Page tree

You are viewing an old version of this page. View the current version.

Compare with Current View Page History

« Previous Version 8 Next »

General

Where can I get the toolkit?

Please see the Support section on how to obtain the toolkit.

How do I install the toolkit?

Please see the tutorial on how to install the toolkit.

How do I run the provided example?

Please see the tutorial on how to run the example.

Is the toolkit open source?

Some components are open source, while others are binary only. We might release the source code of these in the future, though. SmartBody has its own SourceForge page which you can access here.

Which platforms does the toolkit support?

The toolkit as a whole is targeted for the Microsoft Windows platform, including XP and Windows 7, both 32 and 64 bit. Some components are multi-platform, most notably the NPCEditor which drives the characters, and the Launcher and Logger. Some other components could be ported to support other platforms, like SmartBody and Ogre.

I know of this great character at ICT that uses toolkit technology. Can I get access to the data driving that character?

With the toolkit we are aiming to provide a technical platform allowing researchers to more quickly create their own virtual humans. Whether project specific agents will be made available to the public will be determined on a case by case basis, but will usually fall outside the scope of the toolkit.

Installation Process

Why do some of the 3rd party installers not start?

Not all 3rd party software come with their own installers. Some might just copy over files to the correct location.

The installer tells me I should do something with the Registry; how do I do that?

If you installed the MS SAPI voices and you are running Windows XP 64 bit, some changes in the registry are required to be able to use the installed voices. You can do this by double clicking "Microsoft-Voices.reg" in "C:\vhtoolkit\scripts". You can check to see if the changes worked by going to the Speech settings in Windows (Start -> Control Panel -> Speech). Under "Voice Selection" you should have several options, including "Microsoft Mike".

Running the Example Character Brad

Why can't I run the Launcher? It came up fine after the installation

It might be that the executable used to start the Launcher can not be found. Please make sure that javaw.exe is part of your path. The location is usually "C:\Program Files\Java\jre1.6.0_10\bin" or "C:\Program Files (x86)\Java\jre1.6.0_10\bin". The path is defined in your Environment Variables; right-click on My Computer, go to the Advanced tab and click Environment Variables.

Why do I get many errors and are all the Launcher rows orange?

Orange means that the Launcher has tried to launch a component, but hasn't gotten a confirmation message back indicating it is up and running. This is likely due to not running the ActiveMQ service. Please make sure it is running (Right click My Computer, select Manage, Services and Applications, Services). If it is not running, turn it on (the Play button at the top). If it is not present, please install using the toolkit installer with 3rd party software.

Why won't the renderer load?

  1. Be sure to install all 3rd party software that comes with the installer. Gamebryo needs specific versions of DirectX to run, which comes with the installer.
  2. If this doesn't work, make sure that you have the latest video drivers installed.
  3. Make sure that your video card supports shader model 3 or higher. This is usually the case with newer dedicated video cards. Integrated video cards might not work.
  4. See if the problem happens with both Gamebryo and Ogre. If Ogre works in OpenGL mode, it might be a DirectX problem.

Why don't I see Brad, even though a renderer (Unity, Ogre) is running?

Is SmartBody running? You can see this by looking at the SmartBody row in the Launcher. If it's green, it is running, if it is orange, its status is unknown (usually, the Launcher tried running the component, but did not get confirmation it is actually up and running), and if the color is neutral it means it is not running. Make sure the renderer is up and running before you start SmartBody. Please also make sure the renderer is completely up and running before you start SmartBody. If you use the Run All button, you might have to kill SmartBody and launch it manually.

Why don't I hear Brad?

See if Brad is moving his lips. If so, make sure your volume is not all the way down, that it is not muted, and that you know where sound should come from (i.e., desktop speakers, or a head phone that has just been plugged in). If Brad is not moving his lips, make sure the Text To Speech module is running. If it is, look in its console window for any error messages that might give you an idea of what is going wrong. The TTS executable is called "MsSpeechRelay.exe".

Why did Brad stop reacting to me?

It might not have the correct answer for you, especially if you have tried asking the same question a couple of times. There might also be a delay in some of the modules. Try waiting a minute after trying again. If there is no response at all, make sure all modules are still green in the Launcher. If a module that is checked is neutral, try launching it. If one of the checked rows is orange, the module might have crashed. Find the associated window or process and kill it manually, then restart that module. If all checked rows are green, try to restart SmartBody, the NVBG and the NPCEditor. If this doesn't help, kill all modules and restart them. As a last resort, try rebooting your computer and start all modules.

SmartBody

NPCEditor

I am asking my character a predefined question, but he doesn't respond. How can I solve that?

