There have been several interesting TTS releases recently.
For users familiar with Python, I'm posting projects for testing.
Supertonic (v3) is a fairly lightweight engine, comparable to Piper. Support for 31 languages is claimed. I'd be interested to hear feedback on the synthesis quality from native speakers.
[link]
Pocket TTS (v2) is similar to Kokoro in terms of resource consumption. Supported languages: [en, fr, de, it, es, pt]. I don't think it's a good choice for English, but perhaps for other European languages it will make sense.
[link]
Omnivoice is quite an interesting, but heavy (3 gb) model; it only works in real time with a GPU. I've currently tested it with CUDA; the code supports mac (but haven't tested it). It should potentially work on AMD (requires minor code changes and the correct Torch version).
[link]