What is ModelTalker?
The ModelTalker System is a revolutionary speech synthesis software package developed by the Nemours Speech Research Laboratory and designed to benefit people who are losing or who have already lost their ability to speak. It allows people who use a Speech Generating Device (SGD) to communicate with a unique personal synthetic voice that is representative of their own voice.
To learn more, explore the information in our new Help Center under the Help menu, listen to our Demo voices under the Voices menu, or Register to get started right away!
News
9/12/24 — Anticipated System Outage
Wednesday, 9/18/24 from 5:00 – 7:00pm US Eastern Time
We have been notified that our internet service will be down for about 30 minutes on September 18th, between 17:00 and 19:00 (US EDT) to perform necessary updates to our network. We do not know exactly when the outage will occur so please plan to avoid recording or attempting to audition voices within that time frame. Apologies in advance for any inconvenience.
5/19/24 — iOS (iPhone and iPad) System Voices Ready for Testing!
We are incredibly pleased to announce the first beta release of our Gen3 voices that are installable as iOS system voices. This will allow you to install your voice directly onto an iPhone or iPad and use it just like any of the voices that Apple provides. If you would like to try using your Gen3 voice on an iPhone or iPad, please send an email to beta@modeltalker.org and provide the email address you use with Apple so we can correctly identify you. You will receive an email with the code that you can redeem in TestFlight (a free app that you must also install to access beta software). You can read more about how to use the iPhone/iPad app on here.
5/18/24 — Updated ModelTalker Gen3 voices for Windows Apps/Devices
We are excited to announce an update to our Gen3 voices for Windows PCs and Windows-based devices running Windows 10 or newer. With this update Gen3 voices are now compatible with most current AAC software for Windows, including recent versions of Grid 3, Communicator 5, MindExpress, and other popular applications. Gen3 voices are still computationally demanding and may work less well on older or slower hardware. If you would like an updated version of your Gen3 voice for Windows, please let us know.
1/22/24 — New shorter Gen3 inventory now available!
We are pleased to announce a new voice-banking inventory especially suited for use with our new Gen3 voice technology. This new inventory consists of 298 sentences that are designed to elicit examples of expressive speech while providing the needed phonetic content to train life-like personal synthetic voices. Voice bankers who are currently using other inventories are welcome to continue using those inventories, but encouraged to switch to or add the Gen3 inventory; anything previously recorded for our Standard or Custom inventories will also be used in building your personal voice. You are, of course, still welcome to add your own custom words or sentences as a Custom inventory that will be merged with your Gen3 inventory when your voice is built.
12/4/23 — Gen3 voices for Android-based SGDs, tablets, and phones now available!
We are excited to announce that our Gen3 voices can now be installed as system voices on Android devices including phones, tablets and Android-based SGDs. Our voices should be compatible with Android 6 and above versions.
11/11/23 — Gen3 Installers for PCs and Windows-based SGDs now available!
We are now able to provide downloadable SAPI 5 ModelTalker Gen3 voices for Windows PCs and Windows-based Speech Generating Devices. For iOS (Predictable and TD Talk/TD Pilot) users, our Gen3 software updates have been provided to the developers at Therapy Box and Tobii-Dynavox. We anticipate that both apps will be compatible with our Gen3 voices via forthcoming updates.
10/19/23 — Auditions now including Gen3 voices!
We have modified our voice construction and audition processes to include the creation of our new AI-based Gen3 voices for inclusion in auditions. Auditions will now typically include examples of all three generations of our voice technology for clients to compare. As always, first and second generation voices are available for installation on most popular SGDs. We should soon be able to provide Gen3 installers for some SGD platforms, with more coming soon thereafter.
10/1/23 — New 3rd Generation ModelTalker voices!
We are thrilled to announce the initial release of our new third-generation (“Gen3”) ModelTalker TTS voices. Live demos of our new voices are now available on a special demonstration page. Coming soon, we will be generating these voices for all voice bankers to audition and installers for personal voices will be downloadable within a few weeks. We have an exciting development roadmap planned for our Gen3 technology. Check back frequently for news about new features such has shorter and more dynamic inventories, enhanced voice expressiveness, voice repair for dysarthric speakers, and more!
