best voice record codec? Options

[usser]

hello

which is the best codec to use in order to record at best quality conversations from a portable voice recorder?

thanks

[Egor]

The best quality voice codec is a lossless: FLAC, Wavpack or plain PCM.

[[JAZ]]

The best quality voice codec is a lossless: FLAC, Wavpack or plain PCM.

Ahem, I didn't expect such a reply.

Of course, he doesn't specify how does he record them, and where does he want to play them (and more importantly, use them for), but.. sincerely.... a lossless for a voice recording?

Answer for the scenario "I record to .wav and will play on PC, don't need to edit".

Speech codecs:

speex is a good option, at its highest quality. AFAIR, it works best at 32Khz.

Lossy codecs:

Really any of the popular ones should suit. Don't use HE-AAC. there's no need and may change the sound in an unwanted way.

Don't be fooled by "it's a voice recording, so i use a low setting". That logic is actually backwards. It shouldn't be stupid to use LAME -V 3 or so. Also don't play with the stereo setting. The default method should work it out.


You just asked for the best quality, without talking about the scenario. Absolutely best quality is obviously lossless. Here you have the "not that much needed" option.

[chrizoo]

Don't use HE-AAC. there's no need and may change the sound in an unwanted way.

where did you get this info from ??? HE-AAC is supposed to be more advanced then mp3, especially HE-AAC v2.

[[JAZ]]

Yes, it is. When the decoder supports it. Which is not always the case (Still today, iPods and the likes don't play it). Also, note that HE-AAC works at 44khz. Voice is not necesarily sampled at that sampling rate.

Comparing it to mp3 may be misleading. HE-AAC and HE-AACv2 target at having acceptable quality at low to very low bitrates (64kbps down to ...24kbps). Of course mp3 can't compete there, but HE-AAC cannot compete either with LC-AAC at 160Kbps or more.




Edit: Now that i've re-read the OP's question, my answer may not have been the best, but we will never know since he didn't come back.

Concretely, a given hardware recorder would give a few options to select from. Given those, we could have suggested which ones was more adequate.
It may be posible that it can record to wav (would need to know at which quality), or may only give the option to record to a lossy codec like 3gp.

If the question was "which codec support should I look for when buying a new hardware recorder", then the answers may have been different aswell.

This post has been edited by [JAZ]: Mar 13 2010, 13:22

[chrizoo]

sorry, my bad. I am new to AAC and was just told the same thing in another thread.
So by "change the sound in an unwanted way" ... were you referring to a situation where the bitrate was set extremely low, I guess?

[[JAZ]]

No. I was talking about the situation where the decoder would play it at a lower sampling rate (which is what happens if the decoder only knows LC-AAC but not any of the HE-AAC variants).
Also, I've added the comment about supported sampling rate, which could have added an extra step.

[chrizoo]

Oh I see. Good to know, thank you.

I'm afraid your addition ("HE-AAC works at 44khz. Voice is not necesarily sampled at that sampling rate.") bears a new question for me. Is he-aac confined to 44khz only? and what is a more intelligent sampling rate for voice according to you?

[[JAZ]]

According to wikipedia, HE-AAC support has been added recently to iPod's, so I correct my previous comment (note that AFAIK, HE-AACv2 is still unsuported):

http://en.wikipedia.org/wiki/HE-AAC

As of iTunes 9, iPhone OS 3.1 and Fall 2009 iPods, support for HE-AAC playback is included.

Also, seems that i have to correct my statement of "only 44Khz".


http://www.ebu.ch/fr/technical/trev/trev_305-moser.pdf

Typical configurations are 16/32 kHz, 22.05/
44.1 kHz or 24/48 kHz sampling rates, while 8/16 or even 48/96 kHz are also possible.

(Where in X/Y Khz, X means samplerate of the LC-AAC part and Y means the input/output samplerate)

Also, i'll post in the LAME equivalent in AAC world, some more info that this document shows

This post has been edited by [JAZ]: Mar 13 2010, 14:46

[chrizoo]

many thanks for your posting your infos. good to know!

[warduck]

Hi, I have a folder full of phone conversations in wav format (16 kHz, 256 kbps, mono). I would like to archive them but they take up about 3 GB across 300 files. From this thread, it seems a lossy codec is fine and I can choose from MP3, Ogg, Speex and AAC. I wish I had time to sample them all out. If you had to choose, which codec would you recommend and with what parameters? I'm not looking for the smallest file size; just a size more manageable without quality loss and a format that will be playable in 40 years. Appreciate any input, thanks.

This post has been edited by warduck: Apr 16 2010, 23:17

[lvqcl]

and a format that will be playable in 40 years

WAV is the most adequate choice then.

[warduck]

WAV is the most adequate choice then.

Hi lvqcl. Thanks for your reply. Keeping the WAV files does seem the most obvious solution. I've just tried to compress a 1.3 GB sub-folder with 7zip at maximum and it yielded a 937 MB file. Still too large.

I suppose a lossless codec such as FLAC could be a better idea? I don't know how well it would compress them. But I believe it should at least allow me to convert back to WAV without degradation in the event FLAC support appears to be dying.

This post has been edited by warduck: Apr 17 2010, 03:27

[dyneq]

If you had to choose, which codec would you recommend and with what parameters? I'm not looking for the smallest file size; just a size more manageable without quality loss and a format that will be playable in 40 years. Appreciate any input, thanks.

40 years!? Who knows what will be around then? Leave it as WAV and back it up to multiple media (NAS, DVD or multiple CD's).

However, if you'd like to experiment with lossy methods, try the following with LAME:

CODE

-V8 --resample 24

. If you can ABX at this quality level (I have never been able to with my speech files), then go up a quality level and re-ABX until you can't tell the difference.

[Fedot L]

The quality of sound recorded in noisy and/or reverberating conditions depends much more on the position of the microphone (placed the closer possible to the speaker(s), and in this case, well protected with a “pop-filter”, and right facing him (not aside), than on the codec selected.

The quality of speech recorded in those optimal conditions, even with AAC 16 kbit/s, will be better than recorded even in WAV, but in noisy and/or reverberating conditions at a distance from the mic of two, three or more metres. And each 30-40 cm more of distance worsening the situation…

[Notat]

There aren't any voice codecs popular enough to bet on being around for 40 years. WAV will be must durable. MP3 (128 kb, CBR) would be a good second choice. You're also going to have to worry about durability of media over that timeframe. So take dyneq's advice and store on multiple formats and physical locations: put copies on USB sticks, hard drives and CD and or DVD ROMs. Do not worry about the space so much. 1.5 GB will fit on a $10 USB stick, blank media costs a quarter and hard disk space is less than $0.10 per gigabyte. None of this will get you 40 years without some stewardship on your part. To be sure they remain readable, you'll have to keep rolling your recordings to new media and formats as they emerge.

[warduck]

You are all right. I shall keep the WAV's and find more storage space across multiple formats. In 40 years, petabyte storage should be the norm anyway. Thanks very much for your replies.