Artificial Intelegent
- Thread starter Hit 94
- Start date
I think you can install another voice into the system and then select it in RadioBOSS.
From the video, it looks like the AI is using Azure. Based on my understanding, it needs to be consumed through an API and does not support Windows TTS.
Text to Speech – Realistic AI Voice Generator | Microsoft Azure
Build apps and services that use AI voice generators to speak naturally using synthesized speech. Engage customers with text readers and text to speech conversions.
azure.microsoft.com
Yes, could be - I'll see if this can be added into RadioBOSS as well.
This seems to be pretty easy to add to RadioBOSS. I did a very quick look at it, but fail to see where the AI is in here. Looks like search and replace - what RadioBOSS has currently with the variables. We just need to add more of them + condition blocks. And use better voices. I scheduled it for the future updates. Thank you for the idea.
Thank you
Fragments with different voices is also a great idea.
AI could be implemented in the future, read news via RSS and with chatGPT "Write it for radio" and filter sad news from the news for example. I didn't see that in that software but it's an interesting idea that occurs to me.
There's a danger it will read something odd on air
But technically, yes, that's definitely possible. One thing is that any GPT solutions usually come at a price, so this will require some additional payment I guess.Exactly, its usefulness depends on how low that percentage of errors is
Yes, even Microsoft Azure voices have a subscription based on frequency of use. But any subscription will be insignificant compared to hiring a host for a nighttime program (for example), saving the big differences with AI
Viafiber.se
Member
Kund of off topic but ...
How about using https://cross-plus-a.com/bconsole.htm Bakabolka, dd the variables, they could be used to send to the external program
How about using https://cross-plus-a.com/bconsole.htm Bakabolka, dd the variables, they could be used to send to the external program
I found the documentation and it is interesting. AI is used to automatically rewrite the scripts and make them more dynamic, plus there is a replacement of conflicting words:
AI Voice Links:Myriad has a brand new "AI Voice Links" feature that can used to automatically generate voice links to keep your listeners updated even when a station is unattended. These start from simple station name and time checks, but can also be expanded to include Weather and simple song information. They can even use Songfacts.com information if available. You can also use Open AI (aka Chat GPT) to rewrite either manually created or auto generated scripts to give those links a more natural feel. This new feature is ideal for live shows to give a second "voice" to an show, for example reading out the weather rather than it always being the presenter, but of course can also be used in fully automated stations to broadcast more dynamic content.
Another example could be if you have a Title or Artist that the AI Voice generation service does not handle correctly, you can correct that in the Replacement list using the Click here to add option.
In this example we have added the artist 'N-SYNC' but replaced it with a phonetically spelled version so that the AI Voice generation service can pronounce it correctly.
AI Voice Links:Myriad has a brand new "AI Voice Links" feature that can used to automatically generate voice links to keep your listeners updated even when a station is unattended. These start from simple station name and time checks, but can also be expanded to include Weather and simple song information. They can even use Songfacts.com information if available. You can also use Open AI (aka Chat GPT) to rewrite either manually created or auto generated scripts to give those links a more natural feel. This new feature is ideal for live shows to give a second "voice" to an show, for example reading out the weather rather than it always being the presenter, but of course can also be used in fully automated stations to broadcast more dynamic content.
- You can use the built in Windows voices for free, or you can use voices from ElevenLabs, Microsoft Azure, or Google - you'll just need to signup for a subscription direct with them.
- You can use Open AI (aka ChatGPT) or Azure OpenAI (recommended) if you want to use GPT script rewriting.
- There are several new Commands that allow you to pre-create Voice links in advance, but can also be used to only re-create a Voice Link if the script has changed - ideal for content like weather reports where the weather provider might only update the script every 4 hours or so, but you want to make sure you always have the latest version every hour.
