import struct

def get_chunk(f):
 header = f.read(8)
 id, size = struct.unpack('<4sI', header)

 return id, size, f.read(size)

f = open('music.riffogg', 'r')

header = f.read(12)

riff, size, type = struct.unpack('<4sI4s', header)

assert riff == 'RIFF'

print 'RIFF file:'
print 'File size: %d' % size
print 'File type: "%s"' % type

def get_vorbis(f):
 while True:
 id, size, data = get_chunk(f)
 print id

 if id == 'data':
 out = open('vorb.ogg', 'wb')
 out.write(data)
 out.close()
 print 'wrote vorbis data to vorb.ogg'
 return data

d = get_vorbis(f)

• fmt: A WAVE chunk, it seems to be filled with valid data: format: 65535 channels: 2 rate: 48000 bitrate:6000


• vorb (around 40 bytes): No idea. Probably important for something.


• data: the Vorbis stream, you can find vorbis headers in it. starts with 0x01 + vorbis, 0x03 + vorbis, 0x05 + vorbis for the first three ones. problem is how to find out where all the other vorbis packets begin and end...

Originally by: Anders1

---

• data: the Vorbis stream, you can find vorbis headers in it. starts with 0x01 + vorbis, 0x03 + vorbis, 0x05 + vorbis for the first three ones. problem is how to find out where all the other vorbis packets begin and end...

---

Originally by: Sleepkevert

---

Originally by: Anders1

---

• data: the Vorbis stream, you can find vorbis headers in it. starts with 0x01 + vorbis, 0x03 + vorbis, 0x05 + vorbis for the first three ones. problem is how to find out where all the other vorbis packets begin and end...

---

There is some file layout definitions up on msdn that might help you a bit in finding that out.

---

Originally by: Hinaj Katsu

---

Did you made any progress?

---

struct eve_vorb_chunk {
 uint32_t unknown1; /* could be number of samples
 * as this divided by the rate (48000)
 * seems to give reasonable lengths,
 * like 5 minutes etc.
 */

 uint32_t unknown2; /* the rest I haven't figured out yet */
 uint32_t unknown3;
 uint32_t unknown4;
 uint32_t unknown5;
 uint32_t unknown6;
 uint32_t unknown7;
 uint32_t unknown8;
 uint32_t unknown9;
 uint32_t unknown10;
 uint32_t unknown11;
} typedef eve_vorb_chunk_t;

field offset size value hex
unknown1 0 32 bits 15802618 00f120fa
unknown2 4 32 bits 15738 00003d7a
unknown3 8 32 bits 0 00000000
unknown4 12 32 bits 3893 00000f35
unknown5 16 32 bits 633 00000279
unknown6 20 32 bits 2203653 0021a005
unknown7 24 32 bits 0 00000000
unknown8 28 32 bits 3893 00000f35
unknown9 32 32 bits 356 00000164
unknown10 36 32 bits 15392 00003c20
unknown11 40 32 bits 15760 00003d90

Originally by: Anders1

---

For some reason all the EVE music is no longer in mp3, but now in Ogg. Nothing will play these files, which seemed odd, so I looked into it. Turns out these

---

Originally by: Catari Taga

---

Originally by: Anders1

---

For some reason all the EVE music is no longer in mp3, but now in Ogg. Nothing will play these files, which seemed odd, so I looked into it. Turns out these

---

It was always in Ogg vorbis format, but a pure ogg format. Anyway I just wanted to link you to this thread, might have some helpful info like the link to the sdk for the audio engine.

---

Originally by: Sleepkevert

---

Sooooo, I did a lot of reading this weekend and I toyed around with the new files.

It looks like the fmt is just fake placeholder data to get it to be valid (hence the 0xFFFF dev codec header). I have no idea what the vorb actually contains but I'll be taking a second look at that this afternoon to determine if there is anything in there that will give the sample rate.

The data contained in the data bit is (from what I'v been able to determine) a raw vorbis stream. Now, with some fiddeling about and determening the audio packet boundries as specified in the vorbis documentation 4.3 Audio packet decode and synthesis. You'll probably be able to re-pack the vorbis stream into an OggS container.

The only problem I want to figure out before I start coding the damn thing is how to get the PCM audio samples of the stream, since you need that to set the granule position of the OggS headers.
How to re-code the vorbis stream into OggS containers is all described in A Embedding Vorbis into an Ogg stream of the manual if you haven't find that already.

