[jingai]

pvr-function: I found a pull request on git. But before I try it myself (as said, I'm really new to this) I'd like to know if one can expect an implementation soon?

This PR will require some reworking/refactoring soon, as we changed a lot under the hood for 2.5 (coming soon).

music-genre: This is really important to me. I found a request somewhere in this thread. But it seems that it wasn't ever implemented. I'd love to say something like "Alexa, ask kodi to play some rock". How about that?

It will be implemented eventually, but for now, I personally just make smart playlists for my genres. I actually prefer it this way because I can lump a bunch of genres together into one thing I can actually remember.

language-specific:
Well, I'm from Germany. So I'd really love Alexa to speak german "Umlaute" (äöü). I found a post from popy who stated he had done some changes concerning "Umlaute". But I just don't know where to start. It sounds silly when alexa sais "playing die Arzte" instead of "die Ärzte"

German support is coming in 2.5.

Next thing, and I don't know if it's really language specific, but currently it's the most annoying for me: alexa doesn't get the right matches for music artists with "special" names.
For example, there's a band called STS. Alexa understands "S.T.S." and can't find the right artist. In the tags and of course in the kodi db it's "STS".
Second example: 2Pac - also in Germany you say it in english. But alexa either matches some other artist or doesn't respond to it anyway. Same thing for a german band with an english name "H-Blockx".
ASP and SOil are other examples and, pretty funny, the german punk band WIZO. It's spoken like "wieso" and the alexa app states that it understood "w.i.z.o.". That's really wired because I would understand it if she understood "wieso". But she did it right with the "z" and without the "e" but she added the dots between the letters (just like for STS). Does this have to do with the fuzzy matching?

Yeah.. this is just kind of how it is for music and voice recognition.

The fuzzy matching tries to catch things like this, but musicians have a tendency to go overboard with their 'creativity'...

One thing I have to add: if I just ask alexa, not kodi, to play 2pac or sts (on prime music) she instantly recognizes the right artist and starts playing.

Amazon has a massive database of alternate pronunciations, spellings, etc. We do not.

They (Amazon) have introduced the idea of "built-in" intents and utterances to allow us to utilize the aforementioned database, but it's only available in the US at the moment. When it's made available worldwide, it's definitely something we're going to look into again.

Custom slots:
I used your generator to populate the custom slots and except in some cases, like mentioned above, it works pretty well. But what I've learned is that if I just use like 500 movie titles (like mentioned somewhere here) alexa does not really match that often if I ask for a movie not mentioned in the slot.
I used "Terminator" as an example. As long as the movie was in the list, she found it. When I excluded this movie she found "Anatomie" (a film that was still in the list).

I am personally an advocate of including everything in the slots, up to the limit, anyway.

The answer really is: do what works for you.

Last thing I wonder is how long should it take alexa for the action? I read about a guy somewhere here who stated that it took about 15 - 20 seconds. That's my experience, too. But I think I also read something about only a few seconds. So, just wondering what's normal and what one can do to improve the reaction time.

When hosting the skill in 'the cloud,' it will take longer. Your Kodi instance has to send lists of files from your library up to the skill, which is mostly dependent on your uplink bandwidth.

I host the skill on my own web server, which is co-located with the Kodi box, so I've never seen a request take > 1 second. Hosting it yourself, however, is obviously a lot more involved.