The format can contain up to8 bits per pixel, allowing a single image to reference its ownpalette of up to 256 different colors chosen from the24-bitRGB color space. It can also represent multiple images in a file, which can be used foranimations, and allows a separate palette of up to 256 colors for each frame. These palette limitations make GIF less suitable for reproducing color photographs and otherimages with color gradients but well-suited for simpler images such as graphics or logos with solid areas of color.
GIF images are compressed using theLempel–Ziv–Welch (LZW)lossless data compression technique to reduce the file size without degrading the visual quality. While once in widespread usage on theWorld Wide Web because of its wide implementation and portability between applications and operating systems, usage of the format has declined for space and quality reasons, often being replaced with newer formats such asPNG for static images andMP4 for videos. In this context, short video clips are sometimes termed "GIFs" despite having no relation to the original file format.[3]
"Under construction" animated GIFs were a common feature of unfinished websites in the late 90s and early noughts.Animated GIFs like this one were once a common decorative feature of personal websites in the late 90s and early 2000s.
CompuServe introduced GIF on 15 June 1987 to provide a color image format for their file downloading areas. This replaced their earlierrun-length encoding format, which was black and white only. GIF became popular because it usedLempel–Ziv–Welchdata compression. Since this was more efficient than the run-length encoding used byPCX andMacPaint, fairly large images could be downloaded reasonably quickly even with slowmodems.
The original version of GIF was called 87a.[1] This version already supported multiple images in a stream.
In 1989, CompuServe released an enhanced version, called 89a,[2] This version added:
allowing text labels as text (not embedding them in the graphical data). However, this feature is rarely used. Modern browsers do not support it, and there is little control over fonts and styling.
The two versions can be distinguished by looking at the first sixbytes of the file (the "magic number" or signature), which, when interpreted asASCII, read "GIF87a" or "GIF89a", respectively.
CompuServe encouraged the adoption of GIF by providing downloadable conversion utilities for many computers. By December 1987, for example, anApple IIGS user could view pictures created on anAtari ST orCommodore 64.[4] GIF was one of the first two image formats commonly used on Web sites, the other being the black-and-whiteXBM.[5]
The feature of storing multiple images in one file, accompanied by control data, is used extensively on the Web to produce simpleanimations.
The optionalinterlacing feature, which stores image scan lines out of order in such a fashion that even a partially downloaded image was somewhat recognizable, also helped GIF's popularity,[6] as a user could abort the download if it was not what was required.
As anoun, the wordGIF is found in the newer editions of many dictionaries. In 2012, the American wing of theOxford University Press recognizedGIF as averb as well, meaning "to create a GIF file", as in "GIFing was the perfect medium for sharing scenes from theSummer Olympics". The press's lexicographers voted it theirword of the year, saying that GIFs have evolved into "a tool with serious applications including research and journalism".[12][13]
A humorous infographic announcing the 2013 launch of aTumblr account for theWhite House suggests pronouncingGIF with a hardg.
The pronunciation of the first letter ofGIF has been disputed since the 1990s. The most common pronunciations in English are/dʒɪf/ⓘ (with asoftg as ingin) and/ɡɪf/ⓘ (with ahardg as ingift), differing in thephoneme represented by the letterG. The creators of the format pronounced the acronymGIF as/dʒɪf/, with asoftg, with Wilhite stating that he intended for the pronunciation to deliberately echo the Americanpeanut butter brandJif, andCompuServe employees would often quip "choosy developers choose GIF", a spoof of Jif's television commercials.[14] However, the word is widely pronounced as/ɡɪf/, with ahardg,[15] and polls have generally shown that this hardg pronunciation is more prevalent.[16][17]
The disagreement over the pronunciation has led to heated Internet debate. On the occasion of receiving a lifetime achievement award at the2013 Webby Awards ceremony, Wilhite publicly rejected the hard-g pronunciation;[15][27][28] his speech led to more than 17,000 posts onTwitter and dozens of news articles.[29] TheWhite House[15] and the TV programJeopardy! also entered the debate in 2013.[28] In February 2020,The J.M. Smucker Company, the owners of the Jif brand, partnered with the animated image database and search engineGiphy to release a limited-edition "Jif vs. GIF" (hashtagged as #JIFvsGIF) jar of peanut butter that had a label humorously declaring the soft-g pronunciation to refer exclusively to the peanut butter, andGIF to be exclusively pronounced with the hard-g pronunciation.[30]
GIFs are suitable for sharp-edged line art with a limited number of colors, such as logos. This takes advantage of the format's lossless compression, which favors flat areas of uniform color with well defined edges.[31] They can also be used to store low-colorsprite data for games.[32] GIFs can be used for small animations and low-resolution video clips, or as reactions in online messaging used to convey emotion and feelings instead of using words. They are popular on social media platforms such asTumblr,[33]Facebook andTwitter.[34]
Conceptually, a GIF file describes a fixed-sized graphical area (the "logical screen") populated with zero or more "images". Many GIF files have a single image that fills the entire logical screen. Others divide the logical screen into separate sub-images. The images may also function as animation frames in an animated GIF file, but again these need not fill the entire logical screen.
