Years ago I’d written my Audio Files addin to grab audio files from websites as I did not like the TTS voices of the time. If you’d like feel free to take that as a basis for an addin of your own.
1 Like
disclaimer: I’m the developer of HyperTTS.
On the topic of TTS voices, when Azure voices came out with their Neural voices (around 2019), that was the turning point for most people where they finally found computer-generated voices high quality. There are still some edge cases in certain languages that cannot always be fixed (save for instructing the TTS engine how to pronounce). I think notably Japanese is problematic with TTS ( Optimizing Japanese text-to-speech with Amazon Polly | Artificial Intelligence ). Some people still have trouble accepting TTS voices, but I would say they are a minority.
Most recently the new style of AI generated voices such as OpenAI and ElevenLabs have become indistinguishable from human speech, for English, with varying results for non-English languages (they tend to have an american accent in other languages).
With HyperTTS, I am keen to cover as many services as possible and I’m working on adding self-hosted TTS services such as Kokoro, but keep in mind a large portion of HyperTTS users are novices so the experience has to cater to them. Perhaps I can also think of implementing a “power users” services which would have everything under the sun, but more difficult to configure (self hosted GPU-powered TTS engines come to mind).
By the way, did you know it’s trivial to implement a service in HyperTTS ? You pretty much just drop a python file in the right place, and HyperTTS picks it up automatically. The python code needs to match the signature of other services, but an LLM can generate this for you trivially. So you know a tiny bit of python, and you know how to use an AI-codegen LLM, you can already implement support for any TTS service inside HyperTTS and have an easy UI to generate the audio inside Anki.
2 Likes