diff --git a/index.js b/index.js
index 44d1151..cbb197f 100644
--- a/index.js
+++ b/index.js
@@ -311,8 +311,21 @@ function tinf_inflate_uncompressed_block(d) {
   return TINF_OK;
 }
 
+/* read an integer from a byte array in little-endian order */
+function tinf_readle(source, start, len) {
+  var res = 0;
+  for (var i = 0; i < len; ++i) {
+    var dat = source[start + i];
+    /* verify in bounds */
+    if (typeof dat === 'undefined')
+      throw new Error('out of bounds');
+    res += dat << (8 * i);
+  }
+  return res;
+}
+
 /* inflate stream from source to dest */
-function tinf_uncompress(source, dest) {
+function tinf_inflate_base(source, dest) {
   var d = new Data(source, dest);
   var bfinal, btype, res;
 
@@ -357,6 +370,100 @@ function tinf_uncompress(source, dest) {
   return d.dest;
 }
 
+/**
+ * Decompresses deflate data. Similar to `pako.inflateRaw()`
+ * @param {Uint8Array} source The deflate data
+ * @param {Uint8Array} [dest] Where to copy the uncompressed data to. If none
+ *                            is provided, a new Uint8Array will be returned.
+ *                            If the decompressed size is known, passing in an
+ *                            empty Uint8Array of that size reduces memory
+ *                            usage.
+ * @returns {Uint8Array} The original data
+ */
+function tinf_inflate(source, dest) {
+  if (dest)
+    return tinf_inflate_base(source, dest);
+  return new Uint8Array(tinf_inflate_base(source, []));
+}
+
+/**
+ * Decompresses gzip data. Similar to `pako.ungzip()`
+ * @param {Uint8Array} source The gzip data
+ * @param {Uint8Array} [dest] Where to copy the uncompressed data to. If none
+ *                            is provided, a new Uint8Array will be returned.
+ *                            If the decompressed size is known, passing in an
+ *                            empty Uint8Array of that size reduces memory
+ *                            usage.
+ * @returns {Uint8Array} The original data
+ */
+function tinf_gunzip(source, dest) {
+  var len = source.length;
+  if (len < 18 || source[0] !== 31 || source[1] !== 139 || source[2] !== 8)
+    throw new Error('invalid gzip data');
+  var flg = source[3];
+  var start = 10;
+  if (flg & 4) {
+    try { start += tinf_readle(source, start, 2) + 2; }
+    catch(e) { throw new Error('invalid gzip data'); }
+  }
+  /* skip FNAME, FCOMMENT (0 terminated) */
+  for (var zs = (flg >> 3 & 1) + (flg >> 4 & 1); zs > 0; zs -= (source[start++] === 0)) {}
+  /* skip 2 bytes if FHCRC */
+  start += flg & 2;
+  if (!dest) {
+    /* use header-provided size */
+    dest = new Uint8Array(tinf_readle(source, len - 4, 4));
+  }
+  return tinf_inflate_base(source.subarray(start, len - 8), dest);
+}
+
+/**
+ * Decompresses zlib data. Similar to `pako.inflate()`
+ * @param {Uint8Array} source The zlib data
+ * @param {Uint8Array} [dest] Where to copy the uncompressed data to. If none
+ *                            is provided, a new Uint8Array will be returned.
+ *                            If the decompressed size is known, passing in an
+ *                            empty Uint8Array of that size reduces memory
+ *                            usage.
+ * @returns {Uint8Array} The original data
+ */
+function tinf_decompress(source, dest) {
+  var len = source.length;
+  if (len < 6 || source[0] & 15 !== 8 || source[0] >> 4 > 7)
+    throw new Error('invalid zlib data');
+  if (source[1] & 32)
+    throw new Error('invalid zlib data: dictionaries not supported');
+  return tinf_inflate(source.subarray(2, -4), dest);
+}
+
+/**
+ * Decompresses deflate/gzip/zlib data. If format autodetection fails, try
+ * `inflate.inflate()`, `inflate.gunzip()`, and `inflate.decompress()`.
+ * @param {Uint8Array} source The deflate/gzip/zlib data
+ * @param {Uint8Array} [dest] Where to copy the uncompressed data to. If none
+ *                            is provided, a new Uint8Array will be returned.
+ *                            If the decompressed size is known, passing in an
+ *                            empty Uint8Array of that size reduces memory
+ *                            usage.
+ * @returns {Uint8Array} The original data
+ */
+function tinf_uncompress(source, dest) {
+  if (source[0] === 31 && source[1] === 139) {
+    /* data is gzipped */
+    return tinf_gunzip(source, dest);
+  }
+  if (source[0] & 15 !== 8 || source[0] >> 4 > 7) {
+    /* data cannot be zlib, assume deflate */
+    return tinf_inflate(source, dest);
+  }
+  /* data should be zlib - in rare cases can still be deflate */
+  return tinf_decompress(source, dest);
+}
+
+tinf_uncompress.inflate = tinf_inflate;
+tinf_uncompress.gunzip = tinf_gunzip;
+tinf_uncompress.decompress = tinf_decompress;
+
 /* -------------------- *
  * -- initialization -- *
  * -------------------- */
diff --git a/readme.md b/readme.md
index dd8c408..3f30d7e 100644
--- a/readme.md
+++ b/readme.md
@@ -1,7 +1,7 @@
 # tiny-inflate
 
