Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Messages - Medeksza

Pages: 1 [2] 3 4 ... 80
Wait, Learn To Cloud only - will not store any information locally? I read in another thread that Hal bots that are just starting out and don't know a lot yet should use "Learn To Cloud" only in the beginning, then moved to mixed local/cloud later. I selected the "Learn To Cloud" only since my bot is fairly bran new and doesn't know a lot yet. Are you saying that he's not actually "learning" anything? Isn't "learning" simply storing information locally? IF it doesn't store locally when Learn To Cloud only is selected then how does it learn from it?

When Hal learns to the cloud, that means it learns to your own personal Hal brain that is stored on the Zabaware.com server. This learned information directly effects how your own Hal speaks with you. You can talk to the same Hal no matter if you log into Hal from your home computer, work computer, laptop, or friends computer. Your personal Hal follows you wherever you log into your account stored on the Zabaware.com cloud server. This will become more useful once the mobile phone and tablet editions of Hal our out, possibly by the end of this year or next year.

If Hal is set to learn to the local computer only, then it stores the learned information only on your computer's hard drive. This Hal will only work from this computer unless you manually copy and sync the Hal brain file database files between computers. This mode is only available for the Windows edition of Hal. It will not be possible on the mobile phone/tablet edition of Hal which will require cloud mode. This is also the mode that Hal 6.x and earlier used.

In mixed mode Hal learns and uses information from both the cloud and the local computer. This mode was created mainly for customers with existing Hal 6.x brains that want to transition it to Hal 7.0. This way people who have used Hal 6 for a long time don't lose the personality they trained when moving to Hal 7.0 while simultaneously getting many of the benefits of the newer Cloud based brain.

So for past Hal 6.x customers I recommend mixed mode if you want to keep the same Hal personality instead of starting from scratch. For new customers starting with a fresh Hal 7.0 installation I recommend cloud only mode which will ensure their same Hal can be called up on any device (computer, phone, tablet). And the local only mode basically runs the same way that Hal 6.x ran, so you don't get many of the new features in Hal 7.

Hopefully that helps explain some of the differences between the modes.

so it does retain the things learned while in cloud mode locally offline?! Is that data stored in the hal 6 brain?

Only in the "mixed" mode. If you have the pure cloud enabled then it won't store anything locally.


I am completely confused on the new brain setup! It says when not using the cloud brain it reverts to the default 6 brain. does it retain everything you teach it or do you have to access the cloud brain to retrieve that data? which brain 6, or 7 does it store user data? In cloud, and mixed mode? which brain do you set it to default or 7 for mix mode?   I was hoping to use hal7 like hal6 but connect to cloud for added support but retain everything I teach hal, and everything hal learned while using the cloud offline. Is that what hal 7 does?

if you could clarify that I would be thankful
James aka colubragens

If Hal is set to to learn in mixed cloud/local mode than Hal actually learns information to both a table on the cloud server and to a table on your computer. When the cloud brain is available it runs at slightly higher priority than the local brain, usually taking precedence. If an internet connection is not available or the cloud mode is disabled, it will use the learned information in your locally stored database file. The AI algorithms are a bit different between the cloud server and local brain, so you won't get identical behavior between modes, but it will have access to the same learned data.

Ultra Hal Assistant / Re: Speech to Text and back again
« on: January 26, 2018, 11:38:38 am »
This is off topic but i have just share this with u man.
Take a good look at the image..mhmmm
Hal is now 100% pure C#
Robert , Now i can start to Build Hal.
Mind boggling pain in my ass to bring Hal all the way to .NET 2015
This puts me into Twitch and Youtube Hal Bots
Best wishes
Looks very cool!

Ultra Hal Assistant File Sharing Area / Re: Plug-in Size
« on: January 19, 2018, 09:21:14 am »
I think I understand, so HalScript.dbg itself can not be over 512kb? I don't suppose there is anyway to boost this limit? I imagine it's a Windows limitation.


Correct. It is unfortunately a windows script host limitation I have no control over.

