Rendered at 12:14:00 GMT+0000 (Coordinated Universal Time) with Cloudflare Workers.
joshstrange 2 days ago [-]
Tried this out, and it might be due to other apps on my computer or something but the audio it records has a echo. When I go to play back a speaker's voice it sounds I can hear what sounds like 2 audio streams, slightly offset. Maybe it's something to do with SoundSource/ARC or my other meeting transcribing app.
One more feature I'd love is the ability to export the text/transcript in markdown automatically. I have another tool (MeetingTranscriber) that does this and I like that so I can put the transcripts in my Obsidian Vault which I already expose to the various agents I use.
elektor 2 days ago [-]
What is the url of the MeetingTranscribe tool? Google is giving me generic results.
On the echo, a few questions so I can pin it down:
1. Which version are you on (Parrot menu > About Parrot)? 0.24.0 fixed a bug where the mic track and the system audio track could start a few seconds apart, which sounds exactly like two copies of the same voice. If you're on something older, updating fixes new recordings (old ones stay as they are, sorry).
2. Were you on headphones or speakers? On speakers the mic also picks up the other person. The echo canceller removes most of it but not all, so you can get a faint second copy a split second behind.
3. Roughly how far apart are the two copies? A short slapback points at mic bleed, a clear repeat of whole words points at timing or routing.
If you have a minute, one short test call with SoundSource and the other transcriber quit would tell us a lot. Both hook into system audio, and I want to rule out Parrot recording the call twice.
On Markdown: that already exists, it's just tucked away in settings. Settings > Connections, pick a folder (your Obsidian vault works) and switch on automatic saving. Every finished meeting lands there as a note with front matter, summary, next steps as checkboxes and the full transcript. Saving again updates the same note instead of making a copy.
And since you mentioned agents: Parrot can also connect straight to Claude, Cursor or Codex (Settings > Claude & AI Apps), so they can search your meetings directly.
joshstrange 20 hours ago [-]
1. I was running: Version 0.24.2 (202609301306)
2. I am wearing headphones so I wouldn't expect any feedback
I will try to test with other things turned off to see if that makes a difference and I'll update.
turantekin 1 days ago [-]
[dead]
moecables 2 days ago [-]
instead of calling it "co-pilot" maybe call it something like "Coach", Assistant, or anything that doesn't connect it to Copilot from Microsoft b/c that's what I thought it meant
turantekin 22 hours ago [-]
Done it is called Assistant now. :) Try to update the app pls.
turantekin 2 days ago [-]
Agree. I don’t know why I called it copilot assistant sounds good. I will work on that tomorrow thank thank you so much.
1 days ago [-]
turantekin 1 days ago [-]
[dead]
1shooner 2 days ago [-]
I agree, 'copilot' is overloaded even within the Microsoft portfolio.
turantekin 2 days ago [-]
Yeah you are right. I will work on that!
marshalla 23 hours ago [-]
This one looks super nice. Trying it now. I've got one I'm working on full-time and trying to get people to pay for, and in some ways this beats it. I'm using localVQE for echo cancellation. I really just need to give up and get some corporate job
turantekin 22 hours ago [-]
Do not give up! To be frank i didn't built it to turn it in to a business, i have enough troubles with the businesses i am running, i dont need one more :) I just wanted to solve my problem that is it.
On parrot i used SpeexDSP for echo cancelation, didnt try localVQE, are you happy with result?
marshalla 9 hours ago [-]
Oh neat, I hadn't even seen or tested SpeexDSP. Pretty happy with localVQE 1.2, it's a little extra on the CPU, but the cancellation is pretty complete.
serg937 1 days ago [-]
great app, installed it and will try it out. One comment for now: please can you make it possible to attach a folder to documents? I guess, in the meantime I will keep there one file & update it after each meeting with the summaries and recent notes & decisions made
turantekin 22 hours ago [-]
Thank you so much. Please try and share your feedback. Do you mind to submit as a feature request on github? Happy to work on that sounds good.
Hi looks nice but can it handle remote meetings with speaker on? So the audio stream gets recorded back thru the mic? Most of these meeting apps struggle with that.
turantekin 22 hours ago [-]
Hey
Mostly yes. Parrot takes recording straight from your mac so it is always clean. I have pushed a new version 0.25.2 please try, should be fine.
turantekin 1 days ago [-]
[dead]
catoc 1 days ago [-]
Apart from the confusing co-pilot invocation, this looks really cool! Will try it out.
turantekin 1 days ago [-]
Thanks! The name is going, you're the third person to say it. Let me know how it goes.
catoc 18 hours ago [-]
Yes, just installed from source and saw the name change.
Really liked the onboarding!