7/21/23 — New Research Study
We are conducting an IRB-approved study to help us understand the nature of the dysarthria associated with ALS/MND using recordings from the many voice bankers who have allowed us to use some of their recordings for improving personalized text-to-speech synthesis. The study may help us understand ways to compensate for dysarthria in our synthetic voices. Any native English-speaking adult can participate, but we are particularly interested in those who are familiar with dysarthric speech. For more information, please visit the study page or contact us at SpeechRating@ModelTalker.org with questions.
12/7/22 — New iOS option for ModelTalker Voices
There are now two front-line AAC apps for iOS that support ModelTalker voices! We are delighted to announce that you can now use your ModelTalker voice on the free Tobii-Dynavox TD Talk app for iPads. TD Talk is also included with TD Pilot the full-fledged AAC app from Tobii-Dynavox, which supports eye tracking technology.For instructions, see the “Use Voice Banking” video on the TD Talk support page.
8/18/22 — New Mac Installers for ModelTalker Voices
To comply with the latest security requirements for Apple software and hardware, our new voices for Mac laptops and desktops are now signed and notarized for installation on both Intel and Apple silicon based systems. If your voice was created before August 18, 2022 and you would like a version that works on the latest Apple laptops and desktops, please let us know and we can rebuild your voice for this new hardware and software.
7/22/22 — Staff changes at ModelTalker
After two years of wonderful assistance with our voice banking service, Matthew Buzzell will be moving on to begin his graduate studies. We are very grateful for all he has done to assist our voice banking clients and wish him every success with his studies! Our colleague Maxwell Hope will be joining Jason and Tim to carry on the voice banking service while we look to fill Matthew’s position in the lab. If you or someone you know might be interested in a position with our group (part of the Speech Research Laboratory at Nemours Children’s Hospital, Delaware), have a look at our job posting.
6/25/20 — Upcoming Service Interruptions
We have been notified by the University of Delaware, which provides internet access to our ModelTalker.org website that we will experience service interruptions on Monday June 29th from 10:00 PM to midnight and Wednesday, July 8th 10:00 PM – Thursday, July 9th 1:00 AM. All times are US Eastern Time (GMT – 5 or UTC – 4). We apologize in advance for any inconvenience.
3/17/20 — COVID-19
We sincerely hope that everyone is taking the best care possible to minimize their risk of contracting the novel Coronavirus now called COVID-19. If there is any glimmer of positive news related to COVID-19 it is that it is not a computer virus! Our servers remain up and our team is mostly able to continue work as usual although we are switching to telecommuting to minimize social contact as much as possible. You may find our response times a bit slower than usual as we work out all the wrinkles associated with remote connections to servers, but we are doing all that we can to carry on as usual despite the pandemic that is sweeping the globe.
3/11/20 — New Help Center
We are combining our FAQ and various other informational pages under a new searchable Help Center. We will be continuing to add and refine the Help Center over time and we hope it will make it much easier for users to locate information that will be of assistance in recording or to learn about voice banking. Please try it out and let us know if you have any problems or if you can think of additional information that should be added.
8/6/19 — Website update
Our site is now fully updated with the changes noted in the News item below from 7/27/19. See the simplified workflow description for details. Please email us with questions/problems.
7/27/19 — Online Recorder Changes
We are preparing to make some noticeable changes that we hope will simplify the workflow for our voice banking service and the online recorder interface. At the same time, we will also be opening the interface that will allow clients to audition and download donated voices. The changes we will be making are described in this page. The target date for us to go live with these changes is August 3, 2019.
7/5/19 — New Demo Voices
Are you curious what your voice might sound like if you only recorded 400 sentences versus 1600 or more sentences? What are the differences between our normal unit selection voices and the new hybrid DNN voices? This is now illustrated on our Demo page using different versions of our Tom voice. (Note: ‘Tom’ is a professional voice artist who recorded our standard sentences in our lab — These are ‘best case’ examples).
7/1/19 — Donated Voices (Coming soon)
We will soon be adding a new feature to allow registered clients to review and select from our growing list of donated voices. While this feature is intended primarily for AAC users who are unable to record their own personal voice, even clients who have made a personal voice will be able to add selected donor voices to their “My Voices” list of voices. Any voice in a client’s My Voices list can be downloaded directly from our site to a Windows or Mac computer or set as the default voice for installation on iOS (Predictable) or Android. We will eventually be adding a small charge for most donated voices, however the charge will be substantially less than for personal voices since the same donated voice may be used by multiple clients.
