Most efficient audio format for voice recordings

[Don Hyams]

I have many audio recordings I need to archive in digital format. I'm
trying to find the most efficient audio format (i.e., smaller file
size) that won't compromise the clarity significantly.

The original recordings are voices in a meeting, recorded directly to
an MP3 recorder deck in 16-bit, 44100, joint-stereo MPEG 1 Layer 3.
The original source (mixing board output) was mono.

Right now, an hour of audio results in a file of approximately 55MB.
I've experimented with a few different bit rates and played with the
RealPlayer format, but haven't run across a significant improvement.

I guess the bottom line is... which audio format and sampling rate is
most efficient for a mono, non-music voice recording?

Thanks in advance.

--Don

[KingMe]

On 10 Nov 2004 10:57:44 -0800, Don Hyams wrote:

> I have many audio recordings I need to archive in digital format. I'm
> trying to find the most efficient audio format (i.e., smaller file
> size) that won't compromise the clarity significantly.
>
> The original recordings are voices in a meeting, recorded directly to
> an MP3 recorder deck in 16-bit, 44100, joint-stereo MPEG 1 Layer 3.
> The original source (mixing board output) was mono.
>
> Right now, an hour of audio results in a file of approximately 55MB.
> I've experimented with a few different bit rates and played with the
> RealPlayer format, but haven't run across a significant improvement.
>
> I guess the bottom line is... which audio format and sampling rate is
> most efficient for a mono, non-music voice recording?
>
> Thanks in advance.
>
> --Don

You might want to look at some codec specific for voice compression,
depending on what type of quality you want.

[Tony]

On 10 Nov 2004 10:57:44 -0800, (Don Hyams) wrote:

>I have many audio recordings I need to archive in digital format. I'm
>trying to find the most efficient audio format (i.e., smaller file
>size) that won't compromise the clarity significantly.
>
>The original recordings are voices in a meeting, recorded directly to
>an MP3 recorder deck in 16-bit, 44100, joint-stereo MPEG 1 Layer 3.
>The original source (mixing board output) was mono.
>
>Right now, an hour of audio results in a file of approximately 55MB.
>I've experimented with a few different bit rates and played with the
>RealPlayer format, but haven't run across a significant improvement.

That surprises me, as I've found that as long as I use a decent
encoder (Fraunhoffer or Lame, not Blade or Xing) I can archive voice
only (meetings etc) at rates as low as 16kb/s (7.2MB/hr) with
acceptable results. It's possible WMA may be better still at low
rates, but Lame's 16k MP3 format suits me fine.

Tony (remove the "_" to reply by email)

[Bob Cain]

Don Hyams wrote:

> I guess the bottom line is... which audio format and sampling rate is
> most efficient for a mono, non-music voice recording?

Microsoft's WMA format has one of the best reputations right
now for low bitrate encoding. You could probably sample at
half that rate or less, or downsample if you use a DAW,
without any loss of intelligibility.

I don't know, however if WMA supports lower sample rates.


Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein

[Bob Cain]

Don Hyams wrote:

> I guess the bottom line is... which audio format and sampling rate is
> most efficient for a mono, non-music voice recording?

Microsoft's WMA format has one of the best reputations right
now for low bitrate encoding. You could probably sample at
half that rate or less, or downsample if you use a DAW,
without any loss of intelligibility.

I don't know, however if WMA supports lower sample rates.


Bob
--

"Things should be described as simply as possible, but no
simpler."

A. Einstein

[Logan Shaw]

Don Hyams wrote:

> I have many audio recordings I need to archive in digital format. I'm
> trying to find the most efficient audio format (i.e., smaller file
> size) that won't compromise the clarity significantly.
>
> The original recordings are voices in a meeting, recorded directly to
> an MP3 recorder deck in 16-bit, 44100, joint-stereo MPEG 1 Layer 3.
> The original source (mixing board output) was mono.

Well, stereo of any kind seems like a waste for a mono signal, but
hopefully with joint stereo, it's going to be minimized pretty well.

You could probably do a lot better just within the MP3 format by
going to mono, converting the sample rate down to 22050 Hz or down
to 16000 Hz or even 12000 or 8000 Hz, and dropping the bitrate (and
maybe going to variable bitrate). Of course, how much you want to
tweak these parameters depends on how much quality you want to lose.

