diff --git a/Makefile b/Makefile
index 2866d03ab..e87cfb510 100644
--- a/Makefile
+++ b/Makefile
@@ -84,7 +84,6 @@ extract: preconditions md gamesconf
 		done; \
 		rmdir "$$d"; \
 	done
-	(for f in build/X.INDEXED/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a -p build/TOTAL.DATA build/X.INDEXED > build/XSINGLE.IDX
 
 index: preconditions md asmfx asmprelaunch asmdemo compress extract
 #
@@ -103,14 +102,14 @@ index: preconditions md asmfx asmprelaunch asmdemo compress extract
 # note: prelaunch must be first in TOTAL.DATA due to a hack in LoadStandardPrelaunch
 # note 2: these can not be padded because they are loaded at $0106 and padding would clobber the stack
 #
-	[ -f build/index ] || (bin/buildindexedfile.sh build/TOTAL.DATA build/PRELAUNCH.INDEXED < build/GAMES.SORTED > build/PRELAUNCH.IDX)
+	[ -f build/index ] || (bin/buildindexedfile.py build/TOTAL.DATA build/PRELAUNCH.INDEXED "" < build/GAMES.SORTED > build/PRELAUNCH.IDX)
 #
 # precompute indexed files for HGR & DHGR titles
 # note: these are not padded because they are all an exact block-multiple anyway
 #
 	[ -f build/index ] || bin/padto.sh build/TOTAL.DATA
-	[ -f build/index ] || ((for f in res/TITLE.HGR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA res/TITLE.HGR build/HGR.TITLES.LOG > build/TITLE.IDX)
-	[ -f build/index ] || ((for f in res/TITLE.DHGR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA res/TITLE.DHGR build/DHGR.TITLES.LOG > build/DTITLE.IDX)
+	[ -f build/index ] || ((for f in res/TITLE.HGR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA res/TITLE.HGR build/HGR.TITLES.LOG > build/TITLE.IDX)
+	[ -f build/index ] || ((for f in res/TITLE.DHGR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA res/TITLE.DHGR build/DHGR.TITLES.LOG > build/DTITLE.IDX)
 	[ -f build/index ] || bin/addfile.sh res/COVER build/TOTAL.DATA > src/index/res.cover.idx.a
 	[ -f build/index ] || bin/addfile.sh res/TITLE build/TOTAL.DATA > src/index/res.title.idx.a
 	[ -f build/index ] || bin/addfile.sh res/HELP build/TOTAL.DATA > src/index/res.help.idx.a
@@ -118,64 +117,64 @@ index: preconditions md asmfx asmprelaunch asmdemo compress extract
 # precompute indexed files for game help
 # note: these can be padded because they're loaded into $800 at a time when $800..$1FFF is clobber-able
 #
-	[ -f build/index ] || (bin/buildindexedfile.sh -p -a build/TOTAL.DATA build/GAMEHELP < build/GAMES.SORTED > build/GAMEHELP.IDX)
+	[ -f build/index ] || (bin/buildindexedfile.py -p -a build/TOTAL.DATA build/GAMEHELP "" < build/GAMES.SORTED > build/GAMEHELP.IDX)
 #
 # precompute indexed files for slideshows
 # note: these can be padded because they're loaded into $800 at a time when $800..$1FFF is clobber-able
 #
 	[ -f build/index ] || $(PARALLEL) '[ $$(echo "{/}" | cut -c-3) = "ACT" ] && bin/buildslideshow.sh -d build/GAMES.CONF < "{}" > "build/SS/{/}" || bin/buildslideshow.sh build/GAMES.CONF < "{}" > "build/SS/{/}"' ::: res/SS/*
-	[ -f build/index ] || ((for f in build/SS/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -p -a build/TOTAL.DATA build/SS > build/SLIDESHOW.IDX)
+	[ -f build/index ] || ((for f in build/SS/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -p -a build/TOTAL.DATA build/SS "" > build/SLIDESHOW.IDX)
 	[ -f build/index ] || $(PARALLEL) 'bin/buildokvs.sh < "{}" > "build/ATTRACT/{/}"' ::: res/ATTRACT/*
-	[ -f build/index ] || ((for f in build/ATTRACT/[ABCDEFGHIJKLMNOP]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -p -a build/TOTAL.DATA build/ATTRACT > build/MINIATTRACT0.IDX)
-	[ -f build/index ] || ((for f in build/ATTRACT/[QRSTUVWXYZ]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -p -a build/TOTAL.DATA build/ATTRACT > build/MINIATTRACT1.IDX)
