Flipnotes animation files are normally stored on the SD card in the following folder:

/private/ds/app/4B475556/001 - European location

You can save your images in a user folder as long as it's on the same level as the 001 folder. Clicking on the Choose Folder icon on the top of the View Flipnote / SD Card option gives you a list of normal folders, and you can swap to user folders by selecting the normal button again. This could possibly be to overcome a limitation in one of the files (possibly the dirmemo2.lst file).

File Header

Start Length Description
0x0000 4 Magic (should be PARA)
0x0004 4 Size of animation data
0x0008 4 size of audio data
0x000C 2 # of frames
0x000E 2 Unknown - all files looked at so far has this field set to 24 00 hex.
0x0010 2 Lock - 0 open, 1 locked
0x0012 2 Preview frame number
0x0014 22 Original author name (UCS-2)
0x002A 22 Last Edited By Author name (UCS-2)
0x0040 22 User name (UCS-2)
0x0056 8 Original author ID
0x005E 8 Edit Author ID - the last user to save the file
0x0066 18 Original File-name - see notes on format
0x0078 18 File-name - see notes on format
0x008A 8 Previous Editing Author ID
0x0092 8 Partial File name? - see notes on format
0x009A 4 Date stored as the number of seconds since midnight 1 Jan 2000.
0x009E 2 Filler - 00 00
0x00A0 1536 Preview Bitmap (4 bits/pixel).

The file name seems to be stored in the file header 3 times:

  • Original file name 0x0066
  • Current file name 0x0078
  • Partial file name 0x0092

The file name is stored in 3 parts:

  • 3 hex bytes, the first byte's high nibble is not the same as the first byte of the file name
  • 13 bytes for the central part of the file name
  • 2 bytes for the last part of the file name.

For the file name : G35B20_0909841CDBEB1_002

  • First location: D3 5B 20 30 39 30 39 38 34 31 43 44 42 45 42 31 00 00
  • Second location: D3 5B 20 30 39 30 39 38 34 31 43 44 42 45 42 31 02 00
  • Last location: D3 5B 20 09 09 84 1C DB

The 3 sections of the file name seem to be stored as:

  • Last 6 digits of your Flipnote Studio ID
  • A random number, possibly generated from a date stamp
  • The version number of the file

Palette

Hex Color DS RGB Approximation
0x0 Not used / White 31, 31, 31
0x1 Dark Grey 10, 10, 10
0x2 White 31, 31, 31
0x3 Light Grey 20, 20, 20
0x4 Pure Red 31, 0, 0
0x5 Dark Red 15, 0, 0
0x6 Light Red / Pink 31, 15, 15
0x7 Pure Green 0, 31, 0
0x8 Pure Blue 0, 0, 31
0x9 Dark Blue 0, 0, 15
0xA Light Blue 15, 15, 31
0xB Pure Green 0, 31, 0
0xC Magenta / Purple 31, 0, 31
0xD Pure Green 0, 31, 0
0xE Pure Green 0, 31, 0
0xF Pure Green 0, 31, 0

The preview image is a 64x48 image stored in a 16 colour tile format, with each 8x8 tile taking up 32 bytes of the file.

Here's a photo taken of the altered preview image, reduced to about the correct size.

 

Link to the original picture, warning image is 2,048 × 1,536.

Here's two preview images that blasty decoded from some PPM files.

   

You can use pbsds's PPMtool to extract the preview image yourself

Animation Data Section

The animation section starts at offset 0x06A0, with a header section.

Animation Sequence Header

Start Length Description
0000 4 Size of the offset table
0004 4 Unknown
0008 4 * number of frames A list of offsets to the frames in the order to play them.

The offsets are all relative to the end of the offset list, and the frames are not stored in any specific format in the file, so you have to read this table for the next offset as I've seen the list point to a frame at 00023572 and the next frame was at 00000000.

Animation Frame

Start Length Description
0000 1 Pen and Paper information
0001 48 Layer 1 line encoding
0031 48 Layer 2 line encoding
0061 ? The frame data stored layer 1 then layer 2

The pen and paper byte at the start is encoded as follows:

 

The paper bit indicates what color the paper is - 0 for black and 1 for white.

The Frame encoding is 0 for a change between last frame and this one, and 1 for a totally new frame.