There could be a number or reasons why this is happening:

  1. First, see if the answer you expect comes up on the Chat panel in the NPCEditor. If it does, you know the right answer is selected by the NPCEditor and that something further down the pipeline might be going wrong; see points 7 and higher for more options. If the answer does not show up, continue with the other steps:
  2. Check that the question has an answer associated with it. You can see this by looking at the column that has the hash symbol at top ('#'). This shows you the number of answers linked to this question.
  3. Check that you are asking the question exactly as it appears on the left hand side of the Utterance panel. Even though the NPCEditor can handle questions that are only similar to predefined questions, this is a good step to test if something is wrong.
  4. Make sure the ID for the human speaker (the person interacting with the character) is consistent between the NPCEditor and the AcquireSpeech client. This means that the Domain variable on the question (left-hand) side of NPCEditor is either Anybody (i.e., all speakers are accepted), or matches exactly the ID of the human speaker (the Speaker ID as defined in the AcquireSpeech module). You can change the AcquireSpeech speaker ID in the batch file startup script in \tools\launcher-scripts.
  5. Select the question you are asking on the left hand side of the NPCEditor, and order the rows by the strength of the link between questions and answers. You can do this by clicking on the associated column name at the top. This is likely the third column, which has a dash at the top ('-'), in between the error column and the Score column. Click on the column name (the dash) to order all rows. Do the same thing on the right hand side, the answers. On the answer side, select all answers associated with the particular question. First, select a single answer that is linked to the questions you have selected. This will unselect any previous answers. Now, hold CTRL and click all answers that have a link value (this value is usually 6). This in turn will highlight all questions that are linked to these answers. While holding CTRL, also select all the questions that are linked to this selection of answers. You now have selected a set of questions, together with the set of all linked answers. You can scroll through both the questions as well as answers list. Ideally, you only see white and green rows (although any selected row will be blue). If you see yellow rows, it means that the selected sets are not completely linked 1 to 1. Although this is not always necessary, the NPCEditor might have trouble if sets of answers and questions are overlapping. Try creating unique sets of questions, that link to unique sets of answers.
  6. If you have added a lot of new data to the NPCEditor, you might want to retrain it. To do this, go to the classifier, select the domain you want to train (this is the Addressee variable, which corresponds to the character), make sure the 'Test on training data' box is checked, and click 'Start Training'. If you see more than one Addressee, it means you have defined more than one domain on the answer side. This is likely due to the default Anybody domain. If you only have one character, you should usually only have one domain. In order to achieve this, go to the Utterances tab, select all answers (CTRL + A) and select one value from the Domain pull down list at the bottom.
  7. Make sure that all modules use the same character ID (like 'brad' in the example). For SmartBody, this is defined in the sequence file. For the Nonverbal Behavior Generator in the startup batch file, and for the NPCEditor in the Speaker variable (accessible from the Speaker tab).
  8. Make sure that on the answer side, all rows have a Speaker defined. This is the character ID that will be send out to the Nonverbal Behavior Generator (NVBG). By default the speaker will be empty, which results in the NPCEditor sending out a request to the NVBG, which the latter ignores, because it is only listening for request for specific characters.

Why do I hear Brad's voice twice?

  1. Ensure that you have only one NPCEditor open. In the case where you have more than one NPCEditors running and you ask a question, this is sent to both NPCEditors that are running and they both would send a reply therefore duplicating Brad's voice.
  2. If after closing one NPCEditor and that doesn't solve the problem, kill the NPCEditor and launch it again.

How do I delete local or user specific NPCEditor data?

If you have installed previous versions of the toolkit and used it before on your system, then the NPCEditor would most likely have created user setting files on your machine. Once in a while, a change from our side to NPCEditor's settings will need to be propogated to the users, which will require them to delete these user files on their machines so that they don't interfere/override the new settings.

The NPCEditor creates .plist files which it stores in the user directory for Windows. These files contain NPCEditor specific information.
So if you need to delete these user settings from your machine, follow the below steps.

Depending on which OS you are using, these files will be present in the mentioned folders

Windows 7:

If you are using Windows 7, then you should look under

Your Primary Drive:\Users_Your user name_\AppData\Roaming\NPCEditor\people

You should see files with the extension ".plist". These are the local files created by the NPCEditor on your machine to save user settings. Delete these files. If these files don't exist, then it shouldn't matter.

Windows XP:

Under Windows XP, These files should be present under

Your Primary Drive:\Documents and Settings_Your user name_\Application Data\NPCEditor\people

These are the local files created by the NPCEditor on your machine to save user settings. Delete these files. If these files don't exist, then it shouldn't matter.