GIF files start with a fixed-length header ("GIF87a" or "GIF89a") giving the version, followed by a fixed-length Logical Screen Descriptor giving the pixel dimensions and other characteristics of the logical screen. The screen descriptor may also specify the presence and size of a Global Color Table (GCT), which follows next if present.
Thereafter, the file is divided into segments of the following types, each introduced by a 1-byte sentinel:
An image (introduced by 0x2C, an ASCII comma',')
An extension block (introduced by 0x21, an ASCII exclamation point'!')
The trailer (a single byte of value 0x3B, an ASCII semicolon';'), which should be the last byte of the file.
An image starts with a fixed-length Image Descriptor, which may specify the presence and size of a Local Color Table (which follows next if present). The image data follows: one byte giving the bit width of the unencoded symbols (which must be at least 2 bits wide, even for bi-color images), followed by a series of sub-blocks containing the LZW-encoded data.
Extension blocks (blocks that "extend" the 87a definition via a mechanism already defined in the 87a spec) consist of the sentinel, an additional byte specifying the type of extension, and a series of sub-blocks with the extension data. Extension blocks that modify an image (like the Graphic Control Extension that specifies the optional animation delay time and optional transparent background color) must immediately precede the segment with the image they refer to.
Each sub-block begins with a byte giving the number of subsequent data bytes in the sub-block (1 to 255). The series of sub-blocks is terminated by an empty sub-block (a 0 byte).
This structure allows the file to be parsed even if not all parts are understood. A GIF marked 87a may contain extension blocks; the intent is that a decoder can read and display the file without the features covered in extensions it does not understand.
The full detail of the file format is covered in the GIF specification.[2]
GIF is a palette-based image format: each frame contains up to 256 colors chosen from the full 24-bit RGB color space. These colors are defined in a table (palette), and each pixel refers to an index in this palette. Originally, this was appropriate for hardware with limited color support; today, it makes GIF ideal for simple graphics, line drawings, logos, and basic animations. To approximate more colors, dithering techniques are sometimes used, but these can reduce image clarity or increase file size.
A GIF file may have a global color table, and each frame may also have a local color table. To conserve space, the specification allows color tables of 2n colors for anyn from 1 through 8. Most graphics applications will read and display GIF images with any of these table sizes; but many do not support all sizes whencreating images. Tables of 2, 16, and 256 colors are widely supported.
GIF supports transparency, allowing one color to be marked as transparent so that backgrounds or layered effects can show through.
Although GIF is almost never used fortrue color images, it is possible to do so.[35][36] A GIF image can include multiple image blocks, each of which can have its own 256-color palette, and the blocks can be tiled to create a complete image. Alternatively, the GIF89a specification introduced the idea of a "transparent" color where each image block can include its own palette of 255 visible colors plus one transparent color. A complete image can be created by layering image blocks with the visible portion of each layer showing through the transparent portions of the layers above.
An animated GIF illustrating a technique for displaying more than the typical limit of 256 colors
To render a full-color image as a GIF, the original image must be broken down into smaller regions having no more than 255 or 256 different colors. Each of these regions is then stored as a separate image block with its own local palette and when the image blocks are displayed together (either by tiling or by layering partially transparent image blocks), the complete, full-color image appears. For example, breaking an image into tiles of 16 by 16 pixels (256 pixels in total) ensures that no tile has more than the local palette limit of 256 colors, although larger tiles may be used and similar colors merged resulting in some loss of color information.[35]
Since each image block can have its own local color table, a GIF file having many image blocks can be very large, limiting the usefulness of full-color GIFs.[36] Additionally, not all GIF rendering programs handle tiled or layered images correctly. Many rendering programs interpret tiles or layers as animation frames and display them in sequence as an animation[35] with most web browsers automatically displaying the frames with a delay time of 0.1 seconds or more.[37][38][better source needed]
Microsoft Paint saves a small black-and-white image as the following GIF file (illustrated enlarged). Paint does not make optimal use of GIF; due to the unnecessarily large color table (storing a full 256 colors instead of the used 2) and symbol width, this GIF file is not an efficient representation of the 15-pixel image. Although the Graphic Control Extension block declares color index 16 (hexadecimal 10) to be transparent, that index is not used in the image. The only color indexes appearing in the image data are decimal 40 and 255, which the Global Color Table maps to black and white, respectively.
