Did you know HyperTTS is designed from the ground up to allow easily writing new TTS services ? All you need to do is drop a python file in the right place to introduce a new service.
I created a new repository here GitHub - Vocab-Apps/anki-hyper-tts-extensions: Community contributions for HyperTTS, extra services and documentation · GitHub which contains examples and instructions on how to create new services, and has AI coding agent instructions designed to help easily creating services.
This repository will also contain community-contributed services , documentation and code. It could be any of the following. If you submit a merge request, I will approve it very quickly with no restrictions (unless there’s something malicious like malware).
- custom TTS services, whether local, API based, cloud based.
- instructions on how to self host TTS services, could be anything from how to install espeak-ng to running large neural TTS models using GPU.
- instructions on how to use the free tiers for Google and Azure and others.
Why a different repository and not the main HyperTTS repository ? It has to do with my my support workload in helping users of HyperTTS with their issues. HyperTTS has a decently large user base and I try to respond to all support emails within 12 hours. This means I have to be very strategic in what I choose to support, so I would like to keep the core services very high quality. I won’t be able to support people’s self hosted TTS service setup and I think it’s important to state clearly what I commit to supporting and what will remain on an ad-hoc basis.
For the core HyperTTS code and services, I commit to being extremely attentive to any bugs, support issues. For this new “HyperTTS extensions” repository, it will be done on a best effort basis without guarantees.
FAQ
Does this mean contributions are not accepted on the main HyperTTS repository ?
No, contributions are still accepted, but they have to be of sufficient quality. The code contributed must make sense for the general HyperTTS audience, there should not be any bugs or strange behavior.
Can HyperTTS be used with any TTS service ?
Yes, in theory you can use any TTS service, but the difficulty in integrating can vary. The easiest thing to do is to use REST API based cloud services. For those, an AI coding agent can usually generate the code in one shot. For those that rely on local programs, it’s a bit more complicated particularly if python modules must be installed, but there are examples in the HyperTTS repository.
Can you add XYZ service ?
Usually I can and it’s not too difficult if it’s an online cloud API. However I need to be able to sign up for the service. If I can’t sign up (happened for some services in asia, or mainland china for example) , I can’t test it. Also if money is involved (such as when spinning up very large models from hugging face), I’ll be OK to pay for GPU hosting costs IF it’s likely to interest the wider community for the duration of testing, but if it’s a one-off very obscure language, then I may ask you to share the cost.
Is HyperTTS free ?
Yes HyperTTS is free and comes with free services: Free Services - HyperTTS - Vocab.Ai . The code is GPL licensed. What is NOT free are the premium services, those cost money, I can’t give access to those for free, otherwise I would go bankrupt very quickly. If you want to use premium services such as Google, Azure, ElevenLabs, you can either bring your own key, or you can choose the HyperTTS Pro services which is optional. If you want to just use free services it’s possible and nothing is stopping you.
Have some ideas of exotic TTS services you’d like to support ? Let me know !