Once you delete these files, then the NPCEditor will, by default, use the .plist files that the installer has copied within the toolkit folder, with the new settings.

Note: The AppData/ApplicationData folder under both OS's are hidden folders. So you will not be able to see them unless you make sure that hidden folders are visible in Windows Explorer. In order to do this, open any Explorer window and in the dropdown menu on top, go to 'Tools->Folder Options'. Go to the 'View' tab. Under 'Advanced settings', check the button that says "Show hidden files, folders, and drives".
This will display hidden folders in Windows Explorer.

Watson

Unity

What Keyboard Commands Can I Use In Unity?

  • W,A,S,D - camera movement. Q & E - camera up/down
  • J - mouse visibility toggle - mouse look mode
  • L - toggles the fake recognizer text box
  • O - toggles the user's recognized text
  • M - (fullscreen window only) - toggles speech recognition mode. When on, click and hold and talk in the mic. Release to quit talking.
  • X - reset camera
  • Z - show debug statistics
  • I - Toggles sbm character subtitles
  • P - Toggle entire GUI
  • Alt-enter - toggle windowed / fullscreen
  • Escape - quit

What Console Commands Can I Use In Unity?

Use the ~ key to bring up the console.

Commands:

  • q - quit
  • play_intro - Play intro sequence
  • vhmsg ... - Send vhmsg out to system. 'vhmsg sbm ...' for sending a smartbody command
  • set_resolution x y - Set_resolution to 'x' x 'y'. Example 'set_resolution 1024 768'
  • toggle_fullscreen - toggle windowed / fullscreen
  • set_loco_char_name - ...

Ogre

To what extent Is Ogre supported?

Ogre is only provided with the toolkit to show a source code proof of concept integration with SmartBody. It is a very bare bones implementation. The main renderer for the toolkit is Unity.

How can I look around in the world / move the camera?

You can use the A, S, W and D keys to move the camera. Hit the J key to toggle using mouse look (you can move the mouse to look around).

How do I get my mouse cursor back?

Hit the J key, or use ALT + TAB to switch to another window.

How do I quit Ogre?

You can either use the Kill button on the Launcher, or hit the Q key on your keyboard, while in the Ogre window.

Text To Speech

How do I change the voice of a character?

The voice ID of a character is defined in the SmartBody sequence file. For the Brad character this is 'BradVoice', as defined in \data\sbm-toolkit\scripts\init-toolkit-characters-Gamebryo.seq:

set character brad voice remote BradVoice

This line tells SmartBody that it should send a request to a TTS engine for any speech (as opposed to using prerecorded speech), and that the voice ID to be used in that request is 'BradVoice'.

There usually is a distinction between the actual TTS engine, and a small relay program that sits in between SmartBody and the TTS engine. By default, the toolkit uses the MSSpeechRelay module for that. MSSpeechRelay offers the functionality of mapping the requested voice ID to an internal voice ID. In the case of the Brad character, it maps the generic 'BradVoice' to the actual Windows voice 'Microsoft Mike', as defined in \tools\launch-scripts\run-toolkit-tts.bat:

start msspeechrelay -m BradVoice "Microsoft Mike"

By default, MSSpeechRelay will try to use the defined voice, but it will fall back to the Windows default voice if the former can not be found.

If you want to change the voice, you can map 'BradVoice' to a different, already installed voice in 'run-toolkit-tts.bat'. Alternatively, you can map the voice to an unknown voice with the result that the default Windows voice will be used. You can change the default Windows voice by going to the Windows Control Panel (click Start) and selecting Speech. In the Text To Speech tab, select a voice. You can only select voices that you have installed.

See the components section for an overview of voices that are compatible with the toolkit.

Speech Recognition / Text Input

How can I talk to my character?

You can push the Start button on the first tab. This will bring the user to the Recorder tab and will start a new session, in which all speech input will be saved on disk. The toolkit only comes with PocketSphinx, which has not been optimized yet for use with the toolkit.

Where can I type in my input for the character?

The AcquireSpeech client starts on the first tab in which you can modify some of the default settings. The Recorder and Player tabs allow users to type in their utterances. The Player tab also loads a script with some example questions which can be used by double clicking on it.

Launcher

How can I bring up console / debug information?

In the Advanced menu, click Show Console option (CTRL + T).

How can I disable the periodic component status pinging / vrAllCall messages?

In the Advanced menu, disable Enable vrAllCall (CTRL + SHIFT + T).

How do I save changes to a certain profile?