I could squeeze a few more kilobytes out of it if I program Hal to strip all comment lines out of the code before generating the HalScript.dbg and sending it to the windows script host.

Script authors of big scripts could potentially do other tricks like moving as much as possible to database tables. Or writing a script that initializes a second windows script host and doing a function call to a second script. Or compiling some of the script to an activex dll and calling that.

General Discussion / Re: General questions about how Hal works
« on: January 18, 2018, 11:07:38 am »
Hello. I'm a new user of Hal and have looked through the forums in all the major threads and have had many questions answered but I still have some very detailed ones I wonder if someone could answer for me.

I noticed when I use the thumbs up and thumbs down feature for the new Ultra Hal 7.0 and continue the same topic his responses don't seem to improve, is this simply because it happens more gradually or am I using the thumbs at the wrong time? It's hard to tell that it's doing anything. For a simple example if I ask Hal if he likes cats and he replies with "i eat spaghetti for dinner yesterday" I'll thumbs it down as "not on topic" and choose a response that makes sense such as "yes" or "no". Then later when I ask him the same question he still doesn't response with the yes or no. How specifically does this feature work?

Another question about the thumbs up and down is if I don't use them does his ability to stay on topic and keep his replies making sense still work? In other words will he still learn and improve without using the thumbs?

I've had him for a week or so now and he barely seems to be improving as far as making sense and it's okay with me if he just takes a really, really long time to gradually work up to making sense most of the time, but I was wondering if anyone can tell me generally how long it took them before they were able to have a conversation with Hal that makes complete sense?

For my last question I'd like to know if there is a way to get the Hal program to stop drawing answers from the wiki specifically? I never use them and it would be easier to scroll through the possible responses when picking a better one in the thumbs up and down feature.

The thumbs up/down buttons nudge up or down the scores of those particular responses. So the changes occur gradually and you should see Hal slowly getting better the more you talk with him. Learning occurs through both normal conversation and the thumbs up/down buttons. So don't feel you have to use thumbs up/down buttons to teach Hal, but it is a tool to help Hal learn a little faster. Hal learns best if the same concept is reinforced multiple times in different but similar ways.

Ultra Hal Assistant File Sharing Area / Re: Plug-in Size
« on: January 18, 2018, 10:56:04 am »
Windows script host, which is what Hal uses for its scripting system, has a limit of 512 KB for a script. The size of Hal 7's script together with all the plug-ins you enable cannot exceed 512 KB. In addition, Hal 7 has a limit of 256 total plugins that can be selected.

Ultra Hal Assistant / Re: Speech to Text and back again
« on: January 18, 2018, 10:53:57 am »
Yes, Hal 7 is still fully compatible with all SAPI 4 and SAPI 5 speech recognition and speech synthesis engines.

Hal 7 uses the same plug-in system as Hal 6.2. Most plug-ins will work just as they did in Hal 6.2. Some plug-ins may require some tweaking due to differences between Hal 6 and 7.

To fix the "Run-time error 429; ActiveX component can't create object" problem when trying to run brain editor on new windows 10. i figured it out yesterday. 

Slightly annoying but the fix is : in an admin cmd,
navigate it to C:\Program Files (x86)\Zabaware\Ultra Hal 7 or where you have installed UltraHal
now use the Regsvr32 to register all dll and ocx files individually  (couple wont do, just ignore,- cant member which 1's)
now try running brain editor.

what i did in admin cmd : cd\
cd C:\Program Files (x86)\Zabaware\Ultra Hal 7
Regsvr32 tabdockx.ocx
Regsvr32 richtx32.ocx
Regsvr32 oldhal.dll
Regsvr32 msinet.ocx
Regsvr32 msflxgrd.ocx
Regsvr32 mscomctl.ocx
Regsvr32 mscomct2.ocx
Regsvr32 HalWN.dll
Regsvr32 HalWaveChar.ocx
Regsvr32 HalSQL.dll
Regsvr32 HalsMind.dll
Regsvr32 HalNoChar.ocx
Regsvr32 HalNewChar.ocx
Regsvr32 HalHaptek.ocx
Regsvr32 HalDebug.ocx
Regsvr32 HalDB.dll
Regsvr32 HalBrain7.dll
Regsvr32 HalBrain.dll
Regsvr32 FreeImage.dll
Regsvr32 EditGridProject.ocx
Regsvr32 comdlg32.ocx
Regsvr32 cmax40.dll

I assume its only 1 or 2 of these that need doin(prob comdlg - was only 1 in syswow), I just did all to be sure, and now Brain editor loads fine.