Sample image (enlarged), actual size 3 pixels wide by 5 high
The hex numbers in the following tables are inlittle-endian byte order, as the format specification prescribes.
Table of example GIF image values
Byte # (hex)
Hexadecimal
Text or value
Meaning
0
47 49 46 38 39 61
GIF89a
Header
Logical Screen Descriptor
6
03 00
3
Logical screen width
8
05 00
5
Logical screen height
A
F7
GCT follows for 256 colors with resolution 3 × 8 bits/primary, the lowest 3 bits represent the bit depth minus 1, the highest true bit means that the GCT is present
B
00
0
Background color: index #0; #000000 black
C
00
0
Default pixel aspect ratio, 0:0
Global Color Table
D
00 00 00
R (red)
G (green)
B (blue)
0
0
0
Global Color Table, color #0: #000000, black
Bytes Dh to 30Ch in the example define a palette of 256 colors. The indexes used in the sample image for black and white are 28h and FFh.
10
80 00 00
R (red)
G (green)
B (blue)
128
0
0
Global Color Table, color #1: transparent bit, not used in image
...
...
...
Global Color Table extends to 30A
30A
FF FF FF
R (red)
G (green)
B (blue)
255
255
255
Global Color Table, color #255: #ffffff, white
Graphic Control Extension
30D
21
'!'
An Extension Block (introduced by an ASCII exclamation point'!')
30E
F9
A Graphic Control Extension
30F
04
4
Amount of GCE data, 4 bytes
310
01
Transparent background color; this is a bit field, the lowest bit signifies transparency
311
00 00
Delay for animation in hundredths of a second;not used
313
10
16
Color number of transparent pixel in GCT
314
00
End of GCE block
Image Descriptor
315
2C
','
An Image Descriptor (introduced by 0x2C, an ASCII comma',')
316
00 00 00 00
(0, 0)
North-west corner position of image in logical screen
31A
03 00 05 00
(3, 5)
Image width and height in pixels
31E
00
0
Local color table bit, 0 means none
Image Data
31F
08
8
Start of image, LZW minimum code size
320
0B
11
Beginning of first data sub-block, specifying 11 bytes of encoded data to follow
321
00 51 FC 1B 28 70 A0 C1 83 01 01
<image data>
11 bytes of image data, see field 320
32C
00
0
Ending data sub-block, specifying no following data bytes (and the end of the image)
The image pixel data, scanned horizontally from top left, are converted byLZW encoding to codes that are then mapped into bytes for storing in the file. The pixel codes typically don't match the 8-bit size of the bytes, so the codes are packed into bytes by a "little-Endian" scheme: the least significant bit of the first code is stored in the least significant bit of the first byte, higher order bits of the code into higher order bits of the byte, spilling over into the low order bits of the next byte as necessary. Each subsequent code is stored starting at the least significant bit not already used.
This byte stream is stored in the file as a series of "sub-blocks". Each sub-block has a maximum length 255 bytes and is prefixed with a byte indicating the number of data bytes in the sub-block. The series of sub-blocks is terminated by an empty sub-block (a single 0 byte, indicating a sub-block with 0 data bytes).
For the sample image above the reversible mapping between 9-bit codes and bytes is shown below.
Reversible mapping
9-bit code
Byte
Hexadecimal
Binary
Binary
Hexadecimal
100
1 00000000
00000000
00
028
00 0101000
01010001
51
0FF
011 111111
11111100
FC
103
1000 00011
00011011
1B
102
10000 0010
00101000
28
103
100000 011
01110000
70
106
1000001 10
10100000
A0
107
10000011 1
11000001
C1
10000011
83
101
1 00000001
00000001
01
00000001
01
A slight compression is evident: pixel colors defined initially by 15 bytes are exactly represented by 12 code bytes including control codes.The encoding process that produces the 9-bit codes is shown below. A local string accumulates pixel color numbers from the palette, with no output action as long as the local string can be found in a code table. There is special treatment of the first two pixels that arrive before the table grows from its initial size by additions of strings. After each output code, the local string is initialized to the latest pixel color (that could not be included in the output code).
