diff --git a/output/dll/pizza.wbx.gz b/output/dll/pizza.wbx.gz
index 5937922ea3..cbc7062d17 100644
Binary files a/output/dll/pizza.wbx.gz and b/output/dll/pizza.wbx.gz differ
diff --git a/waterbox/pizza/lib/sgb.c b/waterbox/pizza/lib/sgb.c
index a006431c78..fbb71b6e5f 100644
--- a/waterbox/pizza/lib/sgb.c
+++ b/waterbox/pizza/lib/sgb.c
@@ -75,7 +75,7 @@ typedef struct
 
 	// frame data
 	uint8_t frame[160 * 144];		// the most recent obtained full frame
-	uint8_t frozenframe[160 * 144]; // the most recent saved full frame (MASK_EN)
+	uint32_t frozenframe[256 * 224]; // the most recent saved full frame (MASK_EN)
 	uint8_t attr[20 * 18];			// current attr map for the GB screen
 	uint8_t auxattr[45][20 * 18];   // 45 attr files
 
@@ -433,7 +433,8 @@ static void cmd_mask(void)
 		case 2:
 		case 3:
 			sgb.active_mask = 1;
-			memset(sgb.frozenframe, 0, sizeof(sgb.frozenframe));
+			for (int i = 0; i < 256 * 224; i++)
+				sgb.frozenframe[i] = sgb.palette[0][0];
 			break;
 		}
 	}
@@ -862,32 +863,10 @@ static void do_vram_transfer(void)
 	}
 }
 
-// 160x144 32bpp pixel data
-// assumed to contain exact pixel values 00, 55, aa, ff
-void sgb_take_frame(uint32_t *vbuff)
-{
-	for (int i = 0; i < 160 * 144; i++)
-	{
-		sgb.frame[i] = 3 - (vbuff[i] >> 6 & 3); // 0, 1, 2, or 3 for each pixel
-	}
-	if (sgb.waiting_transfer != TRN_NONE)
-	{
-		if (!--sgb.transfer_countdown)
-		{
-			do_vram_transfer();
-			sgb.waiting_transfer = TRN_NONE;
-		}
-	}
-	if (!sgb.active_mask)
-	{
-		memcpy(sgb.frozenframe, sgb.frame, sizeof(sgb.frame));
-	}
-}
-
 static void sgb_render_frame_gb(uint32_t *vbuff)
 {
 	const uint8_t *attr = sgb.attr;
-	const uint8_t *src = sgb.active_mask ? sgb.frozenframe : sgb.frame;
+	const uint8_t *src = sgb.frame;
 	uint32_t *dst = vbuff + ((224 - 144) / 2 * 256 + (256 - 160) / 2);
 
 	for (int j = 0; j < 144; j++)
@@ -951,12 +930,35 @@ static void sgb_render_border(uint32_t *vbuff)
 	}
 }
 
+// 160x144 32bpp pixel data
+// assumed to contain exact pixel values 00, 55, aa, ff
+void sgb_take_frame(uint32_t *vbuff)
+{
+	for (int i = 0; i < 160 * 144; i++)
+	{
+		sgb.frame[i] = 3 - (vbuff[i] >> 6 & 3); // 0, 1, 2, or 3 for each pixel
+	}
+	if (sgb.waiting_transfer != TRN_NONE)
+	{
+		if (!--sgb.transfer_countdown)
+		{
+			do_vram_transfer();
+			sgb.waiting_transfer = TRN_NONE;
+		}
+	}
+	if (!sgb.active_mask)
+	{
+		// render the frame now
+		for (int i = 0; i < 256 * 224; i++)
+			sgb.frozenframe[i] = sgb.palette[0][0];
+		sgb_render_frame_gb(sgb.frozenframe);
+		sgb_render_border(sgb.frozenframe);
+	}
+}
+
 void sgb_render_frame(uint32_t *vbuff)
 {
-	for (int i = 0; i < 256 * 224; i++)
-		vbuff[i] = sgb.palette[0][0];
-	sgb_render_frame_gb(vbuff);
-	sgb_render_border(vbuff);
+	memcpy(vbuff, sgb.frozenframe, sizeof(sgb.frozenframe));
 }
 
 void sgb_render_audio(uint64_t time, void (*callback)(int16_t l, int16_t r, uint64_t time))
diff --git a/waterbox/pizza/lib/sgb.h b/waterbox/pizza/lib/sgb.h
index 4c1009ac08..35e76dcad9 100644
--- a/waterbox/pizza/lib/sgb.h
+++ b/waterbox/pizza/lib/sgb.h
@@ -1,16 +1,36 @@
 #pragma once
 #include <stdint.h>
 
+// whenever a time is asked for, it is relative to a clock that ticks 35112 times
+// per nominal frame on the GB lcd, starts at 0 when emulation begins, and never resets/rebases
+
+// write to MMIO ff00. only bits 4 and 5 are used
 void sgb_write_ff00(uint8_t val, uint64_t time);
 
+// read from MMIO ff00. supplies data for all 8 bits
 uint8_t sgb_read_ff00(uint64_t time);
 
+// set controller data to be used by subsequent controller reads
+// buttons[0] = controller 1, buttons[3] = controller 4
+// 7......0
+// DULRSsBA
 void sgb_set_controller_data(const uint8_t* buttons);
 
+// initialize the SGB module.  pass an SPC file that results from the real S-CPU initialization,
+// and the length of that file
 int sgb_init(const uint8_t* spc, int length);
 
+// call whenever the gameboy has finished producing a video frame
+// data is 32bpp 160x144 screen data.  for each pixel:
+//31                          7      0
+// xxxxxxxx xxxxxxxx xxxxxxxx DDxxxxxx -- DD = 0, 1, 2, or 3.  x = don't care
 void sgb_take_frame(uint32_t* vbuff);
 
+// copy the finished video frame to an output buffer.  pixel format is 32bpp xrgb
+// can be called at any time, including right after sgb_take_frame
 void sgb_render_frame(uint32_t* vbuff);
 
+// call to finish a frame's worth of audio. should be called once every 35112 time units (some jitter is OK)
+// callback will be called with L and R sample values for various time points
+// between the last time sgb_render_audio was called and now
 void sgb_render_audio(uint64_t time, void(*callback)(int16_t l, int16_t r, uint64_t time));