1. First try uploading ANY replay had failed from my side, I had a message the length can not be established (tried every mod and nomod). Then after another person put a video the upload went okay without problems (tried same replay as previously). Just pointing this out. 2. The sound seems a little off, like a few ms early.
1. First try uploading ANY replay had failed from my side, I had a message the length can not be established (tried every mod and nomod). Then after another person put a video the upload went okay without problems (tried same replay as previously). Just pointing this out. 2. The sound seems a little off, like a few ms early.
Anyways, this is great. Good job there.
1. I applied a fix for that just there now, I think the replay format changed at some point in the past because all the replays you uploaded are missing the length info that's present in all the examples I've put through it until now. I'll have to see about decoding the 7z (I think) stream that most of the replay consists of and/or scraping the highscores page to make a proper fix at some point
2. Increased the offset from +12ms to +16ms, might be a little better now (inb4 I've gone the wrong way )
Shiro wrote:
I am in awe. This thing is absolutely amazing. For the notifications, you can disable them in the options.
Unfortunately that doesn't work, there's a handful of options that don't save correctly under Wine/Linux
I've fixed it so it works as guest now though, not logged into my account any more when recording
Seph
Seph
amazing that it works on ctb too, though no online scoreboards ;;
For some reason it says it detected DOUBLETIME/NIGHTCORE in nomod replay, causing the video to end too early.
Removed DT/NC 'support' (I have a sneaking suspicion it was coincidental it worked at all ) entirely, better to record past the end of DT than to stop early on non-DT
I'll look into the proper method of length reading (ie decoding the 7z) tomorrow, if my server is still in one piece!
EDIT: i can also help you adjust the offset if need be, looks a wee bit late
You really don't want to, it's horrendous
Probably not going to open source it any time soon, but I'll send you a wee pm to peek through when I get a chance to clean it up tomorrow
As for the offset, late/early always confuses me: assuming I'm padding the start of the audio track with 16ms of blank, should I increase or decrease that padding?
Offset is a tad bit late.. but other than that it's fine. I also tested a taiko replay on it because all of the other replays looked good. Not a good choice to make because the play field is constantly moving so you can tell when frame rate drops(?) or if there is lag(?), sometimes the taiko play field starts shaking like a leaf in a blizzard
Offset is a tad bit late.. but other than that it's fine. I also tested a taiko replay on it because all of the other replays looked good. Not a good choice to make because the play field is constantly moving so you can tell when frame rate drops(?) or if there is lag(?), sometimes the taiko play field starts shaking like a leaf in a blizzard
Aye the GPU and hard disk can't always keep up sadly, if only I had a basic discrete gpu in there along with twice the ram
I've taken out the 16ms padding entirely, see if that's any better now offset-wise
should be, UO is rarely needed except on very old sound cards in my experience
also, regarding messy code: you've never seen my contributions to osu! source code.. i feel like an elephant living in a bird colony when compared to the other devs sometimes!
should be, UO is rarely needed except on very old sound cards in my experience
Aye you're probably right, the audio stack it's going through scares me though
Various alsa hooks, probably a bit of pulseaudio thrown in along the way, hnnnnng
At least I'm not trying to asynchronously capture from the pulseaudio equivalent of stereo mix and sync it up with a variable frame rate video (any more), now THAT was madness!
Also speaking of the osu source code, would it be possible to get the code for decoding replays? I'm at a total loose end with determining length/doubletime properly, I guess what I need is in the mysteriously compressed part
Offset is a tad bit late.. but other than that it's fine. I also tested a taiko replay on it because all of the other replays looked good. Not a good choice to make because the play field is constantly moving so you can tell when frame rate drops(?) or if there is lag(?), sometimes the taiko play field starts shaking like a leaf in a blizzard
Aye the GPU and hard disk can't always keep up sadly, if only I had a basic discrete gpu in there along with twice the ram
I've taken out the 16ms padding entirely, see if that's any better now offset-wise
Better, I also tested a song that was more than 6 minutes long. I was seeing how far into the replay it would go.
Update: thanks to statementreply, doubletime/nightcore detection now works perfectly - no more running on way past the end
Additionally, I've added code which scrapes the length from the online page for the correct difficulty if it can't be determined from the HP/time graph data in the .osr file, so basically no more recordings stopping early.
Update: thanks to statementreply, doubletime/nightcore detection now works perfectly - no more running on way past the end
Additionally, I've added code which scrapes the length from the online page for the correct difficulty if it can't be determined from the HP/time graph data in the .osr file, so basically no more recordings stopping early.
Nice fix, would that also include songs longer than 5 minutes? That "test" I ran capped the replay at 6:08, which wasn't too far off though. The replay only had a few more seconds left on it.
Edit: The offset is better too might need some fine tuning but other than that, it was fine.
You should add a historical display of previous conversions on the page somewhere, in my opinion.
Yeah I've been recording the data for that, just need to add a page to display it
In fact the whole UI needs redone properly, it's literally a 30 minute job as it stands
Kuro wrote:
Darkimmortal wrote:
Update: thanks to statementreply, doubletime/nightcore detection now works perfectly - no more running on way past the end
Additionally, I've added code which scrapes the length from the online page for the correct difficulty if it can't be determined from the HP/time graph data in the .osr file, so basically no more recordings stopping early.
Nice fix, would that also include songs longer than 5 minutes? That "test" I ran capped the replay at 6:08, which wasn't too far off though. The replay only had a few more seconds left on it.
That's probably due to the second thing I fixed, so it's worth trying again
Are you sure you're uploading the replay(.osr) and not the .osu?
All of your replays should be saved in C:\Program Files (x86)\osu!\Data\r or C:\Program Files\osu!\Data\r depending on your processor and OS Make sure the extension ends in .osr