[Jitsi] Configure Appearance of Closed Captions/Subtitles

[alexk2037]

Hello, I was wondering how I could configure the appearance of the closed captioning on Jitsi Meet?

Currently it looks like this for me:

You can see that the subtitles are stacked on top of one as the text-to-speech program processes the audio. However, I would like the closed captioning to appear more like Google Meet, in that the subtitles stay in one line and get overwritten in place. For example, this video about Jitsi closed captioning demonstrates the behaviour I am looking for.

I am currently using Vosk Server to handle the text-to-speech. It looks to me like the subtitles are showing multiple lines of partial text at once, however, I’m not sure where I can configure the UI to display one line of partial text at a time, or even just the full text for a sentence.

On the Jitsi community forums, I received a reply stating:

I think there is something wrong with the vosk implementation - like some transcriptions are not coming as a correction of the previous results, but like new transcriptions. Corrections should replace the previous one if visible and not duplicate them …

I can write down my Vosk Server implementation if that may be useful. Any help, advice, or links would be appreciated. Thanks.

[alexk2037]

My Vosk Server Implementation

Machine: AWS EC2 Server Operating System: Ubuntu 18.04 Python Version: Python 3.8

Steps

Create Python virtual environment

$ python3 -m venv venv
$ source venv/bin/activate

Download Vosk Server

$ git clone https://github.com/alphacep/vosk-server.git

Install dependencies

$ python3 -m pip install aiortc aiohttp aiorpc vosk websockets

Download Speech Recognition model:

$ cd
# Download accurate wideband model
$ wget http://alphacephei.com/vosk/models/vosk-model-en-us-0.20.zip
$ unzip vosk-model-en-us-0.20.zip

# Rename the downloaded model
$ mv vosk-model-en-us-0.20 model

Create System Daemon

$ mkdir -p ~/.config/systemd/user
$ touch ~/.config/systemd/user/vosk_service.service

$ systemctl --user list-unit-files | grep vosk_service
>>> vosk_service.service           masked

Write to .service File

$ nano ~/.config/systemd/user/vosk_service.service

[Unit]
# Human readable name of the unit
Description=Vosk Service

[Service]
# Command to execute when the service is started
WorkingDirectory=/home/ubuntu
ExecStart=/home/ubuntu/venv/bin/python3 /home/ubuntu/vosk-server/websocket/asr_server.py

# Disable Python's buffering of STDOUT and STDERR, so that output from the
# service shows up immediately in systemd's logs
Environment=PYTHONUNBUFFERED=1

# Automatically restart the service if it crashes
Restart=on-failure

# Our service will notify systemd once it is up and running
Type=simple

[Install]
# Tell systemd to automatically start this service when the system boots
# (assuming the service is enabled)
WantedBy=default.target

Activate service

$ systemctl --user daemon-reload
$ systemctl --user start vosk_service
$ systemctl --user status vosk_service

Enable Vosk to start on bootup:

$ systemctl --user enable vosk_service

Check Logs

$ journalctl --user -u vosk_service
# Or
$ journalctl --user -u vosk_service --since today

[nshmyrev]

Same issue as https://github.com/alphacep/vosk-server/issues/89

[LuisFerchx]

Does this work just in Linux? Is there another way for windows?

[nshmyrev]

@LuisFerchx it should work on Windows as well. Please try and let us know