(kudos - definitely not something that could be one-shot vibe-coded)
turantekin 1 days ago [-]
Hey! Took your point on the name, it's the Assistant as of 0.25.1.
nilleb 2 days ago [-]
amazing! Great initiative, thank you! I have contributed to a similar app (called Recap) a few months ago and wanted to change it to something similar to Parrot - you anticipated me, thank you!
turantekin 2 days ago [-]
Great! Well I had been using it for myself for a long time for my work. Plus I am enjoying building it, so much fun. Any feedback please feel free to reach me.
satvikpendem 2 days ago [-]
There are so many of these already out there.
turantekin 1 days ago [-]
I know, I've tried most of them myself for my business. Each time there was something I didn't like, or something I needed was missing. That's why I built this one for myself.
librasteve 1 days ago [-]
kudos for going open source … good business decision too
turantekin 1 days ago [-]
Thank you. I love open source.
vivzkestrel 1 days ago [-]
- anyone got a real time "original voice" to "AI voice" implementation using parrot or parakeet or whatever bird is flying this month?
- bonus if it works with OBS studio
dv35z 1 days ago [-]
Can you explain more what you are looking for, and what you are hoping to achieve?
vivzkestrel 23 hours ago [-]
- i do game recordings
- my actual voice is terrible and i have a lot of noise in vicinity to sit and process it
- what i want is to be able to talk into the mic and an AI voice is generated in real time with a very large degree of accuracy
- bonus if it works with OBS as a plugin or something
dv35z 13 hours ago [-]
That's interesting -
I have been playing with Whisper speech-to-text recently (both CLI, and also . Depending on the model & computer performance, it seems like it can stream (listen and transcribe to text) audio. Maybe you can pipe that to a text-to-speech service (remote or local), and configure that as an audio input to OBS.
Are you live-streaming, or just recording video file to edit/publish later? If it's not live, that may relax some of the latency requirements.
Somewhat related, Descript mentions it does dubbing. I recently used this tool to edit a video, and I was impressed by how easy it was, and accurate too. It may not fit your realtime use-case, but might be worth exploring...
One more feature I'd love is the ability to export the text/transcript in markdown automatically. I have another tool (MeetingTranscriber) that does this and I like that so I can put the transcripts in my Obsidian Vault which I already expose to the various agents I use.
On the echo, a few questions so I can pin it down:
1. Which version are you on (Parrot menu > About Parrot)? 0.24.0 fixed a bug where the mic track and the system audio track could start a few seconds apart, which sounds exactly like two copies of the same voice. If you're on something older, updating fixes new recordings (old ones stay as they are, sorry).
2. Were you on headphones or speakers? On speakers the mic also picks up the other person. The echo canceller removes most of it but not all, so you can get a faint second copy a split second behind.
3. Roughly how far apart are the two copies? A short slapback points at mic bleed, a clear repeat of whole words points at timing or routing.
If you have a minute, one short test call with SoundSource and the other transcriber quit would tell us a lot. Both hook into system audio, and I want to rule out Parrot recording the call twice.
Happy to dig in on GitHub if that's easier: https://github.com/turantekin/Parrot/issues
On Markdown: that already exists, it's just tucked away in settings. Settings > Connections, pick a folder (your Obsidian vault works) and switch on automatic saving. Every finished meeting lands there as a note with front matter, summary, next steps as checkboxes and the full transcript. Saving again updates the same note instead of making a copy.
And since you mentioned agents: Parrot can also connect straight to Claude, Cursor or Codex (Settings > Claude & AI Apps), so they can search your meetings directly.
2. I am wearing headphones so I wouldn't expect any feedback
3. They are ever so slightly off. My voice is perfect, it's the other attendees that I hear with a slight echo, less than second and a slight static-y sound to it. I can't find. Example: https://www.dropbox.com/scl/fi/yditx2ykv4zgjvoa0v2f3/Example...
I will try to test with other things turned off to see if that makes a difference and I'll update.
On parrot i used SpeexDSP for echo cancelation, didnt try localVQE, are you happy with result?
https://github.com/turantekin/Parrot/discussions/categories/...
Mostly yes. Parrot takes recording straight from your mac so it is always clean. I have pushed a new version 0.25.2 please try, should be fine.
Really liked the onboarding! (kudos - definitely not something that could be one-shot vibe-coded)
- bonus if it works with OBS studio
- my actual voice is terrible and i have a lot of noise in vicinity to sit and process it
- what i want is to be able to talk into the mic and an AI voice is generated in real time with a very large degree of accuracy
- bonus if it works with OBS as a plugin or something
I have been playing with Whisper speech-to-text recently (both CLI, and also . Depending on the model & computer performance, it seems like it can stream (listen and transcribe to text) audio. Maybe you can pipe that to a text-to-speech service (remote or local), and configure that as an audio input to OBS.
Are you live-streaming, or just recording video file to edit/publish later? If it's not live, that may relax some of the latency requirements.
Related links:
- https://github.com/openai/whisper (speech-to-text for terminal)
- https://buzzcaptions.com/ (graphical interface to Whisper)
Somewhat related, Descript mentions it does dubbing. I recently used this tool to edit a video, and I was impressed by how easy it was, and accurate too. It may not fit your realtime use-case, but might be worth exploring...
https://descript.com/
Good luck!
- what exactly is the architecture of speech to speech?
- is it speech to text and text to speech or am I badly understating the actual mechanism