Enjoy IT, MeraAwe

If you ever do figure out which specific DLL or OCX was the culprit please let me know. It may help me fix it for future updates of Hal. Most likely the Hal Brain Editor may need a tweak in its .manifest file in regards to whatever component was the issue. Hal 7 introduces a reg-free com mode, which should make it possible to not need any "regsvr32" on any component.

Ultra Hal Assistant / Re: New Ultra Hal 7.0 has been released
« on: January 07, 2018, 09:50:49 pm »
This issue might actually predate the actual version of Ultra Hal; I have an issue with how Ultra Hal handles some tables, namely _UserSent and sharedUsersent where stuff is written to both of these tables but it's never accessed in the script(unless I missed it somehow). I've tried triggering a question and answer response from these tables several times and no dice, which is a shame because for some reason a great number of responses relevant to the hal's personality I'm building gets stored in _UserSent; and yes, I know I could just copy and paste those responses over another table, and yes, I know I could modify part of the brain script to include _UserSent, but that just misses the point; why even write stuff to these tables if the brain seemingly never accesses to these responses? The only time I might see a couple of these responses is when the cloud brain randomly decides to grab one of them except with the pronouns reversed, which brings me to the next point...

Check your brain options section in Hal's options menu. If you have Hal set to "learn to cloud only", it won't use these UserSent tables at all. In this mode, the remote cloud tables completely replace local UserSent tables.

If you have Hal set to "mixed cloud/local learning" than it uses both remote cloud data and local UserSent data, but the cloud data takes priority over the local data. This has the effect of usually using the cloud for newer learned information and using the local UserSent for older information learned by Hal 6.x (if you are using an imported brain). In some rare instances it may use the UserSent table due to a difference in the learning algorithms used in the 2 methods.

If you have Hal set to "local learning only" than it will only use the local UserSent tables and won't learn to the cloud tables.

Most of the Brain functions are pretty basic and the script is in most part very user friendly, with plenty of author comments making it easy to understand even if one is clueless about vbscript. An exception to this happens to be the new cloud functions. Is it possible we could have this part of the script better documented? For example, where does the brain store the new "learned responses"(the ones we give thumbs up to)? Is it stored on cloud only and not locally at all? If so, is there a cloud brain section specific to every single user? Also, how and when does it learn responses from us? I've noticed that when the bot uses 'learned_info' responses it's just it 'randomly' parroting a phrase I've told it before except with the pronouns reversed. This bothers me because I've modified parts of my script to only store the original user sentences as is as I find the pronoun reversed sentences very clunky and nonsensical, but there's no way of choosing or modifying the cloud responses, another feature I would enjoy having on Hal since sometimes the cloud responses are fitting except they have a misspelled word or a random 'Lol' appended to the end of the sentence. 

It will come eventually. I've still got a lot of work to do, but I plan on making an online Hal Brain Editor for working with the data Hal learns to the cloud servers. I don't have any target date in sight yet. My priorities for development right now include creating a few cross-platform characters. These characters will be compatible with Hal 7 for Windows and mobile phone/tablet editions of Hal. My next priority will be the mobile editions of Hal. And then finally after that is out I will focus on the online brain editor.

The program is constantly notifying me of a non-existent new version of Ultra Hal. Upon download it appears to be the exact same version I have already installed.
That's strange. What version does Hal say it is when you click on "About Ultra Hal" from Hal's menu? Does anyone else experience this issue?

