I've got several tapes that i'd like to convert to mp3 format. They're sermons, so there isn't really any complicated sounds in them.

Currently, i'm just playing the tapes on my little rca bookshelf system, and plugging a double-ended male 1/8" plug from the headphone out on the rca to line in on my laptop soundcard. i set my volume to get a good level (they're recorded rather uniformly), and record the wave into Cool edit pro.

now....should i just use cool edit to save the wav, and if so, what bit rate should i use? i've already figured out how to get rid of the tape noise using cool edit, so that's pretty cool. i want to get a quality sound, but i'm not sure at what point it would be overkill (i'm sure the sound from this old tape isn't all that good...my initial guess was that 64kbps would be a good bit rate....do y'all think 32 would work, and if i'm that low, like i said, would the fraunhofer that cool edit uses be best?)

should i look into getting a higher quality tape player (sure i could find one for pretty cheap) with a line out?

or should i get one of those new echo indido i/o cards for my laptop, to get a higher quality signal.


basically; i'm doing this for my church, and another church, but only as a volunteer. i don't want to drop a TON of cash, but i will spend a little money, if it's wisely spent.

thank you for any input that you may have!!!

Generally, the word spread here is ABX, but in your case, you might want to pass. ABX means to encode at different settings and see if you can tell which sounds better when blind. When you've reached a setting that you can't distinguish, then you've found your bitrate.

I would say record a few minutes of it, and then see what you like better. Hydrogen Audio generally uses Lame 3.90.3 for mp3 encoding, although at bitrates that low, I don't know how much of a difference it will make. Lame DOES have a sweet ABR flag that will encode it at about the same size as a constant bitrate file, but the sound will be better. For that reason alone, I'd suggest saving the Cool Edit output as a wave.

If the tapes aren't that good to begin with, then there probably isn't a very good reason to get a high quality tape player, now is there?

As for the Indigo, if you're really interested in audio on your laptop in general, then I'd say go and get it. I've heard nothing but good things about it. But if you're just going to use it for this task, then I'd be reluctant to tell you to get it.

If you're willing to use something OTHER than mp3, Speex might be right appropriate for this task.

yeah....i got the lame stuff...i'm currently ripping all my music for my new nomad jukebox zen (pretty tight little toy) using eac and lame. still playing w/ that, but that's another story.


i guess a clarification is: i've seen some audiobooks encoded at 32kbps (they used to be tapes), but that seems REALLY low, even though it sounds alright to me. i am pretty good at picking out artifacts, but usually on a high quality car stereo, not on the tiny little headphones that i've got now, so there could be artifacts all OVER the place at 19khz, and i wouldn't know it w/ what i'm listening to with.


and...it will have to stay w/ mp3; these have to be very distributable, easy to use files, so they're really user-friendly.

perhaps the answer is: get some better headphones to start with, and then go from there.

other question: is there a nice program that will allow me to abx a selection of files? is it CALLED abx?

that's basically what i planned on doing, but i wanted a place to START shooting from, and if i should go with cbr, abr, or vbr.

sorry that i'm still clarifying my question.


grace,
shawn

Trying to achieve 32kbps ABR for speech with LAME, I came up with this set of options (about a year and a half ago) that worked better for me than anything else I came up with:That's set to 38kbps target ABR, but assuming there are (more-or-less) silent pauses between phrases & sentences, it uses a bunch of 8kbps frames to achieve an actual average bitrate of ~32kbps.

incredibly tight.

i'll try that out.


it seems like speex would work well, BUT, it has to be easily distributable. unfortunately, when you still have to have relatively in-depth instructions to tell people how to play an mp3, you can't really swap to a different format.

(unless i misunderstand how speex works....).

my understanding: i could encode a resultant wav file into the speex format, and then ?somehow? get that onto an audio cd, but playing it in its digital format is not that simple; possible, but not as easy as an mp3 file. am i correct?

That's why I'm in favour of MP3 format for audio that has to be widely distributed - because it's almost universally playable. "But won't <insert favourite format here> sound better?" Probably, but trying to get users who don't really understand the term "download" to get a <favourite format>-compatible player installed is just not worth the trouble. Not to mention Typical User won't notice the difference.

cool....i really appreciate it, man. i'll try that out. going to have to figure out how to encode w/ lame when it's all by itself, and not a part of eac, but i'll figure it out somehow.


now....i've STILL got to figure out the best way to get the wav INTO the computer; assuming that what i'm doing now is not the best way.

I've heard artifacts at 32kbps, but of course, since I'm listening to comedy, it's not that difficult. What I've found is that the voice just doesn't sound as full. But given non-studio settings, that's just about bound to happen anyways.

There IS a possible reason why you might want to stick with CBR. That answer is legacy mp3 players/decoders. I've had a couple mp3s that started off with a low bitrate, and ended up playing really fast, because the player couldn't handle VBR.

As for the question about where to find ABX, RTFFAQ (Read The Fine FAQ)


To use lame, for people who aren't familiar with console interfaces, do a google search on RazorLame.

CoolEdit is probably just as good as anything else you could probably use to get audio off of a tape

I don't know all the flags for lame, but is one of those mono? Generally with speech there isn't much point to stereo.

Often if I'm making a mp3 CD I include copy of winamp on it just
so I will be able to play it on any PC.

