Feedback /bugs with new audio transcription

The new audio transcription feels like it’s going in the right direction but is plagued with UX issues.

I like that it feels more contextual and fluid but it’s currently a poor experience and I’ve lost work multiple times.

The fact that you can type while it transcribes is potentially a good direction to go in as you can then fluidly type and talk but the currently implementation has problems..

  1. if you type and talk for a while and hit the X you lose everything. and cannot undo or get it back
    2.hitting tick moves your cursor
  2. if you type something and move the cursor, the next transcription continues of from previous position, you have to stop and start

I think it would be better if :

  • you just removed X and Tick
  • and the next words transcribed were placed at the cursor.
  • any edits from typing or talking were undoable.

Hey @tomcomp, thanks for the detailed follow-up, and sorry about the lost text.

Quick context on the controls: while recording, X discards the dictation and the check mark stops recording and keeps the text. With the new live transcription, cancelling also rolls the box back to how it looked before you started, which is what is wiping the words you typed in between.

If it helps, holding Ctrl+M to talk avoids the X and check buttons entirely, and using the check mark rather than X keeps everything in the box.

Thanks for the tip on Cntrl-M - I’ll try that.

  • I have also noticed that using the Tick sometimes undoes the manual change made while or after talking.

  • Also I still question the value of the X to clear everything, especially given :

  1. that it cannot be undone
  2. a talking/typing might last considerably time and a lot of typing effort.
  3. The X button is similar to most Stop buttons in normal transcription inputs which don’t clear the whole chat. So easy to press and then lost work.
  4. Undo already works within the live transcript (Great!)

So I think it would be preferably if:

  • just a stop button to leave things as they are

  • leave user to use undo as needed

  • And lastly, being able to continue transcription from cursor position would make it 10x more productive.

Thanks!!

I just tried Cntrl-M - handy for adding short bursts but because you have to hold both keys down while talking, it means :

  • slow to jump between Cntrl M and making changes
  • difficult to navigate computer while talking and then edit transcription mistakes.

So it has it’s purpose but I still advocate for changes above.
Thanks

Here is example of the bug where it ignoring my typing changes on submit. The first 2 transcriptions are fine but the 3rd and 4th attempt undo my change and add some characters to the end.

Sorry to keep pestering but I think having Stop and Submit would be the right thing - like on chatgpt.

Stop - stop recording mode
Submit - stop and submits (actioned by enter too)

At the moment, if you just dictate, you have press tick and then submit. which means I often have stuff I have dictated but forgotton to submit and I come back to it and it’s still there wiating to be submitted.

So in 90% times you would press submit and occassionaly if you still have some images or other content to paste, then you press stop.