Table 9-bitstring --> code code Action #0 | 000h Initialize root table of 9-bit codes palette | : colors | : #255 | 0FFh clr | 100h end | 101h | 100h ClearPixel Local |color Palette string |BLACK #40 28 | 028h 1st pixel always to outputWHITE #255 FF | String found in table 28 FF | 102h Always add 1st string to table FF | Initialize local stringWHITE #255 FF FF | String not found in table | 0FFh - output code for previous string FF FF | 103h - add latest string to table FF | - initialize local stringWHITE #255 FF FF | String found in tableBLACK #40 FF FF 28 | String not found in table | 103h - output code for previous string FF FF 28 | 104h - add latest string to table 28 | - initialize local stringWHITE #255 28 FF | String found in tableWHITE #255 28 FF FF | String not found in table | 102h - output code for previous string 28 FF FF | 105h - add latest string to table FF | - initialize local stringWHITE #255 FF FF | String found in tableWHITE #255 FF FF FF | String not found in table | 103h - output code for previous string FF FF FF | 106h - add latest string to table FF | - initialize local stringWHITE #255 FF FF | String found in tableWHITE #255 FF FF FF | String found in tableWHITE #255 FF FF FF FF | String not found in table | 106h - output code for previous string FF FF FF FF| 107h - add latest string to table FF | - initialize local stringWHITE #255 FF FF | String found in tableWHITE #255 FF FF FF | String found in tableWHITE #255 FF FF FF FF | String found in table No more pixels 107h - output code for last string 101h End
For clarity the table is shown above as being built of strings of increasing length. That scheme can function but the table consumes an unpredictable amount of memory. Memory can be saved in practice by noting that each new string to be stored consists of a previously stored string augmented by one character. It is economical to store at each address only two words: an existing address and one character.
The LZW algorithm requires a search of the table for each pixel. A linear search through up to 4096 addresses would make the coding slow. In practice the codes can be stored in order of numerical value; this allows each search to be done by a SAR (Successive Approximation Register, as used in someADCs), with only 12 magnitude comparisons. For this efficiency an extra table is needed to convert between codes and actual memory addresses; the extra table upkeeping is needed only when a new code is stored which happens at much less than pixel rate.
Decoding begins by mapping the stored bytes back to 9-bit codes. These are decoded to recover the pixel colors as shown below. A table identical to the one used in the encoder is built by adding strings by this rule:
Is incoming code found in table?
Yes
add string for local code followed by first byte of string for incoming code
No
add string for local code followed by copy of its own first byte
shift9-bit ----> Local Table Pixelcode code code --> string Palette color Action100h 000h | #0 Initialize root table of 9-bit codes : | palette : | colors 0FFh | #255 100h | clr 101h | end028h | #40BLACK Decode 1st pixel0FFh 028h | Incoming code found in table | #255WHITE - output string from table 102h | 28 FF - add to table103h 0FFh | Incoming code not found in table 103h | FF FF - add to table | - output string from table | #255WHITE | #255WHITE102h 103h | Incoming code found in table | - output string from table | #40BLACK | #255WHITE 104h | FF FF 28 - add to table103h 102h | Incoming code found in table | - output string from table | #255WHITE | #255WHITE 105h | 28 FF FF - add to table106h 103h | Incoming code not found in table 106h | FF FF FF - add to table | - output string from table | #255WHITE | #255WHITE | #255WHITE107h 106h | Incoming code not found in table 107h | FF FF FF FF - add to table | - output string from table | #255WHITE | #255WHITE | #255WHITE | #255WHITE101h | End
Shorter code lengths can be used for palettes smaller than the 256 colors in the example. If the palette is only 64 colors (so color indexes are 6 bits wide), the symbols can range from 0 to 63, and the symbol width can be taken to be 6 bits, with codes starting at 7 bits. In fact, the symbol width need not match the palette size: as long as the values decoded are always less than the number of colors in the palette, the symbols can be any width from 2 to 8, and the palette size any power of 2 from 2 to 256. For example, if only the first four colors (values 0 to 3) of the palette are used, the symbols can be taken to be 2 bits wide with codes starting at 3 bits.
Conversely, the symbol width could be set at 8, even if only values 0 and 1 are used; these data would only require a two-color table. Although there would be no point in encoding the file that way, something similar typically happens for bi-color images: the minimum symbol width is 2, even if only values 0 and 1 are used.
