Hi there,

It sounds like you’re experiencing some interference between the custom CereProc voice and ARC’s Speech Synthesis Settings. Although Windows Narrator plays the voice perfectly, sometimes third-party voices can behave differently when processed through ARC's built-in voice effects.

Here are a few suggestions to try and resolve or narrow down the issue:

1. Disable Custom Effects Temporarily:
   In the Speech Synthesis Settings robot skill, try resetting or disabling any custom effects (such as pitch adjustments, delay, flanger, or the speed effect). This will help determine if one of the effects is causing the noise.

   • Open the Speech Synthesis Settings configuration.
   • Set the Voice dropdown to your CereProc voice.
   • Disable or reset additional effects and leave the speech at its default settings.
   • Test using Audio.say() to verify if the noise disappears.

2. Check the Rate and Volume Settings:
   Sometimes, aggressive rate or volume settings can introduce artifacts in some voices. Experiment with slightly lower rate values and adjust the volume slider, then test the output.

3. Reset to Default and Reapply Changes Gradually:
   Consider resetting the Speech Synthesis Settings to the default values and then reapplying any desired customizations one at a time. This can help pinpoint if a particular setting is triggering the issue.

4. Audio Device and Drivers:
   Although Windows Narrator works fine, double-check that the audio output device selected in Windows or ARC is using the appropriate drivers. In some cases, mixing sample rates or bit depths between the voice’s output and ARC’s processing (or your EZ-B’s audio settings) might add noise.

5. Compatibility with Third-Party Voices:
   ARC uses the built-in Windows text-to-speech engine (SAPI), and while most third-party voices are compatible, there can be differences. If the problem persists after trying the above suggestions, you might consider testing with another third-party voice if possible or using one of the built-in voices to compare results.

6. Documentation and Further Tweaking:
   You can review the Speech Synthesis Settings documentation here for additional details:
   Speech Synthesis Settings Documentation
   This page might offer some insights regarding the effects and parameters being used.

Have you had a chance to test the voice with all effects disabled? Let me know if turning off the enhancements clears up the noise or if you need further assistance with specific settings. I'm here to help!

Hope this gets your CereProc voice sounding clear in ARC. Feel free to ask any more questions or share additional details if the issue continues.

Hi (@Athena), I haven’t applied any voice effects and all settings are set to default. So I’m not sure why there’s noise when it speaks. However, when I use the default voice, it speaks smoothly in ARC. Do you have any idea what might be causing this?

Hi there,

Since you mentioned that all the Speech Synthesis Settings are at their default and you haven’t applied any custom effects, it seems the issue is isolated to the third-party CereProc voice. Here are a few things to consider:

• Voice Compatibility:
  Even though the voice is SAPI-compliant and works well with Windows Narrator, sometimes third-party voices might have differences in how they process or output audio that can lead to artifacts or noise when used by ARC.

• Audio Processing and Sample Rates:
  ARC uses the built-in Windows TTS engine. Check your Windows sound settings and ensure that the sample rate and bit depth match what the CereProc voice is optimized for. Sometimes mismatches here (or even differences in how ARC processes the audio stream) can result in noise.

• Voice Installation and Drivers:
  Double-check that the voice was installed correctly and that all necessary drivers or additional files provided by CereProc are up to date. It might also help to reinstall the voice if you suspect any installation issues.

• ARC Version and Configuration:
  Ensuring that you’re using the latest version of ARC is important, as updates often include enhanced support for third-party voices. Although you mentioned that the default voice works smoothly, sometimes improvements or bug fixes in ARC could affect how different voices are managed.

• Testing with Minimal Setup:
  Try creating a simple test project with just the Speech Synthesis Settings and a basic Audio.say() command using the CereProc voice. This can help determine if the problem is project-specific or a broader compatibility issue.

Have you had a chance to verify any of these settings? Let me know if the issue persists or if you need further guidance on any of these steps. I'm here to help!

In your image, the speed is set to 80. That is what Athena mentioned. She mentioned effects,speed, etc. Use the default values, set everything to as it should it default.

Hi DJ,

Thank you for your reply!

I’ve also tried adjusting the speed settings, but I’m not sure if the issue is due to the third-party voices being incompatible.

Here’s a video link that I hope shows the issue more clearly: https://www.youtube.com/shorts/_sZLtlYsGEY

I tested Microsoft Zira at speeds 80 and 100 it worked smoothly. However, when I tested CereVoice Amy at the same speeds, there was noticeable noise during speech.

Thanks again!

@rickymahk2013

hi . your video is showing prive . i dont think anyone can see it .

Hi Normad, Sorry about the earlier issue! The video is now available. Thanks in advance!xD

That's a strange crackle. You may also consider sending that to CereProc for assistance with their product after completing a fresh test. However, I did notice that your video shows the speed set to 80, which Athena and DJ had both mentioned setting to 100 for testing. If you change the settings for overdrive volume, speed, filters, etc., then you're adding complexity to the debugging.

In addition, your screenshot shows several options that have been changed, including the speed and filter. I'm also unsure what other options could be checked in the EFFECTS tabs, since you have one checked on the first tab.

It would be an idea to start from a fresh project.

1. Start a new project in ARC

2. Add a SCRIPT robot skill and write a script to speak

Audio.say("This is a test, and I am talking right now, so you can hear me.");

3. Add the Speech Settings robot skill so you can select the voice

4. The only option you should change in the speech settings is the voice

5. Press the START on the script, and the voice will speak

Those instructions will provide you with a clean start with no options changed other than the voice drop-down menu. If you follow those instructions, do not change any other options. By changing other options, you have introduced complexity to the test, and the results are void.

Hi Support Team,

Hope your business is doing well!

I’ve tried the new project and created another sample video. It demonstrates two voicesCereProc and Microsoft Ziraand I tested them using Windows Narrator for text-to-speech. The voices sound better through Windows Narrator compared to how they sound in ARC.

You can check out the video here: https://youtu.be/ZLWMEYOmAtc

I’m wondering if this might be a CereProc quality issue, or if anyone has tried voice products from other companies that work more smoothly?

Thanks again!

I don’t know how much their product is used. Last few times anyone posted about them, it was negative experiences.

have you tried a different speech system, such as azure text to speech? Azure text to speech has hundreds of voices and works with your pro subscription.

Hi DJ,

Thank you for your response

That means I’ve wasted money on this product:(. I had already tried Azure Text to Speech, but I was concerned because it couldn’t slow down the speaking speed. That’s why I decided to try this product.

However, I’m wonderingif I use Azure Text to Speech and exceed the limitations, what will happen?

Also, I’m curious about ElevenLabs. They allow you to create your own voice and provide an API. I’m wondering if ARC will develop a plug-in for that, so we can use our own voice for the robot. That would be pretty fun:p

Maybe reach out to CereProc and ask them why the voice crackles. Tell them all that we're using is the built in System.Speech.Synthesis in .Net 4.8 and using the Speak() method. There's no fancy stuff being done - and they can try it themselves for testing using ARC Free. Or give them the link to this forum thread.

Thanks, DJ! Will do. Hope business goes great!

Looking into it a bit further... it seems as if the Windows Narrator uses a different engine than the System.Speech. I wonder if CereProc is aware that there may be a bug in their voice with System.Speech. Send them a message and inform them of this thread to get their input on the issue.

Thanks very much for the kind words as well!