Mochis built in TTS generation fails quite a lot. On about a couple hundred cards it was a 40% success rate. Instead of just regenerating it at a later point Mochi places a empty attachment with 0kb which has to be deleted manually to then trigger the regeneration progress (on replaying the audio). This has to be redone manually for each failed card and sometimes it takes up to 5 attempts until it successfully generates. Only tested in Chinese.
Also some times the TTS Quality is really bad but after regenerating it suddenly gives a different/better one.
Context: I added a lot of cards at once via the API. TTS generation only started when opening the deck in Mochi, but per deck it generated up to 50 Cards simultaneously. But if I recall correctly it even failed for decks with about 10 cards.