The code table initially contains codes that are one bit longer than the symbol size in order to accommodate the two special codesclr andend and codes for strings that are added during the process. When the table is full the code length increases to give space for more strings, up to a maximum code 4095 = FFF(hex). As the decoder builds its table it tracks these increases in code length and it is able to unpack incoming bytes accordingly.
A 46×46 uncompressed GIF with 7-bit symbols (128 colors, 8-bit codes). Click on the image for an explanation of the code.
The GIF encoding process can be modified to create a file without LZW compression that is still viewable as a GIF image. This technique was introduced originally as a way to avoid patent infringement. Uncompressed GIF can also be a useful intermediate format for a graphics programmer because individual pixels are accessible for reading or painting. An uncompressed GIF file can be converted to an ordinary GIF file simply by passing it through an image editor.
The modified encoding method ignores building the LZW table and emits only the root palette codes and the codes for CLEAR and STOP. This yields a simpler encoding (a 1-to-1 correspondence between code values and palette codes) but sacrifices all of the compression: each pixel in the image generates an output code indicating its color index. When processing an uncompressed GIF, a standard GIF decoder will not be prevented from writing strings to its dictionary table, but the code width must never increase since that triggers a different packing of bits to bytes.
If the symbol width isn, the codes of widthn+1 fall naturally into two blocks: the lower block of2n codes for coding single symbols, and the upper block of2n codes that will be used by the decoder for sequences of length greater than one. Of that upper block, the first two codes are already taken:2n for CLEAR and2n + 1 for STOP. The decoder must also be prevented from using the last code in the upper block,2n+1 − 1, because when the decoder fills that slot, it will increase the code width. Thus in the upper block there are2n − 3 codes available to the decoder that won't trigger an increase in code width. Because the decoder is always one step behind in maintaining the table, it does not generate a table entry upon receiving the first code from the encoder, but will generate one for each succeeding code. Thus the encoder can generate2n − 2 codes without triggering an increase in code width. Therefore, the encoder must emit extra CLEAR codes at intervals of2n − 2 codes or less to make the decoder reset the coding dictionary. The GIF standard allows such extra CLEAR codes to be inserted in the image data at any time. The composite data stream is partitioned into sub-blocks that each carry from 1 to 255 bytes.
For the sample 3×5 image above, the following 9-bit codes represent "clear" (100) followed by image pixels in scan order and "stop" (101).
The trivial example of a large image of solid color demonstrates the variable-length LZW compression used in GIF files.
Sample compression of a GIF file
Code
Pixels
Notes
No.Ni
ValueNi + 256
Length(bits)
This codeNi
AccumulatedNi(Ni + 1)/2
Relations using Ni only apply to same-color pixels until coding table is full.
0
100h
9
Clear code table
1
FFh
1
1
Top left pixel color chosen as the highest index of a 256-color palette
2
102h
2
3
3⋮255
103h⋮1FFh
3⋮255
6⋮32640
Last 9-bit code
256⋮767
200h⋮3FFh
10
256⋮767
32896⋮294528
Last 10-bit code
768⋮1791
400h⋮7FFh
11
768⋮1791
295296⋮1604736
Last 11-bit code
1792⋮3839
800h⋮FFFh
12
1792⋮3839
1606528⋮7370880
Code table full
⋮
FFFh
3839
The maximum code may repeat for more same-color pixels.Overall data compression asymptotically approaches 3839 ×8/12 =2559+1/3
101h
End of image data
The code values shown are packed into bytes which are then packed into blocks of up to 255 bytes. A block of image data begins with a byte that declares the number of bytes to follow. The last block of data for an image is marked by a zero block-length byte.
Screen capture of an interlaced GIF loading in a web browser
The GIF Specification allows each image within the logical screen of a GIF file to specify that it is interlaced; i.e., that the order of the raster lines in its data block is not sequential. This allows a partial display of the image that can be recognized before the full image is painted.
An interlaced image is divided from top to bottom into strips 8 pixels high, and the rows of the image are presented in the following order:
Pass 1: Line 0 (the top-most line) from each strip.
Pass 2: Line 4 from each strip.
Pass 3: Lines 2 and 6 from each strip.
Pass 4: Lines 1, 3, 5, and 7 from each strip.
The pixels within each line are not interlaced, but presented consecutively from left to right. As with non-interlaced images, there is no break between the data for one line and the data for the next. The indicator that an image is interlaced is a bit set in the corresponding Image Descriptor block.
GIF can be used to display animation, as in this image ofNewton's cradle.