4/5/19 — New Voice Options
Since last summer, we have been testing–and occasionally providing voice bankers with–hybrid Deep Neural Network (hDNN) voices that are based on new machine learning algorithms and parametric speech synthesis, but are compatible with all existing versions of our TTS system. We have now begun routinely providing these hDNN voices as an option for all voice bankers when auditioning their voices. While hDNN voices can sound more machine-like in voice quality, they nonetheless tend to flow more naturally, and are often more intelligible. The hDNN voices are especially promising for voice bankers who struggle with recording or who are unable to complete a full 1600- or 3200-sentence inventory. Don’t be alarmed if the first voice you hear in auditioning is now an hDNN voice; the standard voices will be there too and available for you to choose as your preferred voice!
11/21/18 — Site maintenance downtime
We must do some maintenance to the ModelTalker website that will involve a more extended amount of downtime than usual. We anticipate the site being unavailable for recording or voice downloads from the evening of 11/29/18 until late on 11/30/18 or early on 12/1/18. We apologize in advance for the inconvenience!
8/25/18 — New Reusable Custom Form!
We have updated the Custom Inventory form so it is possible to use and reuse the form as often as needed. It is no longer necessary to add everything at once! Now you can add new words and phrases/messages as you think of them and they will be appended to the end of your current inventory.
6/1/18 — Problems with Chrome…UPDATE
For users who are seeing an error that says:
Error capturing audio: message=null name=OverconstrainedError constraintName=undefined.
We have identified the cause of the problem–a change that Google introduced in a recent version of Chrome–and we are working on a fix. Until we have completed testing the fix, there is a work-around you can use:
- Login to https://modeltalker.org/vrec, but do not fill in the Settings dialog
- Click the movie camera icon in the right side of the chrome address bar. That will bring up a popup box to allow Microphone controls.
- Click the Manage button in the bottom left of the popup
- At the top of the resulting Chrome Settings page click the microphone selector, and choose the entry that says: “Default <your_microphone_name>” Where <your_microphone_name> is whatever the name of your microphone is such as “Sennheiser PC 36 USB Headset” possibly followed by some numbers in parentheses.
- Close the Chrome Settings page, log out of the online recorded, and log back in again
Thereafter, it should be possible for you to measure the silence and proceed with recording.
Very sorry for the inconvenience!!
3/13/18 — Inactive Voice Inventories
Some voice bankers start recording their full inventory, but then seem to discontinue recording. Because recordings take up space on our server that other users may need, we will be instituting a policy of backing off inventories from our server that have been inactive for a long time. We will be introducing this process gradually over the next several weeks. DON’T WORRY!! We will not delete anyone’s recordings. Inactive inventories will just be moved to a safe location and can be restored at any time.
We understand that it is easy to become engaged with other more immediate needs and not have an opportunity to return to complete voice banking for an extended period of time. If you record even a little bit every few months, your inventory will be considered active and remain on the live site. If you happen to be away long enough (at least half a year) that your recordings seem to have disappeared, just let us know and we will be able to restore them quickly.
We also understand that some people find the task too demanding to complete all the sentences we recommend. If you are in this category, please do let us know. We are sometimes able to build usable voices from substantially fewer than 1600 sentences and would be happy to discuss this possibility with you.
1/31/18 — Team Gleason Supports ModelTalker Voice Banking!
Starting February 1, 2018, if you do not have other support, Team Gleason has generously offered to provide support for ModelTalker voice bankers who are US citizens with a confirmed diagnosis of ALS. Persons with ALS who complete the Team Gleason application on our website and are approved by Team Gleason will be able to download their completed ModelTalker voices for free. Visit the Team Gleason announcement for a custom link to register and request support from Team Gleason.
1/21/18 — Online Recorder Downtime Completed
Our service is completed and the website should now be fully functional once again. We have replaced the previous version of the online recorder with a new updated version that has many new features and a somewhat different look and feel. See our Online Recorder Overview for a summary of the changes we have introduced.
We tested the new recorder internally before releasing the changes, but as with any new complex software, there may be some problems we are not presently aware of. Please let us know if you experience any problems with the new interface.
12/21/17 — Happy Solstice! New features coming soon!
