Installing more Desktop voices

Question

@Athena. How can I access more desktop voices? The only voices available to choose in Speech Settings  are Hazel, David and Zira. There are many other voices in Windows Settings but they do not appear in Arc.

Athena · Answer

To access more desktop voices in ARC for your EZ-B v4 robot, you'll need to ensure that the voices are installed and enabled on your Windows system. ARC utilizes the voices available through the Windows Text-to-Speech engine. If you're seeing a limited selection, it's likely due to the voices being restricted to specific applications or not being fully installed for system-wide use. Here's a step-by-step guide to ensure you have access to more voices in ARC: 1. Check Windows Text-to-Speech Settings: - Go to Windows Settings > Time & Language > Speech. - Under the "Manage voices" section, you can see the list of voices available and add more voices by clicking "Add voices". 2. Install Additional Voices: - In the "Add voices" menu, you can browse through the list of available languages and voices. Click on the ones you wish to add and then click the "Add" button. Windows will download and install these voices. 3. Enable Additional Voices for Third-Party Applications: Some voices might be restricted to use only with certain Microsoft applications. To make more voices available for third-party applications like ARC, you might need to edit the Windows Registry. This step involves modifying system settings, so proceed with caution and consider creating a system restore point before making changes. - Press Win + R, type regedit, and press Enter to open the Registry Editor. - Navigate to the following path: Computer\HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Speech_OneCore\Voices\Tokens - Here, you'll see folders for each voice installed on your system. Copy the path of the voice you want to enable for third-party applications. - Now, navigate to Computer\HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\SPEECH\Voices\Tokens - Right-click on the "Tokens" folder, select New > Key, and name it after the voice you're adding. Paste the information from the OneCore voice registry into this new key. - After adding the registry keys, restart your computer for the changes to take effect. 4. Check ARC for New Voices: - Open ARC and go to the Speech Settings to see if the newly added voices are available for selection. Important: Modifying the Windows Registry can have unintended consequences if not done correctly. If you're not comfortable making these changes, seek assistance from someone who is experienced with Windows Registry edits. By following these steps, you should be able to access more desktop voices in ARC for your EZ-B v4 robot. Remember, the availability of certain voices may vary based on your Windows version and the languages installed on your system.

Synthiam Support · Answer

You may wish to use the support section and look at robot skills in the audio category. Because the speech synthesis skill you’re limited to Microsoft voices for that system. I recommend you look at the azure speech synthesis skill which has hundreds of voices.  check for all audio skills here: https://synthiam.com/Support/Skills/Skills-Overview#Audio

afcorson · Answer

I have not looked at using Azure speech synthesis, as I thought it required my robot to be connected to the internet. I have a couple of robots that use ChatGPT and need to be on-line for that, but my other robots never use the internet. An ARC pro subscription is also locked in forever. Nonetheless I will consider Azure as an option. Thanks. Another thought worth mentioning, is that using ChatGPT with an on-line voice, is double handling the robot's dialogue. It takes a few seconds for ChatGPT to respond with text, which would need to be resent through Azure to get a voice response - that's another 2 or 3 seconds. By the time the robot responds, the visitor has walked away.

DJ Sures · Answer

I don't think there are any other options. You can look into purchasing 3rd party applications like Cepstral, I believe it was called. I can't remember, but there was some program people were purchasing to add custom voices to Windows.  The trouble you're facing is that all these companies who make technologies want cloud APIs. Unfortunately, I cannot do anything about that as they're separate companies. It's been a trend for companies to keep their technology in the cloud to prevent competitors from reverse engineering.  Even if you wanted to use one of the 3rd party cloud APIs, like Azure Text To Speech, without a Pro subscription, you'd need to pay them. Synthiam makes some of these apps more affordable because we purchase large quantitiesin bulk. You could buy it yourself, but it'll cost more than a monthly ARC subscription. I feel your pain, though. During the development of ARCx, we had a cross-platform strategy. And I have to say, there's nothing available for Linux or MacOS when it comes to speech synthesis or recognition. Not only are there no speech libraries, but there are a million other missing support for technologies. Most USB devices aren't supported or require a week of work for you to make the kernel compatible. As much as I like the idea of ARCx on low-powered devices, I find the limitations of Linux to be a HUGE PITA.  We still don't know how to make Speech to Text work on Linux for ARCx... we'll figure it out eventually, but right now, we're scratching our heads. So far, we can only see using cloud services for Linux speech-to-text.  Windows may have a reputation for being "bloated" and "resource-hungry," but I have realized it's because Windows actually does stuff.

afcorson · Answer

Thanks for your thoughts. I actually tried one of Cereproc's voices as it only cost $50, but when I tried using it in ARC, it took over 15 secs to respond. Needless to say I uninstalled it and will be getting a refund. Aside from a few Microsoft Desktop voices, the only ones that seem to work with ARC are the old Ivona voices. It is obviously a problem for even robots like Ameca. It took a while to respond to questions and it must have cost millions.

DJ Sures · Answer

Ameca has an excellent physical design, but the software is rudimentary. Your robot can perform better than her! The trouble with Ameca and the other band-wagon robots is that they are writing the software from scratch. So imagine writing ARC and maintaining it while maintaining a robot and fixing bugs, etc... We have a team to keep ARC going... can't imagine how difficult and costly it would be to have a physical robot AND be writing code every day from scratch. I can relate to when I ran EZ-Robot, which was challenging to pay for software development while making hardware. That's why I needed to split the company and focus on Software. Simple math: if you sell one robot for $100 and it costs $80 to the manufacturer, how long do you expect that $20 of profit to last? I wanted to continue creating new software features, but there's no way hardware sales would make it possible. That's why we see companies like iRobot struggling financially. Ameca should be using ARC, but the authors have a chip on their shoulders and want to own the full stack. That's why robot companies are bankrupt. That being said, I hope there's some development with speech libraries in the upcoming years - because that's what is missing with robotics and operating systems. Even games have a difficult time generating speech, so it's all pre-recorded. Microsoft was on the right track when they released the SAPI speech synthesis in Windows 7. I don't know why they didn't continue with it - because it's super awesome running local.

afcorson · Answer

Are these TTS boards any use in addressing the problem?   https://www.ebay.com.au/itm/155862169678

DJ Sures · Answer

That will add hardware to your robot but it'll work fine i imagine. You're stuck with the robotic voice it provides though.

Jeremie · Answer

@afcorson I have used Cereproc voices and I don't recall any slow downs. It should operate just as the Microsoft David voice.

Nomad 6R · Answer

hi all i used the cepstral voice of david but you can also ad aliasses in cepstral window . you can change the voice pitsch ,speed ,light or low . then you can ad this , in your ARC .

afcorson · Answer

I tried Cereproc Lawrence and Millie and my Win11 laptop. The response time was a joke. So there's a mystery there.

Nomad 6R · Answer

maybe deleting your browser hystory or updates on your laptop ?

robo rad · Answer

I just used the basic speech setting skill since we can use all those new effects like chorus,reverb and flanger effect. I was able to make the default David voice sound exactly like the Proteus computer/robot in the Demon Seed movie, very freaky coming out of my Roomba!

afcorson

Installing More Desktop Voices

Upgrade to ARC Pro