@Sebastian_Berlin-WMSE OK, we have changed the lexicon import since then, using the newer lexicon Braxen instead which has another transcription. It's hard to say without context why your version reads as "O" despite the transcription in the NST lexicon. My guess is that it has to do with the audio database and the fact that there may be very few occurences of "och" being pronounced with a "k".

Looks correct in 2026-04:1. Need more context and output audio.

Both these cases have now been added to the textproc in wikispeech-server version 2026-04:1.

"Längden är 123 mm." works in the latest text processning. "Massan är 123 kg." does not expand correctly, since this abbreviation had not yet been added. We will add "kg" to the text processning.

Looks find in the latest textproc version:

I'm not sure where you are looking, but in the lexicon currently used in the latest version of Wikispeech, it is transcribed /" o:/. Currently realised more like /" O/ in TTS (due to how this word is pronounced in the source audio database).

I don't have the full sentence, but here is what I get from the text processing component:

Mishkal needs another configuration, better to add as a separate ticket.

In T263114#10608719, @Sebastian_Berlin-WMSE wrote:

@HannaLindgren were you looking at this? Is this task of any use or can we close it?

Not in scope

Not in scope

Not relevant

Not relevant

MaryTTS-related

In T177209#7226842, @kalle wrote:

Is this still relevant?

js dependencies removed from symbolset

js dependencies removed from pronlex