+	[ -f build/index ] || ((for f in build/ATTRACT/[ABCDEFGHIJKLMNOP]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -p -a build/TOTAL.DATA build/ATTRACT "" > build/MINIATTRACT0.IDX)
+	[ -f build/index ] || ((for f in build/ATTRACT/[QRSTUVWXYZ]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -p -a build/TOTAL.DATA build/ATTRACT "" > build/MINIATTRACT1.IDX)
 #
 # precompute indexed files for graphic effects
 # note: these can be padded because they're loaded into $6000 at a time when $6000..$BEFF is clobber-able
 #
-	[ -f build/index ] || (bin/buildindexedfile.sh -p -a build/TOTAL.DATA build/FX.INDEXED < res/FX.CONF > build/FX.IDX)
-	[ -f build/index ] || (bin/buildindexedfile.sh -p -a build/TOTAL.DATA build/FX.INDEXED < res/DFX.CONF > build/DFX.IDX)
-	[ -f build/index ] || (bin/buildindexedfile.sh -p -a build/TOTAL.DATA build/FX.INDEXED < res/SFX.CONF > build/SFX.IDX)
-	[ -f build/index ] || ((for f in build/FXCODE/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -p -a build/TOTAL.DATA build/FXCODE > build/FXCODE.IDX)
+	[ -f build/index ] || (bin/buildindexedfile.py -p -a build/TOTAL.DATA build/FX.INDEXED "" < res/FX.CONF > build/FX.IDX)
+	[ -f build/index ] || (bin/buildindexedfile.py -p -a build/TOTAL.DATA build/FX.INDEXED "" < res/DFX.CONF > build/DFX.IDX)
+	[ -f build/index ] || (bin/buildindexedfile.py -p -a build/TOTAL.DATA build/FX.INDEXED "" < res/SFX.CONF > build/SFX.IDX)
+	[ -f build/index ] || ((for f in build/FXCODE/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -p -a build/TOTAL.DATA build/FXCODE "" > build/FXCODE.IDX)
 #
 # precompute indexed files for coordinates files loaded by graphic effects
 # note: these can not be padded because some of them are loaded into tight spaces near the unclobberable top of main memory
 #
-	[ -f build/index ] || ((for f in build/FXDATA/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA build/FXDATA > build/FXDATA.IDX)
+	[ -f build/index ] || ((for f in build/FXDATA/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA build/FXDATA "" > build/FXDATA.IDX)
 #
 # precompute indexed files for HGR & DHGR action screenshots
 # note: these can not be padded because they are compressed and the decompressor needs the exact size
 #
-	[ -f build/index ] || ((for f in res/ACTION.HGR/[ABCD]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA res/ACTION.HGR > build/HGR0.IDX)
-	[ -f build/index ] || ((for f in res/ACTION.HGR/[EFGH]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA res/ACTION.HGR > build/HGR1.IDX)
-	[ -f build/index ] || ((for f in res/ACTION.HGR/[IJKL]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA res/ACTION.HGR > build/HGR2.IDX)
-	[ -f build/index ] || ((for f in res/ACTION.HGR/[MNOP]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA res/ACTION.HGR > build/HGR3.IDX)
-	[ -f build/index ] || ((for f in res/ACTION.HGR/[QRST]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA res/ACTION.HGR > build/HGR4.IDX)
-	[ -f build/index ] || ((for f in res/ACTION.HGR/[UVWX]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA res/ACTION.HGR > build/HGR5.IDX)
-	[ -f build/index ] || ((for f in res/ACTION.HGR/[YZ]*;   do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA res/ACTION.HGR > build/HGR6.IDX)
-	[ -f build/index ] || ((for f in res/ACTION.DHGR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA res/ACTION.DHGR > build/DHGR.IDX)
+	[ -f build/index ] || ((for f in res/ACTION.HGR/[ABCD]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA res/ACTION.HGR "" > build/HGR0.IDX)
+	[ -f build/index ] || ((for f in res/ACTION.HGR/[EFGH]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA res/ACTION.HGR "" > build/HGR1.IDX)
+	[ -f build/index ] || ((for f in res/ACTION.HGR/[IJKL]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA res/ACTION.HGR "" > build/HGR2.IDX)