Although GIF was not designed as an animation medium, its ability to store multiple images in one file naturally suggested using the format to store theframes of an animation sequence. To facilitatedisplaying animations, the GIF89a spec added the Graphic Control Extension (GCE), which allows the images (frames) in the file to be painted with time delays, forming avideo clip. Each frame in an animation GIF is introduced by its own GCE specifying the time delay to wait after the frame is drawn. Global information at the start of the file applies by default to all frames. The data is stream-oriented, so the file offset of the start of each GCE depends on the length of preceding data. Within each frame the LZW-coded image data is arranged in sub-blocks of up to 255 bytes; the size of each sub-block is declared by the byte that precedes it.
By default, an animation displays the sequence of frames only once, stopping when the last frame is displayed. To enable an animation to loop,Netscape in the 1990s used the Application Extension block (intended to allow vendors to add application-specific information to the GIF file) to implement the Netscape Application Block (NAB).[39] This block, placed immediately before the sequence of animation frames, specifies the number of times the sequence of frames should be played (1 to 65535 times) or that it should repeat continuously (zero indicates loop forever). Support for these repeating animations first appeared inNetscape Navigator version 2.0, and then spread to other browsers.[40] Most browsers now recognize and support NAB, though it is not strictly part of the GIF89a specification.
The following example shows the structure of the animation fileRotating earth (large).gif shown (as a thumbnail) in the article's infobox.
Structure of GIF
Byte # (hex)
Hexadecimal
Text or value
Meaning
0
47 49 46 38 39 61
GIF89a
Logical Screen Descriptor
6
90 01
400
Width in pixels
8
90 01
400
Height in pixels
A
F7
GCT follows for 256 colors with resolution 3 × 8 bits/primary
B
00
0
Background color: #000000, black
C
00
0
Default pixel aspect ratio, 0:0
D
00
Global Color Table
⋮
30D
21
'!'
An Extension Block (introduced by an ASCII exclamation point'!')
30E
FF
Application Extension
30F
0B
11
Size of block including application name and verification bytes (always 11)
310
4E 45 54 53 43 41 50 45 32 2E 30
NETSCAPE2.0
8-byte application name plus 3 verification bytes
31B
03
3
Number of bytes in the following sub-block
31C
01
1
Index of the current data sub-block(always 1 for the NETSCAPE block)
End of the sub-block chain for the Application Extension block
320
21
'!'
An Extension Block (introduced by an ASCII exclamation point'!')
321
F9
Graphic Control Extension for frame #1
322
04
4
Number of bytes (4) in the current sub-block
323
04
000........001........0........0
(broken into sections for easier reading)
Reserved, 5 lower bits arebit field Disposal method 1: do not dispose No user input Transparent color, 0 means not given
324
09 00
9
Frame delay: 0.09 second delay before painting next frame
326
FF
Transparent color index(unused in this frame)
327
00
End of sub-block chain for Graphic Control Extension block
328
2C
','
An Image Descriptor (introduced by 0x2C, an ASCII comma',')
329
00 00 00 00
(0, 0)
North-west corner position of image in logical screen: (0, 0)
32D
90 01 90 01
(400, 400)
Frame width and height: 400 × 400 pixels
331
00
0
Local color table: 0 means none & no interlacing
332
08
8
Minimum LZW code size for Image Data of frame #1
333
FF
255
Number of bytes of LZW image data in the following sub-block: 255 bytes
334
...
<image data>
Image data, 255 bytes
433
FF
255
Number of bytes of LZW image data in the following sub-block, 255 bytes
434
...
<image data>
Image data, 255 bytes
⋮
Repeat for next blocks
92C0
00
End of sub-block chain for this frame
92C1
21
'!'
An Extension Block (introduced by an ASCII exclamation point'!')
92C2
F9
Graphic Control Extension for frame #2
⋮
Repeat for next frames
EDABD
21
'!'
An Extension Block (introduced by an ASCII exclamation point'!')
EDABE
F9
Graphic Control Extension for frame #44
⋮
Image information and data for frame #44
F48F5
3B
Trailer: Last byte in the file, signaling EOF
The animation delay for each frame is specified in the GCE in hundredths of a second. Some economy of data is possible where a frame need only rewrite a portion of the pixels of the display, because the Image Descriptor can define a smaller rectangle to be rescanned instead of the whole image. Browsers or other displays that do not support animated GIFs typically show only the first frame.