Unsaved changes are automatically saved to a Last Known Config profile, which will be loaded by default on start up. If you want to save changes under a specific name, for instance to save several often used configurations, you can do that in the Profiles menu, Save As New Profile (CTRL + SHIFT + S). This will bring up a new window in which you can enter your profile name.

How do I switch between Build and SVN mode for a certain component?

If supported, right click on any component name and select the appropriate Switch To option. Note that this feature assumes a symmetrical installation of both a build version and a SVN (version control) version of your project. This is currently not supported by the toolkit.

Logger

Where did the Logger go?

The Logger window is minimized on start-up. On the task bar, you can usually find it next to the Launcher.

Why do new messages periodically appear, even when I'm not interacting with a character?

By default, the Launcher sends out status request messages to all modules and tools (vrAllCall). These then respond with a confirmation message, indicating they are online. This enables the Launcher to show which components are still online. You can turn this behavior off in the Advanced menu of the Launcher (Enable vrAllCall option). Note that the Launcher will then not correctly indicate component status anymore. For more information on this particular message protocol, see Messages.

VHMSG

Troubleshooting

The Launcher does not start

It might be that the executable used to start the Launcher can not be found. Please make sure that javaw.exe is part of your path. The location is usually "C:\Program Files\Java\jre1.6.0_10\bin" or "C:\Program Files (x86)\Java\jre1.6.0_10\bin". The path is defined in your Environment Variables; right-click on My Computer, go to the Advanced tab and click Environment Variables.

Known Problems

  • Although Windows 7 is supported, due to non-backwards compatibility of the Microsoft SAPI interface in Windows 7, lip synching is likely off.

Website / Wiki

How to Use Media Wiki

Which terminology to use

The toolkit should be referred to as either 'Virtual Human Toolkit' or 'toolkit'. Please do not use 'VH toolkit', 'Virtual Humans Toolkit' (notice the plural humans), 'vhtoolkit' or 'VHT'. If you want to use some abbreviation, just use 'toolkit'. 'vhtoolkit' is exclusively used as an abbreviation in code, URL's, email addresses, etc., not in documentation. For more terms see the Glossary below.

Glossary

ActiveMQ

General purpose messaging system, used by the toolkit components.

AcquireSpeech

[AcquireSpeech, one of the toolkit modules. Can also be used for typing in questions or for selecting utterances from a predefined script.

Author

Creates characters.

Component

A component is either a module, tool, or library that's part of the toolkit. It is used to refer to a general toolkit element, rather than more specific elements.

Developers

Modifies the toolkit technology in some way, or uses the technology in other systems.

ICT

See Institute for Creative Technologies.

Institute for Creative Technologies

Plain awesomeness.

Library

A software component of the toolkit. Support modules and/or tools and is usually hidden from users and authors.

Module

A runtime toolkit component that is part of the toolkit architecture. Modules are essential for running a virtual character, as opposed to tools or libraries which just play a support role.

NVBG

See Nonverbal Behavior Generator.

Nonverbal Behavior Generator

Generates nonverbal behavior, based on textual input. One of the toolkit modules. Often referred to with acronym NVBG. Developed at the Institute for Creative Technologies.

NPCEditor

Statistical text classifier that takes text input and selects text output. One of the toolkit modules. Serves as a character's brain and source for natural language input and output. Developed at the Institute for Creative Technologies.

Plist

Data file used by the NPCEditor. Although a general XML format, in the toolkit the plist is often used when referring to a character's knowledge.

SBM

Stands for SmartBody Module. See SmartBody.

SmartBody

Procedural animation system. One of the toolkit modules. Sometimes referred to as SBM. Developed at the Institute for Creative Technologies.

Text-To-Speech

Module that generates audio based on textual input. Provides a character's voice.

Tool

A toolkit component that allows users, authors and developers supportive functions.

Toolkit

See FAQ#Virtual Human Toolkit.

TTS

See FAQ#Text-To-Speech.

USC

See University of Southern California.

University of Southern California

University with which the Institute for Creative Technologies is affiliated.

User

Runs the toolkit.

VH

Usually an acronym used to refer to the ICT Virtual Humans group.

VHT

See FAQ#Virtual Human Toolkit.

Vhtoolkit

See FAQ#Virtual Human Toolkit.

Virtual Human Toolkit

Collection of modules, tools and libraries that allow users to create their own virtual humans. This software is being developed at the University of Southern California Institute for Creative Technologies and is freely available for the research community. Sometimes referred to as VHT in text, always at 'vhtoolkit' in code, urls, etc.

Watson

Real-time visual feedback recognition library for interactive interfaces that can recognize head gaze, head gestures, eye gaze and eye gestures using the images of a monocular or stereo camera. One of the toolkit modules.

  • No labels