+	[ -f build/index ] || ((for f in res/ACTION.HGR/[MNOP]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA res/ACTION.HGR "" > build/HGR3.IDX)
+	[ -f build/index ] || ((for f in res/ACTION.HGR/[QRST]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA res/ACTION.HGR "" > build/HGR4.IDX)
+	[ -f build/index ] || ((for f in res/ACTION.HGR/[UVWX]*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA res/ACTION.HGR "" > build/HGR5.IDX)
+	[ -f build/index ] || ((for f in res/ACTION.HGR/[YZ]*;   do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA res/ACTION.HGR "" > build/HGR6.IDX)
+	[ -f build/index ] || ((for f in res/ACTION.DHGR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA res/ACTION.DHGR "" > build/DHGR.IDX)
 #
 # precompute indexed files for GR and DGR action screenshots
 # note: these can be padded because they are not compressed
 #
-	[ -f build/index ] || ((for f in res/ACTION.GR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a -p build/TOTAL.DATA res/ACTION.GR > build/GR.IDX)
-	[ -f build/index ] || ((for f in res/ACTION.DGR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a -p build/TOTAL.DATA res/ACTION.DGR > build/DGR.IDX)
+	[ -f build/index ] || ((for f in res/ACTION.GR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a -p build/TOTAL.DATA res/ACTION.GR "" > build/GR.IDX)
+	[ -f build/index ] || ((for f in res/ACTION.DGR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a -p build/TOTAL.DATA res/ACTION.DGR "" > build/DGR.IDX)
 #
 # precompute indexed files for SHR artwork
 # note: these can not be padded because they are compressed and the decompressor needs the exact size
 #
-	[ -f build/index ] || ((for f in res/ARTWORK.SHR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA res/ARTWORK.SHR > build/ARTWORK.IDX)
+	[ -f build/index ] || ((for f in res/ARTWORK.SHR/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA res/ARTWORK.SHR "" > build/ARTWORK.IDX)
 #
 # precompute indexed files for demo launchers
 # note: these can not be padded because some of them are loaded too close to $C000
 #
-	[ -f build/index ] || ((for f in build/DEMO/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a build/TOTAL.DATA build/DEMO > build/DEMO.IDX)
+	[ -f build/index ] || ((for f in build/DEMO/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a build/TOTAL.DATA build/DEMO "" > build/DEMO.IDX)
 	[ -f build/index ] || bin/addfile.sh build/DEMO.IDX build/TOTAL.DATA > src/index/demo.idx.a
 
 #
 # precompute indexed files for single-load game binaries
 # note: these can be padded because they are loaded at a time when all of main memory is clobber-able
 #
-	[ -f build/index ] || ((for f in build/X.INDEXED/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.sh -a -p build/TOTAL.DATA build/X.INDEXED > build/XSINGLE.IDX)
+	[ -f build/index ] || ((for f in build/X.INDEXED/*; do echo "$$(basename $$f)"; done) | bin/buildindexedfile.py -a -p build/TOTAL.DATA build/X.INDEXED "" > build/XSINGLE.IDX)
 	[ -f build/index ] || bin/addfile.sh build/XSINGLE.IDX build/TOTAL.DATA > src/index/xsingle.idx.a
 #
 # create search indexes for each variation of (game-requires-joystick) X (game-requires-128K)
diff --git a/bin/buildindexedfile.py b/bin/buildindexedfile.py
new file mode 100755
index 000000000..bf6a65d5c
--- /dev/null
+++ b/bin/buildindexedfile.py
@@ -0,0 +1,100 @@
+#!/usr/bin/env python3
+
+# flags
+# -a  append to data file (default off = truncate)
+# -p  pad sizes within data file to next block size (default off)
+
+# parameters
+# stdin - input containing list of files (e.g. FX.CONF)
+# stdout - binary OKVS data structure
+# 1 - output filename for data file
+# 2 - input directory of files to merge into data file
+# 3 - (optional) output filename for log of key,offset,size
