How about we add the ability to translate the user speech to another language, live?

The setup would be the same as we have, in the sense that we capture the audio, but rather than just processing it and echoing it back to the user, we pipe that through one of the classic AI models, which are also capable of translating text to other languages. The challenge here might be in doing the live translation. Similar to this, but a simultaneous translation.

Clearly this would also depend on #1 so that we can speak it back to the user

The initial work on this project's UI has been carried out in Svelte but, in order to remain consistent with the other NF apps, we should convert it to React before beginning any other work

When double clicking the record message button right after recording a message, the recorded message gets duplicated on the screen.

We need to make the application deployable to GCP.

Hint: use the same approach to bundle everything as the llm-playground project

Can't create new chatrooms due to a constraint that is not needed anymore.

The front page needs to be updated according to the provided design.

The Figma designs can be found here

When the participant in the chat room hits the "record" button the app should record and recognise the voice until the user presses the "stop recording" button.

The current behaviour is a bit erratic and shows a few issues:

• while talking with small pauses between words - restarts the voice recognition and everything spoken before is lost
• the "stop recording" button is enabled while voice recognition is still processing the spoken words, this should be disabled

Let's focus only on the Google streaming solution for this.

Let's set up the app so that it speaks the text back to the user as soon as it comes in.

We can use the Web Speech API for this https://developer.mozilla.org/en-US/docs/Web/API/Web_Speech_API

The UI l&f needs to be updated to be consistent/more inline with NearForm design language.

Update the Readme file to explain how to run the project locally.

As the title implies, we need to setup Firebase authentication for the application and set up rules on the RT DB as well.

The server is using Firebase RT DB to store the chatrooms and the events. After setting up rules that only authenticated users can access the DB, server is failing now as it's missing the authentication for the service account.

The chatroom page needs to be updated according to the provided design.

The Figma designs can be found here

          After playing around with #3 , I realize that in practice this would be quite awkward to do, because speaking while you're getting your message spoken back to you in a different language as you speak, doesn't make for a terribly pleasant conversation. In any case let's work on this under the assumption that the recipient of the translated audio would be a different person. We could consider, as an extension of this work (but not as part of the PR for this work) to set up a simple p2p audio conversation so you could:

• speak in English to, say, an Italian
• they would hear Italian voice and talk back in Italian
• you would hear what they said in English

Obviously, replace Italian with any other language

Originally posted by @simoneb in #2 (comment)

• When typing a character in the language dropdown, the selection should jump to the first option that starts with this character
• fix font family
• Add flags next to user names in the chatroom per design.

Set CSS rules for the layout so it supports self-contained scrolling for every message container that contains enough messages to exceed it's initial height.

This behavior is expected on both home page and the chatroom page.

Implement live translation functionality. "Add a voice message" button should be replaced with speech detection toggle (which should be enabled by default).