If you want to ditch the MP3 format, there are several speech-specific
codecs out there. GSM phones compress speech digitally, and I believe
software that the GSM codecs is available. There are also several
other codecs out there, for example the open-source, patent-free
Speex ( http://www.speex.org/ ).

- Logan

[Logan Shaw]

Don Hyams wrote:

> I have many audio recordings I need to archive in digital format. I'm
> trying to find the most efficient audio format (i.e., smaller file
> size) that won't compromise the clarity significantly.
>
> The original recordings are voices in a meeting, recorded directly to
> an MP3 recorder deck in 16-bit, 44100, joint-stereo MPEG 1 Layer 3.
> The original source (mixing board output) was mono.

Well, stereo of any kind seems like a waste for a mono signal, but
hopefully with joint stereo, it's going to be minimized pretty well.

You could probably do a lot better just within the MP3 format by
going to mono, converting the sample rate down to 22050 Hz or down
to 16000 Hz or even 12000 or 8000 Hz, and dropping the bitrate (and
maybe going to variable bitrate). Of course, how much you want to
tweak these parameters depends on how much quality you want to lose.

If you want to ditch the MP3 format, there are several speech-specific
codecs out there. GSM phones compress speech digitally, and I believe
software that the GSM codecs is available. There are also several
other codecs out there, for example the open-source, patent-free
Speex ( http://www.speex.org/ ).

- Logan

[Arny Krueger]

"Bob Cain" > wrote in message

> Don Hyams wrote:
>
>> I guess the bottom line is... which audio format and sampling rate is
>> most efficient for a mono, non-music voice recording?
>
> Microsoft's WMA format has one of the best reputations right
> now for low bitrate encoding. You could probably sample at
> half that rate or less, or downsample if you use a DAW,
> without any loss of intelligibility.
>
> I don't know, however if WMA supports lower sample rates.


http://www.microsoft.com/windows/windowsmedia/howto/articles/OptimizingLowBitRate.aspx

"Many Internet broadcasters and content providers use the Windows Media
Audio 9 codec to deliver audio content at 32 kilobits per second (Kbps) with
quality that is similar to FM radio. Others use bit rates of 64 Kbps to 128
Kbps to deliver quality similar to audio CDs.

"However, there are a large number of Internet broadcasters and other
producers who can take advantage of very low bit rate audio to deliver AM
radio quality music and voice-only streams over slow and unreliable
networks. The standard codec streams as low as 5 Kbps with a sampling rate
of 8 kilohertz (kHz), and is designed to compress a variety of sounds.
However, if your content is voice-only, you may be able to achieve higher
quality at lower bit rates by using the Windows Media Audio 9 Voice codec,
which is designed to optimize quality when the content is voice. The voice
codec can also compress as low as 4 Kbps at 8 kHz. If audio contains a mix
of music and voice, you can configure the voice codec to switch
automatically between the voice and music compression algorithms.

"At very low bit rates, you achieve audio similar to telephone quality. The
secret to getting the best sound at very low bit rates is to understand how
the codec works. You can achieve the best voice quality by making sure you
record voice content in a way that helps the codec do the best job of
compression.

[Arny Krueger]

"Bob Cain" > wrote in message

> Don Hyams wrote:
>
>> I guess the bottom line is... which audio format and sampling rate is
>> most efficient for a mono, non-music voice recording?
>
> Microsoft's WMA format has one of the best reputations right
> now for low bitrate encoding. You could probably sample at
> half that rate or less, or downsample if you use a DAW,
> without any loss of intelligibility.
>
> I don't know, however if WMA supports lower sample rates.


http://www.microsoft.com/windows/windowsmedia/howto/articles/OptimizingLowBitRate.aspx

"Many Internet broadcasters and content providers use the Windows Media
Audio 9 codec to deliver audio content at 32 kilobits per second (Kbps) with
quality that is similar to FM radio. Others use bit rates of 64 Kbps to 128
Kbps to deliver quality similar to audio CDs.

"However, there are a large number of Internet broadcasters and other
producers who can take advantage of very low bit rate audio to deliver AM
radio quality music and voice-only streams over slow and unreliable
networks. The standard codec streams as low as 5 Kbps with a sampling rate
of 8 kilohertz (kHz), and is designed to compress a variety of sounds.
However, if your content is voice-only, you may be able to achieve higher
quality at lower bit rates by using the Windows Media Audio 9 Voice codec,
which is designed to optimize quality when the content is voice. The voice
codec can also compress as low as 4 Kbps at 8 kHz. If audio contains a mix
of music and voice, you can configure the voice codec to switch
automatically between the voice and music compression algorithms.