 This is a port of Joergen Ibsen's [tiny inflate](https://bitbucket.org/jibsen/tinf) to JavaScript.
-Minified it is about 3KB, or 1.3KB gzipped. While being very small, it is also reasonably fast
+Minified it is about 3.7KB, or 1.5KB gzipped. While being very small, it is also reasonably fast
 (about 30% - 50% slower than [pako](https://github.com/nodeca/pako) on average), and should be 
 good enough for many applications. If you need the absolute best performance, however, you'll
 need to use a larger library such as pako that contains additional optimizations.
@@ -12,18 +12,47 @@ need to use a larger library such as pako that contains additional optimizations
 
 ## Example
 
-To use tiny-inflate, you need two things: a buffer of data compressed with deflate,
-and the decompressed size (often stored in a file header) to allocate your output buffer.
-Input and output buffers can be either node `Buffer`s, or `Uint8Array`s.
+To use tiny-inflate, you only need a buffer of data compressed with `deflate`, `zlib`, or `gzip`.
 
+If you have the decompressed size, you can allocate your output buffer ahead of time to save memory.
+(Note that this is unecessary for GZIP data because `tiny-inflate` automatically detects the output
+size from the header.)
+
+Input and output buffers must be `Uint8Array`s. Since `Buffer`s are instances of `Uint8Array`s, you can 
+also pass those in directly.
+
+`tiny-inflate` will try to automatically detect what compression method the data is in, but in rare cases 
+it fails. If that happens, you can use `inflate.gunzip()` for GZIP data (like `pako.ungzip()`), 
+`inflate.inflate()` for deflated data (like `pako.inflateRaw()`), and `inflate.decompress()` for Zlib data 
+(like `pako.inflate()`).
+
+
+Example: decoding a Base64, GZIP string to the source string
+```javascript
+var inflate = require('tiny-inflate');
+
+var compressedBuffer = Buffer.from('H4sIAAAAAAAAA/NIzcnJVyjPL8pJUQQAlRmFGwwAAAA=', 'base64');
+
+var outputUint8Array = inflate(compressedBuffer); /* Can also use inflate.gunzip(compressedBuffer) */
+
+var outputBuffer = Buffer.from(outputUint8Array);
+
+console.log(outputBuffer.toString('utf8')); /* Hello world! */
+```
+Example: efficiently decoding a Base64, Zlib string with known uncompressed length
 ```javascript
 var inflate = require('tiny-inflate');
 
-var compressedBuffer = new Bufer([ ... ]);
-var decompressedSize = ...;
-var outputBuffer = new Buffer(decompressedSize);
+var compressedBuffer = Buffer.from('eJzzSM3JyVcozy/KSVEEAB0JBF4=', 'base64');
+
+var outputSize = 12; /* Assuming you know this previously... */
+
+var outputArray = new Uint8Array(outputSize); /* Then you can create an efficient output space */
+
+/* Output array that is passed in is mutated - no need to extract return value */ 
+inflate(compressedBuffer, outputArray); /* Can also use inflate.inflate(compressedBuffer, outputArray) */
 
-inflate(compressedBuffer, outputBuffer);
+console.log(Buffer.from(outputArray).toString('utf8')); /* Hello world! */
 ```
 
 ## License
diff --git a/test/index.js b/test/index.js
index f4e8c88..29ead89 100644
--- a/test/index.js
+++ b/test/index.js
@@ -5,7 +5,7 @@ var assert = require('assert');
 var uncompressed = fs.readFileSync(__dirname + '/lorem.txt');
 
 describe('tiny-inflate', function() {
-  var compressed, noCompression, fixed;
+  var compressed, noCompression, fixed, zlibCompressed, gzipped;
   
   function deflate(buf, options, fn) {    
     var chunks = [];
@@ -40,6 +40,20 @@ describe('tiny-inflate', function() {
       done();
     });
   });
+
+  before(function(done) {
+    zlib.deflate(uncompressed, function(err, data) {
+      zlibCompressed = data;
+      done();
+    });
+  });
+
+  before(function(done) {
+    zlib.gzip(uncompressed, function(err, data) {
+      gzipped = data;
+      done();
+    });
+  });
   
   it('should inflate some data', function() {
     var out = Buffer.alloc(uncompressed.length);
@@ -72,4 +86,30 @@ describe('tiny-inflate', function() {
     inflate(input, out);
     assert.deepEqual(out, new Uint8Array(uncompressed));
   });
+
+  it('should handle no output array', function() {
+    var out = inflate(compressed);
+    assert.deepEqual(out, new Uint8Array(uncompressed));
+  })
+
+  it('should handle gzip', function() {
+    var out = Buffer.alloc(uncompressed.length);
+    inflate(gzipped, out);
+    assert.deepEqual(out, uncompressed);
+  });
+
+  it('should handle zlib', function() {
+    var out = Buffer.alloc(uncompressed.length);
+    inflate(zlibCompressed, out);
+    assert.deepEqual(out, uncompressed);
+  })
+
+  it('should autodetect format', function() {
+    var outGzip = inflate(gzipped);
+    assert.deepEqual(outGzip, uncompressed);
+    var outZlib = inflate(zlibCompressed);
+    assert.deepEqual(outZlib, uncompressed);
+    var outDeflate = inflate(compressed);
+    assert.deepEqual(outDeflate, uncompressed);
+  })
 });