+
+import argparse
+import os
+import os.path
+import struct
+import sys
+
+kStandardFilename = 'STANDARD'
+kFileMode = {False: 'wb', True: 'ab'}
+
+def build(records, args):
+    if args.append:
+        output_file_size = os.path.getsize(args.output_file)
+    else:
+        output_file_size = 0
+    standard_offset = standard_size = 0
+    standard_filename = os.path.join(args.input_directory, kStandardFilename)
+    with open(args.output_file, kFileMode[args.append]) as output_file_handle:
+        if os.path.exists(standard_filename):
+            standard_offset = output_file_size
+            standard_size = os.path.getsize(standard_filename)
+            with open(standard_filename, 'rb') as standard_handle:
+                output_file_handle.write(standard_handle.read())
+            output_file_size += standard_size
+
+        # yield OKVS header (2 x 2 bytes, unsigned int, little-endian)
+        yield struct.pack('<H', len(records))
+        yield struct.pack('<H', 0)
+
+        for rec in records:
+            filename, dummy, dummy = rec.partition('=')
+            key, dummy, addr = filename.partition('#')
+            filename = os.path.join(args.input_directory, filename)
+            rec_offset = standard_offset
+            rec_size = standard_size
+
+            # yield record length (1 byte, unsigned byte)
+            if addr:
+                # if filename is in the form 'NAME#06ADDR' then create extended index record
+                yield struct.pack('B', len(key)+9)
+                # trim '06' so we get just the starting address
+                addr = addr[2:]
+            else:
+                yield struct.pack('B', len(key)+7)
+
+            # yield key (Pascal-style string)
+            yield struct.pack(f'{len(key)+1}p', key.encode('ascii'))
+
+            if os.path.exists(filename):
+                rec_offset = output_file_size
+                rec_size = os.path.getsize(filename)
+                output_file_size += rec_size
+                if args.pad:
+                    # If offset+size does not cross a block boundary, use file's true size.
+                    # Otherwise, round up size to the next block boundary.
+                    # This padding does not get added to the file; it is just an
+                    # optimization to avoid a partial copy on the last block read.
+                    if (rec_offset // 512) != ((rec_offset + rec_size) // 512):
+                        rec_size = ((rec_offset + rec_size + 511) & -512) - rec_offset
+                with open(filename, 'rb') as input_file_handle:
+                    output_file_handle.write(input_file_handle.read())
+
+            # yield record offset (3 bytes, big-endian, unsigned long)
+            yield struct.pack('>L', rec_offset)[1:]
+
+            # yield record size (2 bytes, little-endian, unsigned short)
+            yield struct.pack('<H', rec_size)
+
+            if addr:
+                # for extended index record, yield load address (2 bytes, little-endian, unsigned short)
+                yield bytes.fromhex(addr)[::-1]
+
+            if args.log_file:
+                with open(args.log_file, 'a') as log_file_handle:
+                    log_file_handle.write(f'{key},{rec_offset},{rec_size}\n')
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Build indexed OKVS structure with links to data stored in TOTAL.DATA file")
+    parser.add_argument("output_file")
+    parser.add_argument("input_directory")
+    parser.add_argument("log_file")
+    parser.add_argument("-p", "--pad", action="store_true", default=False, help="pad file sizes to multiples of 512 (default: use exact size)")
+    parser.add_argument("-a", "--append", action="store_true", default=False, help="append to output file (default: overwrite)")
