Audio recording improvements
complete
D
Dan Brady
I'm using Friday every day and mostly enjoying it however sometimes it does become frustrating so I thought I would add a few feature requests and improvements here.
- Firstly sometimes Friday is not receiving sound from my Apple AirPods. It happens fairly frequently. Sometimes it seems when my AirPods switch to my phone and then back to my Mac, Friday and no longer detect the AirPods as a microphone device. However, this does not affect Apple dictation or my ability to listen to music on the Air Pods, so I believe it is a Friday issue. Disconnecting and reconnecting my AirPods sometimes helps and sometimes doesn't.
- When Friday cannot detect audio but is effectively listening and no audio is sent to ChatGPT the ChatGPT API responds with a generic response (e.g. "Please provide the text you need to be corrected, and I will be happy to help you fix any grammar, spelling, and punctuation errors") which is then used in place of the audio that I had dictated. Friday should detect when it is sending an empty audio file to ChatGPT and simply prevent this.
- I would like Friday to automatically delete audio recordings after a set period of time. Ideally, I would configure this in settings: one hour, one day, seven days, 30 days, never, etc.
- Make it easy to bulk delete all audio recordings, e.g. a 'Delete all recordings' button.
Edited to fix typos caused by using Apple dictation because Friday wasn't hearing my audio :/
Naveen
complete
in v0.30, voice to text functionality has been improved. This issue is now closed. Please submit a new request if you encounter any issues with voice to text. thank you!
D
Dan Brady
Naveen, this is amazing. Thank you. These delete recordings changes (v0.27) are exactly what I needed. The retry transcription option is nice, too.
I have realised that my point number 1 was probably my iPhone hijacking the connection with my AirPods. Since realising this, I deselect the AirPods on my iPhone whenever there are problems with audio reaching Friday GPT and it's worked fine.
On point 2, here's what Friday (via ChatGPT) returned for some empty audio just now (v0.27):
"Thanks for watching, good luck, and bye."
Maybe I should publish a book of ChatGPT's hallucinations!
Keep up the good work, this app keeps improving, and I use it every day :)
Naveen
Dan Brady: I see option 1 is a bit difficult for me to solve. It's good to know that it's not a pain.
I'm still thinking about how to solve empty audio issue. Do you know when you are sending empty audio?
D
Dan Brady
Naveen Empty audio was a problem when I couldn't figure out why my audio wasn't reaching Friday. Now that I've sorted that, it's mostly gone away. However, there is a problem that happens sometimes when I dictate a short piece of audio and – As I have a grammar check chained – it is found to be grammatically correct by ChatGPT. Instead of returning the transcription of the audio, I get a message saying something like, "The text you supplied was grammatically correct."
Naveen
Dan Brady I see. Issue is then with the grammar fix prompt. You can update the prompt saying "if the grammar is correct, just return the text." I would suggest incrementally updating the grammar fix prompt whenever you find these types of cases.
D
Dan Brady
NaveenThanks, that makes sense.
Naveen
in progress
Naveen
I'm still working on improving the voice to text feature. Made a few incremental improvements in v0.27(released today)
- You can now delete all old recordings
- Added a setting to auto-delete old recordings (set this up in Record Settings)
- Retry the transcription in the Recording History
Naveen
planned
Naveen
Thank you for the detailed feedback. I will work on improving the feature
- Will show the audio input and letting user to change the input manually so that this issue occurrence can be decreased
- will handle the case of empty audio
- nice idea, will add it.
- on it