>pretty much all records use the maximum gain so I only had to do this
>calibration once and then save the gain settings to an aumix config
>file. Then I used gramofile to split them into tracks
>(semi-automatically) and turned them to MP3s with lame.

I found gramofile introduced artifacts on my system - skips and repeats.

I connect my turntable to a pre-amp then into the Soundblaster card.

I record using sox, and split into tracks using audacity. I have to do it manually, using the audacity graph to see the breaks between tracks. I highlight the bits between the breaks, and save the selected bit as WAV, then repeat.

I encode using bladeenc, I could do it in one hit in audacity, but didn't have the stuff it needed to do that and haven't bothered getting it.

I recommend a dedicated sound card like the Soundblaster, I have tried a couple of inbuilt audio chips and they just were not as good.

Zebee

|I found gramofile introduced artifacts on my system - skips and
|repeats.

Hmm, a couple of my friends have reported that gramofile can also record at the wrong speed, I haven't experienced this as I don't use its record feature; as mentioned, I use bare arecord. I also don't do any filtering. I live with the pops and clicks.

The most tedious part of track splitting is finding the split points. For pop songs, gramofile does well, and I only need tweak its .tracks a bit. What it doesn't cope well with are: fade in/outs, quiet sections (e.g. classical music), concerts with no breaks, and scratchy records (which make its silence detection fail). When that happens, I listen to the whole side using xmms (or the wav player of your choice) and note the split points. It's not so bad, I use the slider to seek around. But I found that the declared lengths on albums are only vaguely related to the actual length. I generate the metadata file using a quick and dirty script I wrote which takes a list of split points from the command line and generates a .tracks file to feed into gramofile. I split using gramofile, move the tracks to their own directory, and run lame over the lot. After that I enter the album info using a command line tool called mpi and then use xmms to enter the song titles. I run mpi again to rename the files by their title instead of processed01.mp3 and I'm done.

I estimate a real time to play time ratio of about 1.4. But I also get suckered to listening to songs I haven't heard for ages, so it's not an unpleasant task. :-)

>The most tedious part of track splitting is finding the split points.

Yeah. Very easy using audacity as you get a visual guide, as it shows the frequency graph.

No noise, no waves…. you can play that section to be sure.

Zebee

> I record using sox, and split into tracks using audacity.  I have to

any reason why you use sox, rather than record straight into audacity?

audacity's a damn good digital editor, i reckon - particularly as it's fairly early into its development. it only seems to have a few very minor bugs. i'm going to have a go at installing it on a couple of machines at the community radio station i work for. they use pro-tools on the main production machine, and quickedit on the news editor, but i doubt much of the production work done there uses any pro-tools/quickedit functions that audacity can't do. if development keeps going, audacity looks set to become the "gimp" of digital editors.

will

>any reason why you use sox, rather than record straight into audacity?

Because I started doing it before I got audacity, and never changed…

>audacity can't do. if development keeps going, audacity looks set to
>become the "gimp" of digital editors.

It's got a lot of good and simple features, and is powerful and easy to use.

Zebee

> Any suggestions on what is the best way (ie: software wise) to do
> this with Linux?

Use the line-in input to your sound card. They you could use gnoise to copy from the input to files. Unless you copy over one track at a time, it will take a long time (and need much disk space) to edit the sound files.

With gnoise and its interpolate function you can also get rid of the worst of the clicks, but someone really needs to come up with a good automatic program to find and eliminate the clicks. (The RIAA correction on the output to the record player makes this more difficult, as those clicks get averaged out by the equalisation, which makes them harder to find and eliminate.)