The Layer 1 pen and Layer 2 pen information is as follows:

Number Meaning
0 Not used
1 Invert Paper (black on white or white on black)
2 Red
3 Blue

Line encoding is compacted from the 192 lines on the screen down to 48 bytes per layer, each byte within the layer encoding section is stored as 2 bits per line, and 4 lines to the byte. The lines are stored in the same order as the bits in the byte, i.e. line 0 uses bit 0-1, line 1 uses bit 2-3, line 2 uses bit 4-5, line 3 uses bit 6-7. The value for the line encoding determins how many bytes you need to read at a minimum:

Number Meaning
0 Skip this line - there is no data for this line
1 Coded line
2 Inverted coded line
3 Raw line data

Before I explain how to read the line data you must understand that the way flipnote works is in layers, each layer is a bit map of on or off, with on being the pen colour, and off being the paper colour.

The coded and inverted coded lines are stored with a 4 byte header that is used to indicate how many extra bytes to read, and what part of the line they represent. Each byte after the first 4 is then read in a reverse order, i.e. pixel 0 of 8 is bit 0, etc. To understand this you first take the 256 pixels of the line, and divide it by 8 to give 32, this is the maximum number of bytes a line encoding will ever use, and it's also the number of bits in 4 bytes. The first 8 pixels of the line are used only when the bit 0x80000000 is set in the initial 4 bytes.

For example the line is encoded as type 1, and the bytes that's stored is 80 00 00 00 20.

The first 4 bytes is indicating that there is 1 byte effacted in this line, and that it's the first 8 pixels in the line. The next byte is read and once it's expanded the 6'th byte of the line is set to the pen colour. If on the other hand the line was encoded with type 2, that one pixel would be the paper colour and every other pixel would be the pen colour.

Here's some code to show how to decode the line:

u32 decodeLine( u8 *outB, u8 *inB, u32 useByte, u8 paper, u8 pen, u8 invflag )
{
	u32 offset = 0;
	u16 x = 0, i;
	u8 data;

	if( useByte == 0 ) // Set the full line to the current paper colour
		memset( outB, paper, 256 );
	else // Bytes in this line, read and deal with them
	{
		while( useByte != 0 )
		{
			if( ( useByte & 0x80000000 ) == 0x80000000 )
			{
				// This byte exists...
				data = inB[ offset++ ];

				for( i = 0; i < 8; i ++ )
				{
					if( ( data & 0x01 ) == 0x01 )
						outB[ x ] = invflag == 0 ? pen : paper;
					else
						outB[ x ] = invflag == 0 ? paper : pen;

					x ++;

					data >>= 1;
				}
			}
			else
			{
				memset( &outB[ x ], paper, 8 );
				x += 8;
			}

			useByte <<= 1;
		}

		if( x < 256 )
			memset( &outB[ x ], paper, 256 - x );
	}

	return offset;
}

The outB and inB are pointers to pre-allocated memory blocks, the inB is a pointer to the file, and the outB is the output data block. The useByte is the 4 bytes that is at the start of the line, if your reading a type 3 line (full raw data) just pass in 0xFFFFFFFF. The pen and paper values are what was read from the pen and paper information block.

Sound data section

The position of the sound header is 0x6A0 + AnimationDataSize

Length Function Description
Framecount Sound effect usage. Each byte represents a frame. bit #1(from Right to Left) represents Sound effect 1, bit #2 represents Sound effect 2 and bit #3 represents Sound effect 3. the last 5 bits is padding.
0 - 3 Padding 0x00 * (0 to 3). Makes the offset to the next value dividable by 4.
4 Size of the background music. The size of the background music in bytes in the audio data below. 0x0000000 if not used/emtpy.
4 Size of sound effect #1 The size of the sound effect #1 in bytes in the audio data below. 0x0000000 if not used/emtpy.
4 Size of sound effect #2 The size of the sound effect #3 in bytes in the audio data below. 0x0000000 if not used/emtpy.
4 Size of sound effect #3 The size of the sound effect #3 in bytes in the audio data below. 0x0000000 if not used/emtpy.
1 Given Framespeed. This is the speed of the flipnote. It's in the range of 1-8. To get the correct value, you must take 8, and subtract the decimal value of this byte.
1 Current Framespeed when BGM was recorded. This is the speed of the flipnote when the BGM was recorded/modified. It's in the range of 1-8. This is used to resample the sound to the new speed. To get the correct value, you must take 8, and subtract the decimal value of this byte.
14 Padding 0x00 * 14

Then comes the audio data.

  • First comes the BGM(if used)
  • Then comes sound effect #1(if used)
  • Then comes sound effect #2(if used)
  • Then comes sound effect #3(if used)

To decode the sound data, you must multiply each byte by 16 and run it as 32000Hz PCM data.


I have yet to figure out how to get rid of the noice when decoding sound...--pbsds 17:38, 6 June 2010 (UTC)