Whether your day is short or long, this is just a heads-up so you will not be surprised by some changes we will introduce in the near future. We have been working on some improvements to the online voice recorder to make it easier for you to review sentences that have already been recorded and keep track of which ones you might want to try to improve. Soon (possibly as early as next week) we will update the online recorder to show you a list of all the sentences you have recorded in a table below the recorder controls. You will be able to sort the list of sentences in a variety of ways and search for sentences containing specific content. There will be indicators for how optimal the recording is for amplitude, speaking rate, and pronunciation. And you will be able to listen to your recording, or select the sentence to rerecord.
12/8/17 — changes complete
We have completed making changes to the registration process. We have added a new registration category of “proxy” for people registering to provide a voice for a specific individual and simplified the registration process. Please let us know if you encounter any problems.
12/6/17 – Brief system downtimes
Our server team has warned us that there may be a period of 30 – 60 minutes tomorrow morning starting at around 8am EST (1pm GMT) where some services (e.g., downloads of Windows voice installers) are unavailable. If you are unable to download a windows voice, please try again later.
Additionally, we plan to have the ModelTalker.org website inactive for about two hours starting at 9pm US Eastern Standard time (2am, 12/9/17 GMT). The downtime will allow us to make a few changes to the website and registration process. Please avoid attempting to login during that time period.
Our apology for any inconvenience!
6/30/17 – Terms of Service Change Effective 07/01/17 12:01 AM EST
New registrations beginning Saturday 12:01 AM 07/01/17 US Eastern Daylight Time (5:01 am London; 14:01 pm Sydney; 16:01 pm Christchurch) for voice banking clients who want to create and download a personalized voice there will be a $100.00 USD fee to enable the download of completed voices. There will be no change for users who have registered prior to that time. At present, we are continuing the “no charge” policy for voice donors and for clinicians and those registering to learn about the technology so they can assist others. See our Fee Policy FAQ for more details.
6/1/2017 – Changes coming to our service!
For many years, we have provided ModelTalker voices to persons with communication needs for free as an adjunct to our research program. However, as demand for voice banking has grown over the last few years, it is no longer practical for us to meet the increased demand without offsetting staff costs. Consequently, for people registering and beginning the voice banking process after July 1, 2017, we will be charging $100 USD for users to enable download of completed voices.
If you are currently in the process of banking your voice, you will not be charged. Once we begin charging, the registration, recording, and auditioning process will continue exactly as they are now. However, after auditioning your voice, if you are pleased with the result, you will be charged to unlock the download capability. Once downloads are unlocked, there will be no restrictions on the number of downloads for different platforms and you will be eligible for voice upgrades if technology improvements warrant and if we have copies of your original speech recordings.
We will be updating our Fee Policy FAQ and other pages with more information about this change in the coming days, but wanted to be sure the word gets out as quickly as possible. Crucially, we are in discussion with a number of charitable organizations to determine what assistance they may be able to provide to users for whom this charge will be a hardship and will post more information about those opportunities as it becomes available.
5/2/2017 – MTVR update
We have posted a minor version update to the Windows MTVR program to correct several issues:
- Changed a prompt that started with “Look like she…” to the more grammatical “Looks like she…”
- Fixed a bug that sometimes created files with names like *.mp3.wav, which could then interfere with prompt playback.
- Replaced some old references to “modeltalker.com” with references to “modeltalker.org”
- Added Microsoft code signing to MTVR and its components so users will no longer receive a warning that the program is from an untrusted source. The developer will be listed as AgoraNet, Inc., which is the company that maintains our website and software.
Clinics using MTVR should update their software when convenient. Current personal or one-time users of MTVR do not need to update to this version unless they are experiencing problems that could be related to one of these issues.
4/6/2017 – Voice Auditioning
We are introducing a new feature that will allow users to have more say in the process of constructing their synthetic voice. In our lab, we create multiple versions of each user’s voice and then test them to see which one seems best. Our final choice becomes the voice we return to the user. We have now set up a process where users can audition the multiple versions of their voice and help us decide which version sounds best to them. See Voices > Auditioning for more information.
1/19/2017 – Automating Custom Inventories!
Users often ask if they can send us some names and/or phrases to record as part of the voice banking process. It is now possible to create your own custom inventory to record along with our standard inventories for voice banking. At any time while recording your full voice banking inventory, you can select Session > Custom Inventory. This will allow you to enter some names and phrases that you especially want your voice to be able to pronounce correctly. Custom sentences generated from the words and phrases you enter will automatically be added to the end of your active inventory, ready for you to record before we build your voice.