Are there any costs - and if so, which – for releasing a conversational action I've built?
It is obvious that Firebase and/or Dialogflow cost some money, but what if you actually don’t use any of these services? I only use Google Assistant to capture voice input, then I call a webhook (I host the server) and reply with a message, which GA reads aloud. Would I get the STT and TTS without charge in this scenario (I doubt so)?
Note: A similar question has already been asked here
However the answers for this question only cover costs during the development stage.
Google does not charge for use of the Actions Console, Actions Builder (but see below), or Actions SDK.
So assuming you're doing speech-to-text and text-to-speech strictly for running an Action - there will be no additional charge.
As you note, there may be charges for other Google services if you use them. This may include the built-in code editor with the Actions Builder (which uses Firebase behind the scenes)