---

Originally by: Sleepkevert

---

Edited by: Sleepkevert on 04/06/2009 11:17:01
From what I'v been reading yesterday / today, it looks like the granule position is only really used for acurate file length determination, and possibly stuff like searchbars and the like. you can still probably play the files without the granule position (or possibly a faked, number calculated from the total number of samples to at least display song length correctly).

I'v got a basic OggS packet packer going (in c# ) As well as a simple RIFF extractor and a start for a vorbis decoder. Code is ugly as hell but it does what it's intended to do. I now face the callenge to go and determine audio packet boundries. Which should be fun figuring out from the setup header and block size... not to mention the setup header size in the first place. But it's all pretty well documented luckily. I'll see if I can spent a couple more hours on it tonight.

---

Originally by: Anders1

---

For some reason all the EVE music is no longer in mp3, but now in Ogg. Nothing will play these files, which seemed odd, so I looked into it.

---

Originally by: Elaron

---

Originally by: Anders1

---

For some reason all the EVE music is no longer in mp3, but now in Ogg. Nothing will play these files, which seemed odd, so I looked into it.

---

This news story gives the change some perspective, in my opinion.

---

Originally by: Alexeph Stoekai

---

Originally by: Elaron

---

Originally by: Anders1

---

For some reason all the EVE music is no longer in mp3, but now in Ogg. Nothing will play these files, which seemed odd, so I looked into it.

---

This news story gives the change some perspective, in my opinion.

---

I don't see it. Care to point out what's so enlightening?

---

Originally by: Bunyip

---

The LOLqual is a capital-sized joke

---

Originally by: Hinaj Katsu

---

Thank you very much for the work you've done.

---

Originally by: Eugene Spencer

---

Edited by: Eugene Spencer on 05/08/2009 11:58:46
Fantastic! This is great stuff :)

It downloaded super quick. The only issue is the link you provided gives a "malware detected" error in Google Chrome - "Warning: Visiting this site may harm your computer!".

I've uploaded the torrent to my site, so people hopefully won't freak it like I did haha.

http://files.gumpcom.com/EVEMusic-20090803.torrent

---

Originally by: Rootax

---

Edited by: Rootax on 16/08/2009 16:38:40
Thx you for this !

Few seeders tough... Anyway it's downloading :)

I'll seed like forever after it's finished.

---

Originally by: Sleepkevert

---

Edited by: Sleepkevert on 18/08/2009 23:57:31
So you figured it out after all eh? Nice work! And thanks for the torrent :)

Adding an 100mbit seed in there for half a year orso

Bla bla long post deleted

Long story short, the files converted play crappy in foobar2000. Turns out it was due to the specific packaging in the converter. The converter was packaging every audio packet into an OGG page. This adds extra overhead since you can easily put about 30-40 packets into a page (this is usualy done). So I adapted some code from the mutagen audio library to do pack it nicely. Upside is that it plays nice in every player and it's a bit smaller due to the removed overhead from the ogg pages (from 198 mb to 170 mb). Downside is that the progress bar in VLC seems to be broken (foobar and winamp work fine though).

The code I used is avaidable here.

Once again, thanks for writing the converter. I'll be enjoying the EVE tracks once again

---

Originally by: Sleepkevert

---

Edited by: Sleepkevert on 18/08/2009 23:57:31
So you figured it out after all eh? Nice work! And thanks for the torrent :)

Adding an 100mbit seed in there for half a year orso

Bla bla long post deleted

Long story short, the files converted play crappy in foobar2000. Turns out it was due to the specific packaging in the converter. The converter was packaging every audio packet into an OGG page. This adds extra overhead since you can easily put about 30-40 packets into a page (this is usualy done). So I adapted some code from the mutagen audio library to do pack it nicely. Upside is that it plays nice in every player and it's a bit smaller due to the removed overhead from the ogg pages (from 198 mb to 170 mb). Downside is that the progress bar in VLC seems to be broken (foobar and winamp work fine though).

The code I used is avaidable here.

Once again, thanks for writing the converter. I'll be enjoying the EVE tracks once again

---

Usage: wwiseconv.py [options] [args]

Options:
 --version show program's version number and exit
 -h, --help show this help message and exit
 -i FILE, --input=FILE
 WWise Vorbis to convert
 -o FILE, --output=FILE
 save converted audio as FILE
 -D, --debug enable debug output
 -v, --verbose enable verbose output
 -q, --quiet surpress all output
 -d DESTDIR, --destdir=DESTDIR
 save files to DEST, only used with --eve-soundbank

 EVE Online specific options:
 -b FILE, --eve-soundbank=FILE
 use SoundbanksInfo.xml for filenames
 -p FILE, --eve-playlist=FILE
 use resPlaylists.stuff for track titles

Originally by: Sleepkevert

---

Somehow still doesn't play smoothly in foobar. It is definately a lot better though, the speed changes are nearly gone, it sounds like some audio packets, or packet parts are getting dropped.