"At very low bit rates, you achieve audio similar to telephone quality. The
secret to getting the best sound at very low bit rates is to understand how
the codec works. You can achieve the best voice quality by making sure you
record voice content in a way that helps the codec do the best job of
compression.

[Mike Rivers]

In article > writes:

> I have many audio recordings I need to archive in digital format. I'm
> trying to find the most efficient audio format (i.e., smaller file
> size) that won't compromise the clarity significantly.

If you really mean "archive" in its true sense - that you want to be
able to play them 50 years from now - then you'd better byte the
bullet and save them in an uncompressed format like WAV. What's a
little disk space compared to not having the decoder program any
longer?


--
I'm really Mike Rivers )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

[Mike Rivers]

In article > writes:

> I have many audio recordings I need to archive in digital format. I'm
> trying to find the most efficient audio format (i.e., smaller file
> size) that won't compromise the clarity significantly.

If you really mean "archive" in its true sense - that you want to be
able to play them 50 years from now - then you'd better byte the
bullet and save them in an uncompressed format like WAV. What's a
little disk space compared to not having the decoder program any
longer?


--
I'm really Mike Rivers )
However, until the spam goes away or Hell freezes over,
lots of IP addresses are blocked from this system. If
you e-mail me and it bounces, use your secret decoder ring
and reach me here: double-m-eleven-double-zero at yahoo

[Ty Ford]

On Wed, 10 Nov 2004 13:57:44 -0500, Don Hyams wrote
(in article >):

> I have many audio recordings I need to archive in digital format. I'm
> trying to find the most efficient audio format (i.e., smaller file
> size) that won't compromise the clarity significantly.
>
> The original recordings are voices in a meeting, recorded directly to
> an MP3 recorder deck in 16-bit, 44100, joint-stereo MPEG 1 Layer 3.
> The original source (mixing board output) was mono.
>
> Right now, an hour of audio results in a file of approximately 55MB.
> I've experimented with a few different bit rates and played with the
> RealPlayer format, but haven't run across a significant improvement.
>
> I guess the bottom line is... which audio format and sampling rate is
> most efficient for a mono, non-music voice recording?
>
> Thanks in advance.
>
> --Don

What quality level do you need?

Ty Ford



-- Ty Ford's equipment reviews, audio samples, rates and other audiocentric
stuff are at www.tyford.com

[Ty Ford]

On Wed, 10 Nov 2004 13:57:44 -0500, Don Hyams wrote
(in article >):

> I have many audio recordings I need to archive in digital format. I'm
> trying to find the most efficient audio format (i.e., smaller file
> size) that won't compromise the clarity significantly.
>
> The original recordings are voices in a meeting, recorded directly to
> an MP3 recorder deck in 16-bit, 44100, joint-stereo MPEG 1 Layer 3.
> The original source (mixing board output) was mono.
>
> Right now, an hour of audio results in a file of approximately 55MB.
> I've experimented with a few different bit rates and played with the
> RealPlayer format, but haven't run across a significant improvement.
>
> I guess the bottom line is... which audio format and sampling rate is
> most efficient for a mono, non-music voice recording?
>
> Thanks in advance.
>
> --Don

What quality level do you need?

Ty Ford



-- Ty Ford's equipment reviews, audio samples, rates and other audiocentric
stuff are at www.tyford.com

[normanstrong]

>
> Right now, an hour of audio results in a file of approximately 55MB.
> I've experimented with a few different bit rates and played with the
> RealPlayer format, but haven't run across a significant improvement.
>
> I guess the bottom line is... which audio format and sampling rate
is
> most efficient for a mono, non-music voice recording?

I've done this several times. I used WMA at 5kb/s, mono. The file
you mentioned above would have taken about 2.4MB. A single CDR blank
would hold about 340 hours of speech. Perfectly intelligible, too.

Norm Strong

[normanstrong]

>
> Right now, an hour of audio results in a file of approximately 55MB.
> I've experimented with a few different bit rates and played with the
> RealPlayer format, but haven't run across a significant improvement.
>
> I guess the bottom line is... which audio format and sampling rate
is
> most efficient for a mono, non-music voice recording?

I've done this several times. I used WMA at 5kb/s, mono. The file
you mentioned above would have taken about 2.4MB. A single CDR blank
would hold about 340 hours of speech. Perfectly intelligible, too.

Norm Strong