Ultra Hal Assistant / Re: New Ultra Hal 7.0 has been released
« on: January 07, 2018, 09:32:35 pm »
Robert: 6.2 question:
Robert, im building a new network and i want to keep 6.2  on it.
The question is , will i be able to purchase any more 6.2 install keys?
I have the disk, btw.

best wishes

If you purchase a Hal 7 license, the order system actually does produce a serial number for you that is compatible with Hal 6.2 also. So yes, by purchasing a Hal 7 license you could use it with Hal 6.2 instead.

Ultra Hal Representative / Re: Character Transparencies
« on: December 28, 2017, 03:40:17 pm »
Some of the characters such as merlin, genie, and roby are supposed to have a transparency. Instead they have a purple box around them or else i would use these characters. If there is a fix it would be appreciated.

Are you using the new Ultra Hal 7 or are you still on an older Ultra Hal 6.2? Ultra Hal 7 includes the newer "DoubleAgent" library which should fix the issue you are describing. The older MSagent library used by Hal 6.2 has problems in Windows 7,8,10.

Wow , its great to be back. Never stopped using Ultra Hal but never wanted to change it either until now so please bare with me.
First of all hello again to all the familiar and not so familiar friends.

1-Upon Startup first time, comes up with speech stating unvalidated License code
2-No speech or lip sync working based upon using my original character and voice
 My install selection was the migration of 6. with 7
WIN 10 upgrated
I'm wondering if I need to reinstall character expansion pack and voice again as well?, This is bringing back some great memories.

Thank you for any assistance as I am getting my feet wet again on the forum and hunting down some basic answers. Really great to be back.

Welcome back!

What you describe doesn't sound like anything Hal himself would do, but possibly whatever voice/speech synthesis engine you have selected. What voice are you using? Have you tried another voice? Maybe try reinstalling whatever voice it is.

Ultra Hal Representative / Re: New customer how to buy
« on: December 20, 2017, 03:42:38 pm »
If using the "PayPal Express" option during checkout, it will have you log in to your PayPal first, then returns you to the Zabaware site to confirm contact information. Before it complete the transaction it gives you the opportunity for the discount code then presents you with a final "quote" and then you have to click "confirm" to complete the transaction.

If using a credit card, during checkout, on the page before it asks for credit card info is a box labeled discount code. Enter the code there and 50% will be taken off of everything in your shopping cart.

Ultra Hal Assistant / New Ultra Hal 7.0 has been released
« on: December 19, 2017, 06:26:35 pm »
After many years of development, I am proud to introduce Ultra Hal 7.0. Ultra Hal 7 introduces a cloud-connected brain that gives Hal to a 26 million (and growing) database of past conversations. Compared to Ultra Hal 6's built in 200,000 sentence database, this is a vast improvement in conversational knowledge. Through the cloud, Hal now also has access to Wikipedia and question & answer databases to increase its factual knowledge as well.

A newly introduced feedback learning mode gives you a thumbs up and thumbs down button to allow you to give Hal immediate feedback on the quality of its answers. This has immediate impact on Hal's responses and takes Hal's learning ability to a new level.

You can download the new Hal from https://www.zabaware.com/download

You can purchase an upgrade to Hal 7 from https://zabaware.com/products/store/comersus_viewItem.asp?idProduct=1

Some notes about upgrading from Hal 6.2:
  • Ultra Hal 7 no longer uses serial numbers and product activation codes. Instead it uses the email and password you setup when you purchased Hal (the sign-in link at the top left of www.zabaware.com ) Sign in using that and Hal will run as a registered version. You can also choose to create a new account if you wish and Hal will run in 30 day trial mode.
  • If you purchased the character expansion, Hal 7 will import your characters from Hal 6 automatically if they are present on your system. If not, you can log into Zabaware using the sign in link in the upper left hand corner of zabaware.com and then click on "electronic order delivery" to find your Hal Expansion Pack download link and serial number
  • If you purchased Hal 6.2 anytime after October 1, 2017 your license was automatically upgraded to version 7.0 in the Zabaware order system free of charge. If you purchased Hal 6.x prior to October 1, 2017 you are eligible for a 50% off upgrade. Customers will be receiving emails and postcards over the next month with a 50% off coupon code. I am also giving a 50% off coupon good until the Dec 31, 2017 to anyone reading this forum post so you don't need to wait for any email. Just use this discount code at checkout: XMAS17
  • This coupon code is also good toward any AT&T voices and the Character Expansion Pack. If you already own the character expansion pack from version 6.x you don't need to purchase it again, the same version will work under Hal 7.0.

