diff --git a/dither-image.js b/dither-image.js
index 66fa869..1241eb8 100644
--- a/dither-image.js
+++ b/dither-image.js
@@ -654,6 +654,14 @@ function byte2byte(arr) {
     return lines.join('\n');
 }
 
+//let [even, odd] = [0, 1].map((bit) => (arr) => arr.filter((_item, index) => (index & 1) === bit));
+function even(arr) {
+    return arr.filter((_item, index) => !(index & 1));
+}
+function odd(arr) {
+    return arr.filter((_item, index) => (index & 1));
+}
+
 function genAssembly(width, height, nbits, lines, reps) {
     let linesEach = height / reps;
     let stride = width * nbits / 8;
@@ -684,14 +692,20 @@ function genAssembly(width, height, nbits, lines, reps) {
     return `.data
 .export frame1_top
 .export frame1_bottom
-.export frame1_palette1
-.export frame1_palette2
-.export frame1_palette3
+.export frame1_palette1_even
+.export frame1_palette1_odd
+.export frame1_palette2_even
+.export frame1_palette2_odd
+.export frame1_palette3_even
+.export frame1_palette3_odd
 .export frame2_top
 .export frame2_bottom
-.export frame2_palette1
-.export frame2_palette2
-.export frame2_palette3
+.export frame2_palette1_even
+.export frame2_palette1_odd
+.export frame2_palette2_even
+.export frame2_palette2_odd
+.export frame2_palette3_even
+.export frame2_palette3_odd
 .export displaylist
 
 .segment "BUFFERS"
@@ -700,17 +714,29 @@ function genAssembly(width, height, nbits, lines, reps) {
 frame1_top:
 ${byte2byte(frames[0].bitmap.slice(0, half))}
 
-.align 256
-frame1_palette1:
-${byte2byte(frames[0].palette1)}
+.align 128
+frame1_palette1_even:
+${byte2byte(even(frames[0].palette1))}
 
-.align 256
-frame1_palette2:
-${byte2byte(frames[0].palette2)}
+.align 128
+frame1_palette1_odd:
+${byte2byte(odd(frames[0].palette1))}
 
-.align 256
-frame1_palette3:
-${byte2byte(frames[0].palette3)}
+.align 128
+frame1_palette2_even:
+${byte2byte(even(frames[0].palette2))}
+
+.align 128
+frame1_palette2_odd:
+${byte2byte(odd(frames[0].palette2))}
+
+.align 128
+frame1_palette3_even:
+${byte2byte(even(frames[0].palette3))}
+
+.align 128
+frame1_palette3_odd:
+${byte2byte(odd(frames[0].palette3))}
 
 .align 4096
 frame1_bottom:
@@ -722,17 +748,29 @@ ${byte2byte(frames[0].bitmap.slice(half))}
 frame2_top:
 ${byte2byte(frames[1].bitmap.slice(0, half))}
 
-.align 256
-frame2_palette1:
-${byte2byte(frames[1].palette1)}
+.align 128
+frame2_palette1_even:
+${byte2byte(even(frames[1].palette1))}
 
-.align 256
-frame2_palette2:
-${byte2byte(frames[1].palette2)}
+.align 128
+frame2_palette1_odd:
+${byte2byte(odd(frames[1].palette1))}
 
-.align 256
-frame2_palette3:
-${byte2byte(frames[1].palette3)}
+.align 128
+frame2_palette2_even:
+${byte2byte(even(frames[1].palette2))}
+
+.align 128
+frame2_palette2_odd:
+${byte2byte(odd(frames[1].palette2))}
+
+.align 128
+frame2_palette3_even:
+${byte2byte(even(frames[1].palette3))}
+
+.align 128
+frame2_palette3_odd:
+${byte2byte(odd(frames[1].palette3))}
 
 .align 4096
 frame2_bottom:
diff --git a/dither4.s b/dither4.s
index c9f886c..9b8bb23 100644
--- a/dither4.s
+++ b/dither4.s
@@ -34,7 +34,7 @@ bytes_per_line = 40
 pages_per_frame = 32
 lines_per_frame = 262
 ;scanline_offset = 31 + (40 - 24) / 2
-scanline_offset = 48
+scanline_offset = 40
 scanline_max = (lines_per_frame - scanline_offset) / 2
 
 .data
@@ -44,14 +44,20 @@ scanline_max = (lines_per_frame - scanline_offset) / 2
 
 .import frame1_top
 .import frame1_bottom
-.import frame1_palette1
-.import frame1_palette2
-.import frame1_palette3
+.import frame1_palette1_even
+.import frame1_palette1_odd
+.import frame1_palette2_even
+.import frame1_palette2_odd
+.import frame1_palette3_even
+.import frame1_palette3_odd
 .import frame2_top
 .import frame2_bottom
-.import frame2_palette1
-.import frame2_palette2
-.import frame2_palette3
+.import frame2_palette1_even
+.import frame2_palette1_odd
+.import frame2_palette2_even
+.import frame2_palette2_odd
+.import frame2_palette3_even
+.import frame2_palette3_odd
 .import displaylist
 
 audio_high_byte:
@@ -136,16 +142,17 @@ wait_loop:
         sta audiotemp ; 3 cyc
     .endmacro
 
-    .macro inner_scanline palette1, palette2, palette3
+    .macro inner_scanline frame_offset, line_offset
+        ; Y should be VCOUNT at entry
         ; it'll fire on unused lines, but harmlessly
-        ldy scanline ; 3 cyc
-        inc scanline ; 5 cyc
+        ;ldy scanline ; 3 cyc
+        ;inc scanline ; 5 cyc
 
         ; Leisurely memory fetches
-        lda palette1,y ; 5 @FIXME alternate
+        lda frame1_palette1_even + frame_offset + line_offset - scanline_offset / 2,y ; 5 @FIXME alternate
         pha                   ; ...3?
-        ldx palette2,y ; 5
-        lda palette3,y ; 5
+        ldx frame1_palette2_even + frame_offset + line_offset - scanline_offset / 2,y ; 5
+        lda frame1_palette3_even + frame_offset + line_offset - scanline_offset / 2,y ; 5
         tay
         pla
 
@@ -159,7 +166,7 @@ wait_loop:
     .endmacro
 
     .macro audio_play_raw
-        ldy VCOUNT
+        ;ldy VCOUNT ; set on entry
         lda (sample_ptr),y ; 5/6 cyc
         sta AUDC1 ; 4 cyc
     .endmacro
@@ -202,14 +209,17 @@ wait_loop:
         audio_cont:
     .endmacro
 
-    .macro run_frame palette1, palette2, palette3
+    .macro run_frame frame_offset
     .scope
     each_scanline_pair:
         ;audio_prep
-        inner_scanline palette1, palette2, palette3
+        sty scanline
+        inner_scanline frame_offset, 0
+
+        ldy scanline
         audio_play_raw
         ;audio_play_lo
-        inner_scanline palette1, palette2, palette3
+        inner_scanline frame_offset, 128
         ;audio_play_raw
         ;audio_play_hi ; too slow
 
@@ -232,9 +242,9 @@ wait_loop:
     beq run_frame1
     jmp run_frame2
 run_frame1:
-    run_frame frame1_palette1, frame1_palette2, frame1_palette3
+    run_frame 0
 run_frame2:
-    run_frame frame2_palette1, frame2_palette2, frame2_palette3
+    run_frame 8192
 
 .endproc