I'll do some more checking in my break in a few hours, updating foobar, playing with the packer a bit.

---

Originally by: Anders1

---

Edited by: Anders1 on 12/09/2009 01:12:02
Uploaded a torrent: http://dl.getdropbox.com/u/773354/EVEMusic-20090912.torrent

---

/* == typedefs =================================== */

typedef unsigned char u8;
typedef unsigned int16 u16;
typedef unsigned int32 u32;
typedef unsigned int64 u64;

typedef signed char s8;
typedef signed int16 s16;
typedef signed int32 s32;
typedef signed int64 s64;

/* == custom types ================================ */

typedef char ID[4];

typedef char UUID[16] <read=UUIDRead, write=UUIDWrite>;

string UUIDRead(UUID u)
{
 string s;
 SPrintf(s, "{%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x%02x%02x}", u[0], u[1], u[2], u[3], u[4], u[5], u[6], u[7], u[8], u[9], u[10], u[11], u[12], u[13], u[14], u[15]);
 return s;
}

void UUIDWrite(UUID &u, string s)
{
 SScanf(s, "{%02x%02x%02x%02x-%02x%02x-%02x%02x-%02x%02x%02x%02x%02x%02x%02x%02x}", u[0], u[1], u[2], u[3], u[4], u[5], u[6], u[7], u[8], u[9], u[10], u[11], u[12], u[13], u[14], u[15]);
}

/* == structures ============================== */

struct RIFFHeader
{
 SetBackColor( cLtGray );
 ID magic;
 u32 chunkSize;
 ID type;
};

/* Generic chunk */
struct UnknownChunk
{
 SetBackColor( cGray );
 ID chunkID;
 u32 chunkSize;
 char data[chunkSize];
};

struct FormatChunk
{
 SetBackColor( cLtBlue );
 ID chunkID;
 u32 chunkSize;

 u16 format;
 u16 channels;
 u32 sampleRate;
 u32 avgByteRate;
 u16 blockAlign;
 u16 bitsPerSample;
 u16 size;
 if (size == 22)
 {
 u16 validBitsPerSample;
 u32 channelMask;
 UUID subFormat;
 }
};

struct VorbChunk
{
 SetBackColor( cLtGreen );
 ID chunkID;
 u32 chunkSize;

 u32 samples;
 u32 unknown1;
 u32 unknown2;
 u32 unknown3;
 u32 unknown4;
 u32 unknown5;
 u32 seekTableSize;
 u32 codebooksOffset;
 u32 unknown8;
 u32 unknown9;
 u32 unknown10;
 u32 checksum; // of what?
 u32 unknown12;
};

struct DataChunk
{
 SetBackColor( cLtYellow );
 ID chunkID;
 u32 chunkSize;

 local int origPos = FTell();

 u16 codebooksSize;
 u16 whoKnows;
 u16 unknown; /* why is codebooksSize above 32-bits and unknown 16-bits */
 /* "unknown" is used if codebooksSize is 0 because there's a seek table */

 local u32 sz;
 if (codebooksSize == 0)
 {
 sz = unknown - 6; /* what's 6 */
 Printf("unknown: %d unknown - 6: %d\n", unknown, unknown - 6);
 Printf("vorb.unknown3 - sz = %d\n", vorb.unknown3 - sz);
 SetBackColor(cGray);
 char seekTable[vorb.seekTableSize];
 }
 else
 {
 sz = codebooksSize;
 Printf("vorb.unknown3 = %d; codebooksSize = %d\n", vorb.unknown3, codebooksSize);
 Printf("diff: %d\n", vorb.unknown3 - codebooksSize);
 }

 SetBackColor(cLtRed);

 char codebooks[sz];

 FSeek(origPos + chunkSize);
};