The size and color quality of animated GIF files can vary significantly depending on the application used to create them. Strategies for minimizing file size include using a common global color table for all frames (rather than a complete local color table for each frame) and minimizing the number of pixels covered in successive frames (so that only the pixels that change from one frame to the next are included in the latter frame). More advanced techniques involve modifying color sequences to better match the existing LZW dictionary, a form oflossy compression. Simply packing a series of independent frame images into a composite animation tends to yield large file sizes. Tools are available to minimize the file size given an existing GIF.
Metadata can be stored in GIF files as a comment block, a plain text block, or an application-specific application extension block. Several graphics editors use unofficial application extension blocks to include the data used to generate the image, so that it can be recovered for further editing.
All of these methods technically require the metadata to be broken into sub-blocks so that applications can navigate the metadata block without knowing its internal structure.
TheExtensible Metadata Platform (XMP) metadata standard introduced an unofficial but now widespread "XMP Data" application extension block for including XMP data in GIF files.[41] Since the XMP data is encoded usingUTF-8 without NUL characters, there are no 0 bytes in the data. Rather than break the data into formal sub-blocks, the extension block terminates with a "magic trailer" that routes any application treating the data as sub-blocks to a final 0 byte that terminates the sub-block chain.
Welch filed a patent application for the LZW method in June 1983. The resulting patent, US4558302,[44] granted in December 1985, was assigned toSperry Corporation who subsequently merged withBurroughs Corporation in 1986 and formedUnisys.[42] Further patents were obtained in the United Kingdom, France, Germany, Italy, Japan and Canada.
In addition to the above patents, Welch's 1983 patent also includes citations to several other patents that influenced it, including:
two 1980 Japanese patents fromNEC's Jun Kanatsu,[45][46]
a 1981 German patent from Karl Eckhart Heinz.[47][48]
In June 1984, an article by Welch was published in theIEEE magazine which publicly described the LZW technique for the first time.[49] LZW became a popular data compression technique and, when the patent was granted, Unisys entered into licensing agreements with over a hundred companies.[42][50]
The popularity of LZW ledCompuServe to choose it as the compression technique for their version of GIF, developed in 1987. At the time, CompuServe was not aware of the patent.[42] Unisys became aware that the version of GIF used the LZW compression technique and entered into licensing negotiations with CompuServe in January 1993. The subsequent agreement was announced on 24 December 1994.[43] Unisys stated that they expected all major commercial on-line information services companies employing the LZW patent to license the technology from Unisys at a reasonable rate, but that they would not require licensing, or fees to be paid, for non-commercial, non-profit GIF-based applications, including those for use on the on-line services.[50]
Following this announcement, there was widespread condemnation of CompuServe and Unisys, and many software developers threatened to stop using GIF. ThePNG format (see below) was developed in 1995 as an intended replacement.[42][43][49] However, obtaining support from the makers of Web browsers and other software for the PNG format proved difficult and it was not possible to replace GIF, although PNG has gradually increased in popularity.[42] Therefore, GIF variations without LZW compression were developed. For instance the libungif library, based onEric S. Raymond's giflib, allows creation of GIFs that followed the data format but avoided the compression features, thus avoiding use of the Unisys LZW patent.[51] A 2001Dr. Dobb's article described a way to achieve LZW-compatible encoding for data that would compress well under a run-length encoding mechanism without infringing on its patents.[52]
In August 1999, Unisys changed the details of their licensing practice, announcing the option for owners of certain non-commercial and private websites to obtain licenses on payment of a one-time license fee of $5000 or $7500.[53] Such licenses were not required for website owners or other GIF users who had used licensed software to generate GIFs. Nevertheless, Unisys was subjected to thousands of online attacks and abusive emails from users believing that they were going to be charged $5000 or sued for using GIFs on their websites.[54] Despite giving free licenses to hundreds of non-profit organizations, schools and governments, Unisys was completely unable to generate any good publicity and continued to be condemned by individuals and organizations such as theLeague for Programming Freedom who started the "Burn All GIFs" campaign in 1999.[55][56]
The United States LZW patent expired on 20 June 2003.[57] The counterpart patents in the United Kingdom, France, Germany and Italy expired on 18 June 2004, the Japanese patents expired on 20 June 2004, and the Canadian patent expired on 7 July 2004.[57] Consequently, while Unisys has further patents and patent applications relating to improvements to the LZW technique,[57] LZW itself (and consequently GIF) have been free to use since July 2004.[58]
This section needs to beupdated. The reason given is: PNG support/use, modern use and alternatives. Please help update this article to reflect recent events or newly available information.(August 2025)
Portable Network Graphics (PNG) was designed as a replacement for GIF in order to avoid infringement of Unisys' patent on the LZW compression technique.[42] PNG offers better compression and more features than GIF,[59] animation being the only significant exception. PNG is more suitable than GIF in instances where true-color imaging andalpha transparency are required.
