?

[Begog]

Alright then, I believe I can just keep sharing you with the Filemaker tool in the future, and it should be a much better tool in creating chatbot lines.

What you said about Google cloud is something new to me. I thought it will be great to have DialogFlow and CloudTextToSpeech running. Anyway, I think I will still try to program those, and see how it goes. The "personal assistant mode" will be done by voice, like "can you ask google ...", otherwise it will not access the cloud for answer.

I believe Doll33 will be greatly enhanced by your chatbot programming skills. I just need to do more programming and fix the bugs myself. I have literately zero experience in making chatbots.

One more thing, it's about the Google Cloud TextToSpeed API, seems it can speak like a real person in different voices. What do you think about this? It's free for 1 million characters per month, so it's basically free for my app. What do you think about using this instead of the android robotic voice?

Sounds like a plan. Once the bugs are fixed, I'll need to download the new version. Will it overwrite the dialogue I already have in place? That's a concern you'll need to address. If so, I'll not add any more to this version since it's not using it anyway. At some point, I'll need a way to upload the chatbot back to you. as for:
"One more thing, it's about the Google Cloud TextToSpeed API, seems it can speak like a real person in different voices."
I'd be amazed if it actually did, because robots don't do voice inflection and accents. Even Sophia with her advanced software and processing power still has a robotic accent. However, it may be better, and certainly worth a try. It probably sounds more like Siri or Alexa, and may have other functionality drawbacks, but it can't hurt to try. Might be an improvement. Either way, Android VR text-to-speech works quite well, FAR better than the WinVR I have on Windows7.

[MannequinFan]

Here's the Google TTS page where you can demo their voices...
https://cloud.google.com/text-to-speech/

I think they are geared mainly for business applications, but they do sound pretty good.
Just not very sexy. Don't think I'd want to give up my sultry Ivona British Amy voice.

[Begog]

Thanks MF. I will check them out. I doubt I want to replace Kari's sweet voice either.

Phillpw99: If you wish to offer Doll33 with preset personalities, multiple profiles should be near the top of your list of things to do. Here's why:
Most chatbot programmers want a clean slate - a fully functioning but bare bones basic mostly empty bot to be molded into a unique personality. However, many doll owners will not have the skills to create a decent bot, or be willing to invest the many hours it takes to create one, let alone learn the many nuances of chatbot conversation and the Doll33 app. Also, some will prefer a girl next door type personality for their doll, while others (like myself) want a synthetic slut. So, I am making these 3 recommendations:
1) Need to be able to create at least 5 profiles, preferably unlimited.
2) There should be a basic personality for those who wish to program, a sweet, loving personality for those who want a lover, and a synthetic slut personality for the sex doll enthusiast.
3) Upgrades should not wipe out many hours of chatbot programming. No one will want to upgrade to a higher version if it means losing their doll's personality they worked for hours to achieve. If this cannot be done, then your website should be clear on this before anyone overwrites.

[philpw99]

Thanks MF. I will check them out. I doubt I want to replace Kari's sweet voice either.

Phillpw99: If you wish to offer Doll33 with preset personalities, multiple profiles should be near the top of your list of things to do. Here's why:
Most chatbot programmers want a clean slate - a fully functioning but bare bones basic mostly empty bot to be molded into a unique personality. However, many doll owners will not have the skills to create a decent bot, or be willing to invest the many hours it takes to create one, let alone learn the many nuances of chatbot conversation and the Doll33 app. Also, some will prefer a girl next door type personality for their doll, while others (like myself) want a synthetic slut. So, I am making these 3 recommendations:
1) Need to be able to create at least 5 profiles, preferably unlimited.
2) There should be a basic personality for those who wish to program, a sweet, loving personality for those who want a lover, and a synthetic slut personality for the sex doll enthusiast.
3) Upgrades should not wipe out many hours of chatbot programming. No one will want to upgrade to a higher version if it means losing their doll's personality they worked for hours to achieve. If this cannot be done, then your website should be clear on this before anyone overwrites.

Begog, Doll33 have the design of switchable profile since day one. A lot of care was put in so switching to another profile will be easy. The profiles will not be created by normal doll users. The best they can do is some limited editing, and it's not recommended. Only few people like you and me will use the filemaker program to edit the profiles, which will be much faster and easier. Filemaker is great in doing this kind of thing. I am good at Filemaker. It's easier for me to implement those filtering and searching(not so easy in the phone). And we can use Audacity to record and edit the audio, which is much better than the built-in mic of cellphones.

1) The profile numbers will be only limited to your phone storage space. In the phone, you can go to this page, choose a profile to download, then the app will unzip the profile, and put everything in their own folders(programming this feature now). Once the profile is unzipped, switching to it should be super fast.

2) Any way you want, my man, any way you want. Just put the doll's personality in the description of "profile.json". Right now my website has limited way to list profiles. I don't know how to enable visitors submit and post a profile. Maybe TDF will help me to have a sub-forum and host those profiles, like the doll albums? I want to find a way that everyone can contribute, not just me.

3) They are just folders in the app's document, so when the app is upgraded, the profiles should still be there. However, if you remove the app, all profiles are lost. For you dear chatbot programmers, you will have the backup folder, and zip files in your laptop anyway.
I don't know why upgrades will wipe out profile data, but thanks for the warning, I will be careful with it.

[MannequinFan]

I'm not sure how big the profiles are, but you might be able to post them here by just changing the file extension to ".PDF". Then after download change the extension back. I've done that a lot for posting print files (.stl). You can also do that with .zip files.
Of course this was with a computer not a phone.

Any chance of this working with a Windows PC in the future?

[philpw99]

I'm not sure how big the profiles are, but you might be able to post them here by just changing the file extension to ".PDF". Then after download change the extension back. I've done that a lot for posting print files (.stl). You can also do that with .zip files.
Of course this was with a computer not a phone.

Any chance of this working with a Windows PC in the future?

This is the thing. I want a place in web that everyone can unload their profile zip files in it. Changing the extension is possible, but I don't really like the work-around. Really prefer a easy way for users.

Uh... I don't think you need this app for Windows PC, which Voice Attack is doing a great job. Plus android has some things that's not in the Windows platform, like accelerometer, google cloud AI and TTS... Seems this app's best chance is for those cheap phones or boxes. Oh, also you can just run an android simulator in the windows. The speed should be OK.