/* == read =========================== */

LittleEndian();

local char id[5];
RIFFHeader header;

if (header.magic != "RIFF" || header.type != "WAVE")
{
 Warning("File is not a valid WWise file");
 return -1;
}

while (!FEof())
{
 /* peek chunk ID */
 ReadBytes(id, FTell(), 4);
 id[4] = 0;

 switch (id)
 {
 case "fmt ":
 FormatChunk fmt;
 break;

 case "vorb":
 VorbChunk vorb;
 break;

 case "data":
 DataChunk data;
 break;

 default:
 UnknownChunk unknown;
 break;
 }
}

local u32 riffHeaderSize = 12;
local u32 fmtHeaderSize = 8 + fmt.chunkSize;
local u32 vorbHeaderSize = 8 + vorb.chunkSize;
local u32 dataOffset = 8 + riffHeaderSize + fmtHeaderSize + vorbHeaderSize;
Printf("offset to data: %d\n", dataOffset);
Printf("offset to codebooks end: %d\n", dataOffset + vorb.codebooksOffset);

import os
import struct
from cStringIO import StringIO
import sys

from mutagen.ogg import OggPage

class RIFF(object):
def __init__(self, stream):
self.stream = stream
self.magic, self.size, self.type = struct.unpack('<4sL4s', stream.read(12))

class RIFFChunk(object):
def __init__(self, riff):
self.id, self.size = struct.unpack('<4sL', riff.stream.read(8))
self.data = riff.stream.read(self.size)

class FormatChunk(RIFFChunk):
def __init__(self, riff):
super(FormatChunk, self).__init__(riff)
unpacked =struct.unpack('<HHLLHHHHL16s', self.data)
self.format = unpacked[0]
self.channels = unpacked[1]
self.sample_rate = unpacked[2]
self.avg_byte_rate = unpacked[3]
self.block_align = unpacked[4]
self.bits_per_sample = unpacked[5]
self.extension_size = unpacked[6]
self.valid_bits_per_sample = unpacked[7]
self.channel_mask = unpacked[8]
self.sub_format = unpacked[9]

class VorbChunk(RIFFChunk):
def __init__(self, riff):
super(VorbChunk, self).__init__(riff)

unpacked = struct.unpack('<LLLLBBHLLLLLLLL', self.data)
self.samples = unpacked[0]
self.blocksize = unpacked[4]
self.checksum = unpacked[13]
self.audio_offset = unpacked[9]

class AudioPacket(object):
def __init__(self, stream):
self.size, self.granule_pos = struct.unpack('<HL', stream.read(6))
self.data = stream.read(self.size)

class DataChunk(RIFFChunk):
def __init__(self, riff):
self.id, self.size = struct.unpack('<4sL', riff.stream.read(8))
self.data = StringIO(riff.stream.read(self.size))

def main():
f = open(r'C:\Sisi\res\audio\336401897.ogg', 'rb')
riff = RIFF(f)
fmt = FormatChunk(riff)
vorb = VorbChunk(riff)
data = DataChunk(riff)

pages = []

bsizes = {64: 6, 128: 7, 4096: 12, 1024: 10, 8192: 13, 256: 8, 512: 9, 2048: 11}

# hardcoded values
os = StringIO()
os.write('\x01vorbis')
os.write(struct.pack('<L', 0)) # vorbis version
os.write(struct.pack('<B', 2)) # channels
os.write(struct.pack('<L', 48000)) # rate
os.write(struct.pack('<L', 0)) # bitrate max
os.write(struct.pack('<L', 48000)) # bitrate nominal
os.write(struct.pack('<L', 0)) # bitrate min
os.write(struct.pack('<B', 0xc9)) # should be correct.. although i got it to 0x9c first, but then it complains as blocksize_0 > blocksize_1. so use c9.
os.write('\x01')

page = OggPage()
page.packets.append(os.getvalue())
pages.append(page)

os = StringIO()
os.write('\x03vorbis')
vendor = 'WWise unknown'
os.write(struct.pack('<L', len(vendor)))
os.write(vendor)
os.write(struct.pack('<L', 0)) # comment list length
os.write('\x01')

page = OggPage()
page.packets.append(os.getvalue())
pages.append(page)

max_len = 0

stream = data.data
count = 0
while data.size - stream.tell() > 0:
count = count + 1
packet = AudioPacket(stream)
#print "read %d packets" % count

# first packet contains the codebooks
if count == 1:
os = StringIO()
os.write('\x05vorbis')
os.write(packet.data)
page = OggPage()
page.packets.append(os.getvalue())
pages.append(page)
else:
page = OggPage()
page.position = packet.granule_pos
page.packets.append(packet.data)
pages.append(page)

pages[0].first = True
pages[-1].last = True
for n in range(len(pages)):
p = pages[n]
p.serial = 0x31337
p.sequence = n

out = open(r'c:\anders.ogg', 'wb')
for page in pages:
out.write(page.write())
out.close()

if __name__ == '__main__':
main()

• figure out vorb chunk - should be easier because I've got a few tricks for that now


• figure out where the blocksize is stored (4 + 4 bits)

Originally by: Emma Royd

---

Or you would wait till dominion, the music is stored as mp3 then with customizable jukebox so you can play your own music etc

---