diff --git a/anim_encoder.py b/anim_encoder.py
index 0deae29..8255bbf 100755
--- a/anim_encoder.py
+++ b/anim_encoder.py
@@ -128,15 +128,23 @@ def find_matching_rect(bitmap, num_used_rows, packed, src, sx, sy, w, h):
         return None
 
 def generate_animation(anim_name):
-    frames = []
-    rex = re.compile("screen_([0-9]+).png")
-    for f in os.listdir(anim_name):
-        m = re.search(rex, f)
-        if m:
-            frames.append((int(m.group(1)), anim_name + "/" + f))
-    frames.sort()
-
-    images = [misc.imread(f) for t, f in frames]
+    if anim_name.endswith('.lcf'):
+        from lcf import process
+        frames = [(f.delta, f) for f in process(anim_name)]
+        images = [f.get_array() for t, f in frames]
+        delays = [t for t, f in frames] + [END_FRAME_PAUSE]
+        anim_name = anim_name[:-4]
+    else:
+        frames = []
+        rex = re.compile("screen_([0-9]+).png")
+        for f in os.listdir(anim_name):
+            m = re.search(rex, f)
+            if m:
+                frames.append((int(m.group(1)), anim_name + "/" + f))
+        frames.sort()
+        images = [misc.imread(f) for t, f in frames]
+        times = [t for t, f in frames]
+        delays = (array(times[1:] + [times[-1] + END_FRAME_PAUSE]) - array(times)).tolist()
 
     zero = images[0] - images[0]
     pairs = zip([zero] + images[:-1], images)
@@ -204,8 +212,6 @@ def generate_animation(anim_name):
         os.system("mv " + anim_name + "_packed_tmp.png " + anim_name + "_packed.png")
 
     # Generate JSON to represent the data
-    times = [t for t, f in frames]
-    delays = (array(times[1:] + [times[-1] + END_FRAME_PAUSE]) - array(times)).tolist()
 
     timeline = []
     for i in xrange(len(images)):
diff --git a/lcf.py b/lcf.py
new file mode 100644
index 0000000..524c14f
--- /dev/null
+++ b/lcf.py
@@ -0,0 +1,114 @@
+# -*- coding: utf-8 -*-
+from __future__ import division
+
+import zlib
+import struct
+from collections import namedtuple
+
+import numpy as np
+
+
+LCF_VERSION = 0x11CEb001
+
+LcfHeader = namedtuple('LcfHeader', 'version bpp w h bsize_w bsize_h nf cdata_left dsize')
+LcfBlock = namedtuple('LcfBlock', 'hdr frame_deltas frames')
+
+
+class LcfFrame(object):
+    def __init__(self, frame, block, block_idx, frame_idx):
+        self.frame = frame
+        self.block = block
+        self.block_idx = block_idx
+        self.frame_idx = frame_idx
+        self.delta = abs(block.frame_deltas[frame_idx])
+
+    def get_array(self):
+        hdr = self.block.hdr
+        arr = np.ndarray(shape=(0, hdr.w, 3), dtype=np.uint8)
+        for y, ypos in enumerate(xrange(0, hdr.h, hdr.bsize_h)):
+            hei = min(hdr.h - ypos, hdr.bsize_h)
+            line = np.ndarray(shape=(hei, 0, 3), dtype=np.uint8)
+            for x, xpos in enumerate(xrange(0, hdr.w, hdr.bsize_w)):
+                # wid = min(hdr.w - xpos, hdr.bsize_w)
+                line = np.concatenate((line, self.frame[x][y]), axis=1)
+            arr = np.concatenate((arr, line), axis=0)
+        return arr
+
+
+class BadBlock(Exception):
+    pass
+
+
+class CorruptedBlock(BadBlock):
+    pass
+
+
+def read(fd, size):
+    data = fd.read(size)
+    if len(data) != size:
+        raise BadBlock("Cannot read %s bytes from file" % size)
+    return data
+
+
+def read_block(fd):
+    hdr = LcfHeader(*struct.unpack('iiiiiiiii', read(fd, 9 * 4)))
+    if hdr.version != LCF_VERSION:
+        raise BadBlock("Wrong LCF version")
+    if hdr.nf < 1 or hdr.nf > 1024:
+        raise BadBlock("Wrong number of frames")
+    if hdr.bpp != 16:
+        raise CorruptedBlock("Wrong bits per pixel")
+    frame_deltas = [struct.unpack('i', read(fd, 4))[0] for f in xrange(hdr.nf)]
+    cdata = read(fd, hdr.cdata_left)
+    decompress = zlib.decompressobj()
+    data = decompress.decompress(cdata)
+    if hdr.dsize != len(data):
+        raise CorruptedBlock("Wrong uncompressed data size")
+    # Decode slices
+    bytespersample = (hdr.bpp + 7) // 8
+    # Setup frames as blocks of ns_x * ns_y slices
+    ns_x = (hdr.w + hdr.bsize_w - 1) // hdr.bsize_w
+    ns_y = (hdr.h + hdr.bsize_h - 1) // hdr.bsize_h
+    frames = [[[[] for y in xrange(ns_y)] for x in xrange(ns_x)] for f in xrange(hdr.nf)]
+    sp = 0
+    for y, ypos in enumerate(xrange(0, hdr.h, hdr.bsize_h)):
+        hei = min(hdr.h - ypos, hdr.bsize_h)
+        for x, xpos in enumerate(xrange(0, hdr.w, hdr.bsize_w)):
+            wid = min(hdr.w - xpos, hdr.bsize_w)
+            sz1 = hei * wid * bytespersample
+            cnt = 0
+            for i in xrange(hdr.nf):
+                if cnt > 0:
+                    cnt -= 1
+                else:
+                    buff = data[sp:sp + sz1]
+                    if len(buff) != sz1:
+                        raise CorruptedBlock("Wrong size read!")
+                    arr = np.fromstring(buff, dtype=np.uint16)
+                    lvalid = np.zeros(arr.size * 3, np.uint8)
+                    lvalid.view()[0::3] = (arr << 3) & 0xf8
+                    lvalid.view()[1::3] = (arr >> 3) & 0xfc
+                    lvalid.view()[2::3] = (arr >> 8) & 0xf8
+                    lvalid.shape = (hei, wid, 3)
+                    sp += sz1
+                    if i < hdr.nf - 1:
+                        cnt = ord(data[sp])
+                        sp += 1
+                frames[i][x][y] = lvalid
+    if sp != len(data):
+        print "Not all data consumed (%s/%s)!" % (sp, len(data))
+    return LcfBlock(hdr, frame_deltas, frames)
+
+
+def process(filename):
+    blocks = []
+    with open(filename) as fd:
+        while True:
+            try:
+                blocks.append(read_block(fd))
+            except CorruptedBlock as e:
+                print(e)
+                break
+            except BadBlock:
+                break
+    return [LcfFrame(f, b, i, j) for i, b in enumerate(blocks) for j, f in enumerate(b.frames)]