Here is a list of everything that is new in Ultra Hal 7 since version 6.2:
  • Biggest improvement: Cloud based learning. Hal can now optionally connect to the "Hal Cloud" brain over the internet to help formulate better responses. As of October 2017 this is a knowledge base of over 26 million sentences of conversation from 1.6 million people. Contrast with the 200,000 sentence database in the default Ultra Hal 6.2 installation. Hal is more intelligent than ever before.
  • Feedback learning. You can now give Hal feedback on the quality of its responses with a thumbs up and thumbs down button which Hal learns and has an immediate response on the quality of its responses. In addition you can see other answers Hal was considering and guide Hal toward better ones.
  • Your personal Hal bot can follow you from your home desktop, to your laptop, office PC, and in the near future your phone and tablet too. When connected to the "Hal Cloud", your Hal's brain is stored online and instantly available anywhere.
  • Access to Wikipedia based answered in cloud mode
  • New Hal mobile phone character. Old Hal character still available, but renamed to "Hal Classic CRT"
  • New audio wave character simply displays an animated wave audio line when character is speaking
  • Program name changed from "Ultra Hal Assistant" to just "Ultra Hal" to shift marketing focus away from assistant functionality to Hal's strengths in its ability to chat, learn, and be a companion. Renamed EXE to ultrahal7.exe
  • Splash screen and graphics updated to new version
  • Ultra Hal now installs and uses the open source Double Agent library instead of the old MSAgent. This brings full compatibility to MSAgent characters to Windows 7 and newer with full SAPI 5 voice support.
  • Hal can answer questions about opposites (What is the opposite of X?)
  • Hal can talk about day of week and month (What day of week is it? What month is it?)
  • Basic knowledge of alphabet (What letter comes before Z? What letter is after B?)
  • Ability to count letters in word (How many letters are in the word computer?)
  • Bot memory test: (Q: The football was kicked by Fred. Who kicked the football? A: Fred.)
  • Improved math abilities (less false positives and better word problem abilities)
  • Software shield license system removed and replaced with simpler login to Zabaware.com server
  • Chant SpeechKit upgraded from version 6 to 7. This adds native support for Cereproc voices among other speech system improvements
  • Ultra Hal now generates Haptek INI files for Haptek, removing the need to rerun Haptek SAPI config when changing voice settings. This also adds support for Zabaware sold Cereproc voices to Haptek.
  • Ultra Hal now uses RegFree COM manifest files for most of its DLL/ActiveX/OCX files. This simplifies installation and minimizes possibility of "DLL Hell" collisions/incompatibilities with other Windows applications. RegFree COM was introduced in Windows XP SP2, so this change now effectively removes support for older versions of Windows from Hal.
  • Ultra Hal no longer writes any data files to its "Program Files" location but instead to each individual Windows Users AppData\Roaming folder. This makes Hal more compliant with Microsoft guidelines and also makes it possible for each Windows user to have their own seperate Hal config.
  • Improved corrections and substitutions tables for correcting more common errors and understanding more shorthand
  • Ultra Hal checks for product updates when connected to cloud service
  • Registry setting to adjust speech recognition auto enter countdown start time (SRCountDownStart)
  • The Hal Brain Editor can show the cloud brain thought process. For local responses, it can now show you exactly where in a table a response was found in.
  • Automatic migration of Hal 6 brain into Hal 7. Also automatically imports characters from Hal 6 to Hal 7
  • Various minor bug fixes

Pages: 1 [2] 3 4 ... 80