Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Whatsifsowhatsit]

It seems to be back now. Excellent. And thank you so much for the new DF talk, I'm so excited that that's starting up again. I actually just listened to all the old ones in order again not too long ago.

[voliol]

@ToadyOne Have you contacted anyone to transcribe the Talk, or is that up to the community? The transcripts for #21-22 seems have emerged this way, but I want it confirmed before starting on my own.

[Toady One]

Nobody has contact me yet - I can mark you down in the OP to avoid repeats.

[voliol]

Nobody has contact me yet - I can mark you down in the OP to avoid repeats.

Thanks
I'm currently working my way in from the start, but there is a lot to go through, so if anyone wants to pick up a section further along it'd only make me happier.

[Untrustedlife]

The twenty-third one has been posted.

Any comments about the format, content, quality, etc. of the call are welcome.  The levels on this one are pretty terrible, since we were dealing with multiple headset issues and just barely managed to get it recorded at all.  Hopefully that'll be improved for next time.  DF Talk questions are welcome!

We don't currently have arrangements for a transcript, so that'll need to be worked out in some way over the course of the thread, as with the previous episode. voliol has volunteered to do a transcript.  Currently in progress.

happy to see this after 6 years. Hope it comes back in full force.

[Quatch]

I took a pass at cleaning up the audio, if you wanted to use that. https://old.reddit.com/r/dwarffortress/comments/hi1b5y/dftalk_23_remastered/

steps to reproduce:

• In audacity, open the mp3
• Ctrl-a (select all)
• Effect>notch filter. freq: 8038, Q:15.0
• Effect>noise reduction. Get profile. Select one of the longer louder "silences", all settings default, apply.
• Effect>filter curve. draw a rolloff from 1k@0, 3k@-4, etc https://imgur.com/a/te4ZAB3. The extra notch around 8k in that curve probably isn't nessisary, I was hedging against it changing freq or intensity over time before I listened to the whole thing.
• Effect>compressor. thresh -40, noise -60, ratio 4:1, all other default, apply. Do this last so we don't have stuff we don't want to hear influence how loud things should be.
• Export as mp3, 56k

It'd have been better to split it to a track for each person first, but eh, it mostly works.

[voliol]

@Quatch This is a blessing. The beeping was really getting to me.

[Toady One]

I took a pass at cleaning up the audio, if you wanted to use that. https://old.reddit.com/r/dwarffortress/comments/hi1b5y/dftalk_23_remastered/

steps to reproduce:

• In audacity, open the mp3
• Ctrl-a (select all)
• Effect>notch filter. freq: 8038, Q:15.0
• Effect>noise reduction. Get profile. Select one of the longer louder "silences", all settings default, apply.
• Effect>compressor. thresh -40, noise -60, ratio 4:1, all other default, apply.
• Export as mp3, 56k

It'd have been better to split it to a track for each person first, but eh, it mostly works.

Ah, I had just posted the cleaned up version by Sankis after the site came back up.  Thanks for helping out!  I'm not sure I can make much use of the instructions for the chirp -- my ears are too bad to even hear it in the original recording, and I don't know how to isolate the bad frequency to begin with.

[Quatch]

edit, tldr: You can see the beep in audacity if you switch it to spectral mode.

I used audacity to look at the audio in spectral mode instead of the classic waveform. When you have audio opened (or record a fresh bit) click on the black downward triangle that is on the name of the track (right above the mute/solo buttons). Select spectrogram. Make it nice and big, then you can see the frequency decomposition. Right click on the frequency labels in the axis (0-8k) and 'zoom to fit'.

In the same place under spectrogram settings you can enable "spectral selection" (checkbox at the bottom). This lets you draw a box around a bit of the displayed spectrogram and then it will show you the frequency information at the bottom right of the window (center frequency and width by default).

Also in those settings you can set the max frequency, obv we need to make it a bit higher than 8k in this case. There isn't a lot of speech content above that anyway esp. at low quality mp3, so another alternative is just to lowpass the whole thing at 7k.

If instead you meant how to physically track it down in the room, do a binary search by turning stuff off (phys. unplug those switchmode supplies). It's weird how stable it is, and it still being periodic rather than constant. Wouldn't be surprised if it were something in the computer.

manual entry: https://manual.audacityteam.org/man/spectrogram_view.html

This is what it looks like (taken from near the end): https://imgur.com/a/lqyjqKX. Sankis' version is top right. I can't hear what remains of the beep in that one, so we clearly don't need to make it vanish from the plot.

Actually this reminds me. I did add a rolloff in my version. Oops. Edited my earlier post.

I'm more than willing to give a pass to anything if you need it. It's quick to do.

[Toady One]

Excellent, thanks.  I was trying to do analyze->spectrogram, but that just showed me a frequency range that didn't isolate the culprit.  Now that I've got the spectral mode working, I can still play through a short section of the old file over and over and not hear it, ha ha.  My ears must really be shot.  But this should make it possible to fix.

Getting local recordings is a different matter (we had talked about it, but Capn's mic/recording setup was being wonky and we thought we had record-my-discord setup working well enough, ha ha.)  I do have the ability to adjust the speaker vs. mic levels, which should be enough when there's only one other person (that may or may not be the case going forward depending on XYZ), but we had also lowered mine a bit because of an annoying buzz in my mic, which promptly went away around minute 7, due to some kind of "these headphones charge up" issue (and we can see the noise change and the beep start right there in the spectrogram, so maybe it is my headset that makes the beep when fully charged.)  I should probably test different headsets when I can, but obtaining anything is slightly complicated now, despite there being a store across the street.

[Pillbo]

It was nice to see this pop up in my podcaster, thanks. I hope you continue to do it, seems like a good way to draw interest in the premium release.  Too bad I didn't wait a few days for someone to improve the audio though!

I guess I'm team 'fix shit'... My biggest want it is the Magic release, but since it's many years away I'd prefer the current features worked better during the wait.

[Quatch]

Yeah, that's the same way I tried to find the spectrogram too. I wouldn't have known about it save I saw someone else use it

You can probably record in audacity at the same time you're using the mic to talk (works for me at least). It'd only get your side of things (barring finangling) but if both parties did that then you'd have the raw audio to mix, which would be better for levelling.

[Inarius]

Thanks voliol for your work, it's really appreciated, i can't wait to read all it !

[voliol]

I've gotten through 24 minutes; the first segment and then some. I underestimated transcribing for sure, especially as I'll have to go back to revise what I've written, evening out the punctuation and so forth. There is a certain risk of me not finishing, but I've yet to throw in the towel so we'll see.

The dog video is here: https://twitter.com/2Husketeers/status/1274091818737598466
Or here: https://www.facebook.com/watch/?v=2252524971741859
It's not this one (marbles not replaced), but it's really charming: https://www.youtube.com/watch?v=93neR8L-wuo
Or this one (only one person playing at once): https://www.youtube.com/watch?v=IWXK5eTwQds

[Quatch]

is there a tool or format you're using? I could jump in and do a bit, but it's not something I've done before.