Important Information: This article refers to features that are still in development and are not yet available. These features may be withdrawn or altered at any time, therefore this information is subject to change at any time and without notice.
Overall feature description
Myriad Playout v6.3 introduced AI Voice Links to allow you to add fully automated, synthesized voice links to your live and automated playout.
The aim of AI Voice Links is to provide a voiced element to shows without the need for a presenter to manually plan and record the content.
Possible applications include:
- Time checks
- News & weather
- Automated DJ links
- What's on guides
Important additional information: There are several other help articles that relate to AI Voice Links and you should review them all before using this feature. You can see a list of these at the bottom of this article.
There are two (and a half) steps to generating AI Voice Links:
-
Create a script - By using some basic criteria about what content could or should be included in the script, Myriad Playout will generate a script using a combination of pre-built phrases and live content from the Log, Smart Info and Script files.
1.5. Use AI service to re-write script - As an option, you can also use a commercial AI service (such as OpenAI (Chat GTP) or Azure OpenAI) to rewrite scripts for infinite variation (subscription required).
- Generate voice audio based on script - Once the script is generated, Myriad will use your preferred 'text to speech' (TTS) service to render the Voice Link and place it in the Log ready for playout. You can use the built in TTS service in windows or use a 3rd party AI Voice service (such as ElevenLabs, Google Voice or Azure AI Voice services). 3rd party services will require a subscription.
AI Script Generation
Myriad currently has support for using either OpenAI or Microsoft Azure OpenAI to automatically generate or refine scripts.
Broadcast Radio currently recommend using Azure OpenAI instead of OpenAI (aka ChatGPT) as Azure OpenAI allows direct use of the GPT4 language model and is also considerably faster at script generation.
Advice on getting the best from Script Generation:
Time announcements
The time that is spoken is based on the Estimated Start Time, and because AI Voice links are usually created several minutes ahead of time, you should avoid using time announcements in places where the expected log time might change between creation and being used On-Air - for example, at the start of the hour, if you are using Soft Times to allow previous hours to overrun.
News & Weather Script Files
AI Voice Link scripts can include the content from text files in the 'scripts' folder. This allows you to update new and weather for inclusion in AI Voice Links.
Information Tags
In addition to script files, your AI Voice Links can also include special that allow you to include key information including:
- Time
- Previous & Next Title, Artists, Year
- Station Name & Tagline
- Presenter Name etc
Voice Generation
CAUTION: If you are using a 3rd party voice generation services, it is your responsibility ensure you fully understand their pricing structure to avoid unexpected bills!
Myriad currently supports several different TTS Voice generators (aka "Text To Speech"):
- Windows built in Text To Speech (Sapi v4/ Sapi v5) (see Home | CereProc Text-to-Speech to purchase more local voices)
- Azure AI voices
- Google Text to Speech
- ElevenLabs
Optimising performance:
Because the Text To Speech process can take a few moments, Myriad will try and pre-create the voice audio a few minutes before it's needed. If your scheduled log contains a sequence of several very short items in a row (for example, a short jingle, an Absolute Time, then an Hour start, then a short AutoHook item) then you may find that the Voice Link is not generated in time and Myriad pauses briefly while it's being created. To help with this, go into Station Settings, and on the Scheduled Log tab increase the "Additional number of log items ahead to cache" by a small amount. This will then tell Myriad to look slightly further ahead when considering when an AI Voice Link needs to be created.
Azure AI Voices
Azure AI Voices provides a wide range of quality AI driven, 'natural' voices with options for male and female voices from around the world. Myriad Playout is able to use individual voices for different sections of a single AI Voice Link so it is possible to have a link that starts with a presenter 'back announcing' the previous song, then pass to a different voice to read the weather. You could even go back to the first voice to complete the link. Please note that multiple voices in a single AI Voice Link is not possible when combined with AI Script rewriting by Azure OpenAI or Chat GTP.
To signup for an Azure AI Voice account, click here - Text to Speech – Realistic AI Voice Generator | Microsoft Azure
The signup process is a little convoluted and is regularly updated so unfortunately we are not able to assist with creation and setup of your Azure AI Voice service account.
Once you have completed the process, you will receive an API key which can be added to Myriad Playout to allow Myriad to use your account to generate audio.
If you receive an error while the audio for the voice links is being created then you may need to install the Visual C++ 2022 Runtimes from Microsoft - https://aka.ms/vs/17/release/vc_redist.x86.exe
Useful links:
- Microsoft Speech Portal including demonstration voices
- Microsoft Visual C++ 2022 runtimes
Google Text To Speech
Google TTS service is very similar to Azure AI Voice services in that Google TTS service provide high quality, AI derived male and female voices from around the world. Google TTS can also handle multiple voices in a single AI Voice Link making links more engaging and dynamic. Please note that multiple voices in a single AI Voice Link is not possible when combined with AI Script rewriting by Azure OpenAI or Chat GTP.
Also like Azure AI Voices, Google TTS service is also a convoluted process to setup a subscription, so unfortunately we are not able to assist with creation and setup of your Azure AI Voice service account.
To get started click here - Text-to-Speech AI: Lifelike Speech Synthesis | Google Cloud
Once your Google TTS service is setup, you can download a 'Json configuration file' which can be added to Myriad Playout to allow access your Google TTS account.
Important: If you have setup Google TTS correctly but receive a message that your Google voice link could not be created when using multi-part voice links, you may need to change the voices you have selected - for some reason, Google does not allow mixing of "WaveNet" voices and "Neural" voices in the same speech. You may also find that Google does not allow mixing and matching of multiple different types of Voice Links during a single link - e.g. mixing WaveNet and Neural voices in the same link
Useful links:
- Overview of Google Text To Speech service including voice examples:
- Getting a a Google "Credentials file":
ElevenLabs
ElevenLabs is a web service dedicated to AI voice creation services. Compared to Azure AI Voices and Google TTS services, ElevenLabs is much easier to subscribe to as you only need to use your email address to setup an account. The voices provided by ElevenLabs are also very good quality but they do not offer the facility to have different voices within the same AI Voice Link.
To get started with ElevenLabs, click here - Text to Speech & AI Voice Generator - ElevenLabs
Once you have setup your account and logged in, click on the Profile icon (top right) and select Profile. You can then view your API key by clicking on the 'eye' icon.
This API key can be added to Myriad Playout to allow it to access your ElevenLabs account to generate voice link audio.
Useful links:
- ElevenLabs website including voice testing demonstrator
Related Articles
- Setting Up AI Voice Links In Myriad Playout (this feature is not yet available)
- Adding AI Voice Links To Clocks & The Log
- Myriad AI VoiceLinks - Using placeholder 'Tags'
Attachments (1)
Was this article helpful?
That’s Great!
Thank you for your feedback
Sorry! We couldn't be helpful
Thank you for your feedback
Feedback sent
We appreciate your effort and will try to fix the article