- There are also several new Help articles that explain how to use the new AI features:
- Setting Up AI Voice Links In Myriad Playout
- AI Text & Speech Synthesis Services Supported By AI Voice Links In Myriad 6 Playout
- Adding AI Voice Links To Clocks & The Log
- Myriad AI Voice Links - using placeholder tags
- Futuri AudioAI support: We have worked hard with Futuri and are pleased to say that Myriad can also now use Futuri AudioAI AI voice links, as well as our own native built-in AI voice link system. You can even combine both services on the same Station.
- Replacement Word List - The Replacement Word List allows you to 'tune' AI Voice Links to pronounce words correctly or remove strings of characters entirely.
- To view and edit the Replacement Word List, click on Edit List.
Another example could be if you have a Title or Artist that the AI Voice generation service does not handle correctly, you can correct that in the Replacement list using the Click here to add option.
In this example we have added the artist 'N-SYNC' but replaced it with a phonetically spelled version so that the AI Voice generation service can pronounce it correctly.
Attachments
Last edited by a moderator:
This seems to come down to:
If this feature is included in RadioBOSS it will do the GPT communication part, that will simplify the process a lot. I'll add this to the to-do list for the future versions.
- Make GPT generate some text
- Do string search-and-replace in that text to change/remove some words
- Read text using TTS
If this feature is included in RadioBOSS it will do the GPT communication part, that will simplify the process a lot. I'll add this to the to-do list for the future versions.
Yes, it opened a step prior to this sequence of steps, which is to obtain global words to generate a script.
Name of the station, year of the next song, artist of the next song, weather data, etc. (And being able to configure the frequency of each information). This is like the basis for GTP to change the text. (I think the idea of AI here is so that the texts do not sound repetitive)
Finally, convert those texts into something better than the Windows TTS, we already discussed that as well.
The problem is that TTS does not sound natural, at least in Spanish. A little over a year ago I made a code in python that generated voicetrack audios with Microsoft Azure, something very similar to this. At that time ChatGTP had not been launched and due to time constraints my project was paused. Although I can return to it, the solution of integrating it into RadioBOSS is much more appropriate since I can easily integrate song names, weather data and everything in the same software. And I also think it increases the characteristics for all users.
If you have such a solution, adding it to RadioBOSS is easy, e.g. you can have an event that runs the generated .mp3 file periodically, or insert it via the API to the playlist.Although I can return to it, the solution of integrating it into RadioBOSS is much more appropriate since I can easily integrate song names, weather data and everything in the same software. And I also think it increases the characteristics for all users.
Yes, generating the prompt is one step, before GPT API is called, of course.
Calling the TTS engine itself is the easy part, we only need to ask user for API keys and that's about it. This is all on the to-do list and will be added fairly soon.
Error code 2 means file not found, looks like TTS engine did not create anything so there's nothing to play. Were you able to select those voices in the TTS box in RadioBOSS? Does it work properly with the default voices installed in Windows?
lughsan
Active member
No, the above error message "Code 2" appears when trying to play one of the voices in RB.
Yes it works well, but I would like to have more variety of voices.
RadioBOSS first generates voice into a file, then plays it. Looks like creating this temporary voice file failed.
Maybe the problem is with the voice, for some reason it can't output to a file.
lughsan
Active member
about the legality of using tts voices.
I understand that to use neural voices you must pay for their use.
Can TTS voices be used on a radio or do I have to pay for their use?
specifically the Microsoft voices, which are the ones that have worked for me with RB.
I haven't been able to find an answer.
Have you thought about the possibility of creating an RB-compatible neural voice?
I think if that were achieved, it would be a great step for RB.
adding more value to the product.
I understand that to use neural voices you must pay for their use.
Can TTS voices be used on a radio or do I have to pay for their use?
specifically the Microsoft voices, which are the ones that have worked for me with RB.
I haven't been able to find an answer.
Have you thought about the possibility of creating an RB-compatible neural voice?
I think if that were achieved, it would be a great step for RB.
adding more value to the product.
