Not really an issue, more QoL

[T1nyRaccoon]

This isn't really an issue, but more a QoL thing.

The TTS options provided are quite difficult to scroll through, especially as you can only go forward not backwards.

Also, are there other TTS options? as in a different system? Because there is a weird amount of mid-word lag sometimes and they all sound pretty much identical. From what I've heard, Piper is a better option.

[FlooferLand]

You can go backwards by shift clicking, I explained it in the tooltip.

I've looked at a ton of different TTS engines, and so far eSpeak has been the only half-decent option. You could use eSpeak with mbrola to get nicer more natural sounding voices, but including mbrola and its voices in my mod is a very shaky legal ground.

Piper uses AI so I'd like to avoid it. Nothing wrong with its use of AI/machine learning specifically as I believe its using it quite ethically, but I think having a very realistic machine-generated human voice is somewhat uncanny and misleading. I think the robotiness to eSpeak makes it fit into Minecraft's aesthetic a lot more with it having a retro artstyle and all, and people won't think the very obvious robot that is speaking is how you sound like in real life.
If people want PiperTTS support I'll be glad to add it, feel free to open a discussion thread and I'll add it if it garners up enough upvotes, but I don't see it happening.
I do want to expose a way for people to add their own TTS engines at some point, like letting ttvoice run a user-made Python file every time you speak, but I'm not sure how transmitting that audio into Minecraft would work then as most Python TTS libraries don't have a way to get the audio data and they instead just play it through the speakers.

One engine I've thought about is festvox/festival as it sounds quite good, but Windows support is very much lacking.

I'm very much open to TTS engine recommendations, I just haven't found any singular one that works well.
Minecraft's own narrator uses Windows' TTS engine, while using Festival on Linux/Mac. That would be really annoying to work with on my end, not only because I would have to deal with platform-specific engines, but the voices and settings would all be completely different depending on what operating system you're on.

[T1nyRaccoon]

Thanks for the reply.

I think the shift click option may have been added a little later after I posted that. I did notice it after and it was very helpful.

I agree with the potential issues of realistic voices, but from my experience unless it’s chat-GPT level then it can be fairly easily told apart, at least from my experience with twitch TTS engines etc.

If it would be easy to implement festival, that does sound like a better option if you can add a seperate linux/ macOS download but If it is difficult I would say it isn’t worth it (or is the voice dependent on both ends running Linux?)

At the end of the day it is your mod and is already a great tool, I just find it is very hard to discern words with the current TTS system unfortunately

[FlooferLand]

That's fair. Another reason for my dislike towards AI is that AI models are huge, a few voices easily take up almost a gigabyte.

Festival as far as I can tell has poor support for Windows, and that's also the reason Minecraft's narrator is using Windows' TTS on Windows instead of Festival, as it already uses Festival for Linux/macOS.

I'll look into MaryTTS as it seems to be the only viable alternative, but it caused me a lot of problems when I had it in the mod previously as for whatever reason running my mod in debug mode causes MaryTTS to freak out and crash.

I plan on adding the ability to change voice parameters in the future, which includes the speaking rate, so it maybe might help if speech can be slowed down a bit.

[naosenai]

I do want to expose a way for people to add their own TTS engines at some point, like letting ttvoice run a user-made Python file every time you speak, but I'm not sure how transmitting that audio into Minecraft would work then as most Python TTS libraries don't have a way to get the audio data and they instead just play it through the speakers.

If you allow it to play audio from a file then I think that would be enough to make it a useful feature for implementing local TTS engines. I personally would be interested in making a script using VOICEVOX, which uses a localhost API. I have done the same for a private discord bot in the past, along with several similar engines, all of which outputted to temporary files on my PC before being uploaded.

I plan on adding the ability to change voice parameters in the future, which includes the speaking rate, so it maybe might help if speech can be slowed down a bit.

Assuming you do end up making something for custom TTS, it would be awesome if you are also able to add support for custom voice parameters too