docker-jitsi-meet icon indicating copy to clipboard operation
docker-jitsi-meet copied to clipboard

Enable VOSK Transcription

Open janonym1 opened this issue 2 years ago • 36 comments

this creates 2 (maybe) new ENV variables to allow for easier VOSK integration. Per default, it is assumed GCLOUD is always used and that makes it hard(er) to set up VOSK within docker-jitsi-meet

I still need to add the GCLOUD check and set for the ENV GOOGLE_APPLICATION_CREDENTIALS /config/key.json within the run script (jigasi/rootfs/etc/services.d/jigasi/run) but I am not sure how to best approach this. Maybe lets define an additional variable (ENABLE_GCLOUD_TRANSCRIPTION) and use that?

janonym1 avatar Jul 05 '22 09:07 janonym1

I added a new var ENABLE_GCLOUD_TRANSCRIPTION but I am not sure, if my code at jigasi/rootfs/etc/services.d/jigasi/run seting the GCLOUD ENV for the key.json is working as intended

janonym1 avatar Jul 05 '22 10:07 janonym1

I also am not sure, if it would be better/nicer if we can chain the check for the GCLOUD creds in the 10-config:

I wanted to AND together the following checks: if [[ ($ENABLE_TRANSCRIPTIONS -eq 1 || $ENABLE_TRANSCRIPTIONS == "true") && ($ENABLE_VOSK -eq 0 || $ENABLE_VOSK == "false") && ($ENABLE_GCLOUD_TRANSCRIPTION -eq 1 || $ENABLE_GCLOUD_TRANSCRIPTION == "true") ]] but that isnt working as I imagined it. I assume I am misunderstanding some syntaxes here?

janonym1 avatar Jul 05 '22 10:07 janonym1

Is there anything left to do? I would be happy to help.

It would be very useful when this gets merged. Thanks

debendraoli avatar Aug 21 '22 06:08 debendraoli

Is there anything left to do? I would be happy to help.

It would be very useful when this gets merged. Thanks

I updated the requested changes (e.g. the variablenaming ENABLE_VOSK_TRANSCRIPTION) but I need to test it out on my testsystem first, which I unfortunately didnt get around to yet. If I botched up something, I also break the GCLOUD transcription ability, which more users seem to use than VOSK :)

Also I am not sure about the logic, that checks if transcriptions are on (but not gcloud) which I will get around to next week hopefully

janonym1 avatar Aug 22 '22 09:08 janonym1

Hi, is there any update on this, and is there anything I can do to help?

charles-zablit avatar Oct 05 '22 14:10 charles-zablit

@janonym1 is there an update on this ? Would love to help if necessary.

rouaidakacem avatar Nov 30 '22 09:11 rouaidakacem

@debendraoli @charles-zablit @rouaidakacem and everyone else asking to help: could you test this PR and report if it works, and in case you have to make any changes, which ones? Because even with my proposed change, I can't get this to work.

HarHarLinks avatar Nov 30 '22 09:11 HarHarLinks

Been patiently waiting for this, hoping to see this in a release soon! Would love to help/test if there's an update on this

saiflakhani avatar Sep 11 '23 13:09 saiflakhani

There is currently a conflict that needs to be fixed.

saghul avatar Sep 11 '23 13:09 saghul

Hi @saghul What needs to be done here? I would like to help. It looks like the changes you requested have not been reviewed by you. Or have I missed something?

How can I/we help to get this merged?

loelkes avatar Jan 30 '24 19:01 loelkes

Yep, conflicts are yet to be solved.

saghul avatar Jan 31 '24 06:01 saghul

@saghul I understand that. But from what I can see, the changes you requested have been made but not reviewed. Can you review them and point out the remaining conflicts?

I don't have write access here, so all I could do was create a new branch/PR to address the conflicts there.

loelkes avatar Jan 31 '24 07:01 loelkes

Please do that. I can't merge the PR as is, even if the feedback was addressed, because there are conflicts.

saghul avatar Jan 31 '24 12:01 saghul

Sorry, I must have overlooked the activity here. I solved the conflicts but I think I have a bug somehwere in the logic. Let me look for my hardcoded values (since it worked for me before) and crosscheck with the effective values here

janonym1 avatar Jan 31 '24 13:01 janonym1

I have a simlar patch, so I'd rather 👍 this than to post my own. I am also willing to help in order to get this merged.

zobadaniel avatar Feb 07 '24 13:02 zobadaniel

I forgot to add, that you have to have a working SIP setup (server+account) and config for VOSK to work. Should I just assume it is setup or make a check (e.g. if JIGASI_SIP_URI is set)?

I also do not seem to get around the GCLOUD restriction; it needs to have a jigasi/config/key.json with valid credentials even when GCLOUD is deactivated. I am not sure how to get around this, maybe mock the credentials?

janonym1 avatar Mar 27 '24 13:03 janonym1

I forgot to add, that you have to have a working SIP setup (server+account) and config for VOSK to work. Should I just assume it is setup or make a check (e.g. if JIGASI_SIP_URI is set)?

This is not a jigasi issue, this is configuration issue and should be fixed, maybe in this PR?

damencho avatar Mar 27 '24 14:03 damencho

I think I fixed that already. If you set ENABLE_TRANSCRIPTIONS the sip config is skipped now.

saghul avatar Mar 27 '24 16:03 saghul