+    args = parser.parse_args()
+    records = [x.strip() for x in sys.stdin.readlines()]
+    records = [x for x in records if x and x[0] not in ('#', '[')]
+    for b in build(records, args):
+        sys.stdout.buffer.write(b)
diff --git a/bin/buildindexedfile.sh b/bin/buildindexedfile.sh
deleted file mode 100755
index 028dd13dc..000000000
--- a/bin/buildindexedfile.sh
+++ /dev/null
@@ -1,95 +0,0 @@
-#!/bin/bash
-
-# flags
-# -a  append to data file (default off = truncate)
-# -p  pad sizes within data file to next block size (default off)
-
-# parameters
-# stdin - input containing list of files (e.g. FX.CONF)
-# stdout - binary OKVS data structure
-# 1 - output filename for data file
-# 2 - input directory of files to merge into data file
-# 3 - (optional) output filename for log of key,offset,size
-
-pad=false
-append=false
-standardoffset=0
-standardsize=0
-while getopts ":ap" opt; do
-    case $opt in
-        a) append=true
-           ;;
-        p) pad=true
-           ;;
-    esac
-done
-shift $((OPTIND-1))
-
-if [ "$append" = false ]; then
-    rm -f "$1"
-fi
-touch "$1"
-
-if [ "${#3}" -ne "0" ]; then
-    rm -f "$3"
-    touch "$3"
-fi
-
-# if there is a file called "STANDARD" in the input directory, add it now
-# because we will reuse it for any files that don't exist
-if [ -f "$2"/STANDARD ]; then
-    standardoffset=$(wc -c < "$1")
-    standardsize=$(wc -c < "$2/STANDARD")
-    cat "$2"/STANDARD >> "$1"
-fi
-
-# make temp file with list of lines that contain keys
-records=$(mktemp)
-tr -d "\r" | awk '!/^$|^#/' > "$records"
-
-# make temp assembly source file that represents the binary OKVS data structure
-source=$(mktemp)
-(echo "*=0"                              # dummy program counter for assembler
- echo "!le16 $(wc -l <"$records"), 0"    # OKVS header
- while IFS="=" read -r filename dummy; do echo "$filename" | {
-     IFS="#" read -r key addr
-     if [ "${#addr}" -ne "0" ]; then     # if filename is in the form 'NAME#06ADDR' then create extended index record
-         addr=$(echo "$addr" | cut -c3-) # trim '06' so we get just the starting address
-         echo "!byte ${#key}+9"          # OKVS record length
-     else
-         echo "!byte ${#key}+7"          # OKVS record length
-     fi
-     echo "!byte ${#key}"                # OKVS key length
-     echo "!text \"$key\""               # OKVS key
-     if [ ! -e "$2/$filename" ]; then    # if file does not exist, use standard offset and size
-         offset="$standardoffset"
-         size="$standardsize"
-     else                                # otherwise calculate offset and size from file and options
-         offset=$(wc -c < "$1")
-         size=$(wc -c < "$2/$filename")
-         if [ "$pad" = true ]; then
-             # If offset+size does not cross a block boundary, use file's true size.
-             # Otherwise, round up size to the next block boundary.
-             # This padding does not get added to the file; it is just an
-             # optimization to avoid a partial copy on the last block read.
-             if [ $(($offset / 512)) -ne $((($offset + $size) / 512)) ]; then
-                 size=$(((($offset + $size + 511) & -512) - $offset))
-             fi
-         fi
-         cat "$2/$filename" >> "$1"      # append this file to the end of the merged data file
-     fi
-     echo "!be24 $offset"
-     echo "!le16 $size"
-     [ "${#addr}" -ne "0" ] && echo '!le16 $'"$addr"
-     [ "${#3}" -ne "0" ] && echo "$key,$offset,$size" >> "$3"
- } done < "$records") > "$source"
-
-# assemble temp source file into binary OKVS data structure, then output that
-out=$(mktemp)
-acme -o "$out" "$source"
-cat "$out"
-
-# clean up
-rm "$out"
-rm "$source"
-rm "$records"