Although support for PNG format came slowly, newweb browsers support PNG. Older versions ofInternet Explorer do not support all features of PNG. Versions 6 and earlier do not supportalpha channel transparency without using Microsoft-specific HTML extensions.[60]Gamma correction of PNG images was not supported before version 8, and the display of these images in earlier versions may have the wrong tint.[61]
For identical 8-bit (or lower) image data, PNG files are typically smaller than the equivalent GIFs, due to the more efficient compression techniques used in PNG encoding.[62] Complete support for GIF is complicated chiefly by the complex canvas structure it allows, though this is what enables the compact animation features.
Videos resolve many issues that GIFs present through common usage on the web. They include drastically smallerfile sizes, the ability to surpass the8-bit color restriction, and better frame-handling and compression throughinter-frame coding. Virtually universal support for the GIF format inweb browsers and a lack of official support for video in theHTML standard caused GIF to rise to prominence for the purpose of displaying short video-like files on the web.
MNG ("Multiple-image Network Graphics") was originally developed as a PNG-based solution for animations. MNG reached version 1.0 in 2001, but few applications support it.
APNG ("Animated Portable Network Graphics") was proposed byMozilla in 2006. APNG is an extension to the PNG format as alternative to the MNG format. APNG is supported by most browsers as of 2019.[63] APNG provides the ability to animate PNG files, while retaining backwards compatibility in decoders that cannot understand the animation chunk (unlike MNG). Older decoders will simply render the first frame of the animation.
The PNG group officially rejected APNG as an official extension on 20 April 2007.[64]
There have been several subsequent proposals for a simple animated graphics format based on PNG using several different approaches.[65] Nevertheless, APNG is still under development by Mozilla and is supported inFirefox 3.0[66][67] while MNG support was dropped.[68][69] APNG is currently supported by all major web browsers including Chrome (since version 59.0), Opera, Firefox and Edge.
EmbeddedAdobe Flash objects andMPEG files were used on some websites to display simple video, but required the use of an additional browser plugin.
WebM andWebP are in development and are supported by some web browsers.[70]
Other options for web animation include serving individual frames usingAJAX, or animatingSVG ("Scalable vector graphics") images usingJavaScript orSMIL ("Synchronized Multimedia Integration Language").[71]
With the introduction of widespread support of theHTML video (<video>) tag in most web browsers, some websites use a looped version of the video tag generated. This gives the appearance of a GIF, but with the size and speed advantages of compressed video.
Notable examples areGfycat andImgur and their GIFV metaformat, which is really a video tag playing a loopedMP4 orWebM compressed video.[72]
HEIF ("High Efficiency Image File Format") is an image file format, finalized in 2015, which uses adiscrete cosine transform (DCT)lossy compression algorithm based on theHEVC video format, and related to theJPEG image format. In contrast to JPEG, HEIF supports animation.[73]
Compared to the GIF format, which lacks DCT compression, HEIF allows significantly more efficient compression. HEIF stores more information and produces higher-quality animated images at a small fraction of an equivalent GIF's size.[74]
In April 2014,4chan added support for silentWebM videos that are under 3 MB in size and 2 min in length,[76][77] and in October 2014,Imgur started converting any GIF files uploaded to the site toH.264 video and giving the link to the HTML player the appearance of an actual file with a.gifv extension.[78][79]
In January 2016,Telegram started re-encoding all GIFs toMPEG-4 videos that "require up to 95% less disk space for the same image quality."[80]
^"GIF".The American Heritage Abbreviations Dictionary, Third Edition. Houghton Mifflin Company. 2005.Archived from the original on 3 September 2011. Retrieved15 April 2007.
^"GIF".The Cambridge Dictionary of American English. Cambridge University Press.Archived from the original on 27 February 2014. Retrieved19 February 2014.
^S. Chin; D. Iverson; O. Campesato; P. Trani (2011).Pro Android Flash(PDF). New York: Apress. p. 350.ISBN9781430232315.Archived(PDF) from the original on 2 April 2015. Retrieved11 March 2015.
^DE patent 3118676, Eckhart, Heinz Karl, "Verfahren zur Kompression redundanter Folgen serieller Datenelemente [Method for compressing redundant sequences of serial data elements]", published 2 December 1982
.gif)