As to whether to seek a better player, it really depends on how it was
recorded.. some 30 year old (now) dictation style portable (no dolby no nuthin) in the 3rd row or a "hi fi" deck plugged in to the sound system?

-a = downmix stereo -> mono (if you input a stereo source, mono source will obviously stay mono)
-m m = mode mono

--lowpass 7.5 = cut off frequencies above 7.5 kHz
--resample 22 = resample input to 22 kHz sampling frequency
-h = "higher quality, but a little slower. Recommended."
-b 8 = minimum allowed frame bitrate is 8 kbps
-B 64 = maximum allowed frame bitrate is 64 kbps
--abr 38 = set target Average Bit Rate to 38kbps (doesn't take into account "silent" portions, so is higher than final target bitrate of 32 kbps). Adjust this number to achieve the target bitrate you want.
--athlower -50 = raises level of "too quiet to hear" to just below quietest speech - allows "silent" portions of recording (which likely contain irrelevant background noise, breathing, paper shuffling, etc) between sentences to be encoded at minimum bitrate (8 kbps), which means you can devote a larger portion of your average 32 kbps to encoding voice rather than "silence". Adjust this number if you're getting a lot of high-quality-encoded "silence", or if quiet words are encoded at 8 kbps. I just looked (with EncSpot) at a dozen samples encoded with that command line and they all have between 15% and 25% of their frames at 8 kbps, depending on the speaking style, how long pauses are between words & phrases.

I almost hate to say it (because I seem to do so a lot), but foobar2000 is freely distributable, and I believe the speex plugin is too. You could set a CD up easily that autoruns a playlist and presents the user with a playlist detailing all the sermons on it. That is, assuming your target audience consists of Windows PCs only. This would take every last in-depth instruction away. Note that this could probably be done with Winamp too, albeit more confusingly.

1. Insert CD.
2. Wait for a window to come up with a list of sermons.
3. If the CD player isn't making any noises and nothing appears to be happening, open "My Computer".
4. Right-click on your CD drive, then click AutoPlay.

You may, assuming that quality is a bit of a concern, try a format with a higher bitrate. Vorbis would do well with an easy speech signal, as would Musepack.

BTW, on the Lame commandline above, wouldn't it make more sense to resample to 16kHz, as your cut-off is at 7.5? (Nyquist limit for that is 15kHz.)

Why bother setting a -b and -B? why not just let the encoder work out how to get a 38kbps average?

Sorry to reccomend this but WMA 9 voice at 24Kbit is good for sermons(curently it's the only voice codec that sounds decent for voice and music). That's what my church has been using for archiving old sermons. Everything else is stored lossless.

WMA 9 isn't good for accessability. FWIW, mp3 is the most playable compressed format.

yeah....one of these churches that i'll be doing it for is in nyc, in manhatten, and will have several media people in it...i'm QUITE positive that the people who would be d/l these sermons will have macs.

not my cup of tea, but that's the deal-e-o.



one general notation: i know from experience, that with m3u2sb and daemon tools (a sneaky way to bypass schmuck-like sony minidisc's OMG jukebox software), the file has to be 44hz, 16bit stereo.

how much file size would this add to the aforementioned LAME mp3? i realize that it's an obscure requirement (being able to put on minidisc easily) but, if it won't add much size, it would be worth doing.

Hmmm... This may be a stupid question, but as the lowpass is 7.5 kHz why not use 16 kHz sampling instead of 22 kHz? I would think that would lower the possible bit rate even further.

A potentially stupid question will get a potentially stupid answer.

My guess is that 22KHz can be extracted easily from a 44.1KHz source, and vice versa.

Storing a mono-signal file as stereo will degrade audio quality. Encoding at a sampling frequency > 24kHz (ie 32, 44 or 48) requires MPEG-1, and your frame bitrate range goes from 8-160kbps for MPEG-2/2.5 up to 32-320kbps, which pretty much negates my method of using 8kbps "silent" frames to gain higher quality encoding, and there's no possibility of ABR < 32kbps. You'll need to find a way to reample to 44/stereo on decode (I assume the aforementioned tools require WAV input, or can they handle MP3 input as well?).

I'm not sure that bitrate (or since this is ABR, quality) would be affected one way or the other by resampling to 16 rather than 22, but for situations where you need to decode to CD-style 44/16/2 I believe it's easier to go 22->44 than 16->44.


I've just been experimenting with --abr 38 vs --alt-preset 38, and the alt-preset bitrate distribution looks better to me than the ABR version (the bell of the curve is concentrated more at lower bitrates). I know modifying presets typically gets you shot around here, but maybe it's acceptable in this case? If so, this would be my current suggestion:
For comparison, here are the two bitrate distributions:

audible sound difference between the 2? file size diff?


i won't shoot you. though.....you *are* a canuck....hmmmmm.

...to get it to stereo in the first place, i just record the tape mono in cool edit, then convert the wave to stereo. both channels are identical.

i initially tried recording in stereo, but the tape is in stereo, and it had some really wierd waveforms....like, one time, at like 75% of the sermon, the right channel went down to almost nothing, and the left was still kicking away.

after that, i started doing it the new way. rec. mono--->conv. stereo--->encode w/ cool edit.