diff --git a/.gitignore b/.gitignore
index 2fe683d..e54c2d1 100644
--- a/.gitignore
+++ b/.gitignore
@@ -13,6 +13,7 @@ __pycache__
 scramble
 cdi4dc
 tools/ttf_outline
+tools/ttf_bitmap
 k_means_vq
 *.blend1
 *.scramble
diff --git a/example/maple_font.cpp b/example/maple_font.cpp
new file mode 100644
index 0000000..f54df9b
--- /dev/null
+++ b/example/maple_font.cpp
@@ -0,0 +1,243 @@
+#include <cstdint>
+#include <bit>
+
+#include "maple/maple.hpp"
+#include "maple/maple_bus_commands.hpp"
+#include "maple/maple_bus_bits.hpp"
+#include "maple/maple_host_command_writer.hpp"
+#include "maple/maple_port.hpp"
+#include "align.hpp"
+#include "sh7091/serial.hpp"
+
+#include "holly/video_output.hpp"
+
+extern uint32_t _binary_font_portfolio_6x8 __asm("_binary_font_portfolio_6x8_portfolio_6x8_data_start");
+
+namespace vmu_display {
+constexpr int32_t width = 48;
+constexpr int32_t height = 32;
+constexpr int32_t pixels_per_byte = 8;
+constexpr int32_t framebuffer_size = width * height / pixels_per_byte;
+}
+
+uint32_t vmu_framebuffer[vmu_display::framebuffer_size / 4];
+
+static inline void render_glyph(uint8_t * dst, const uint8_t * src, uint8_t c, int x, int y)
+{
+  int y_ix = 186 - (y * 6 * 8);
+  for (int i = 0; i < 8; i++) {
+    switch (x) {
+    case 0:
+      dst[y_ix - i * 6 + 5] = src[(c - ' ') * 8 + i];
+      break;
+    case 1:
+      dst[y_ix - i * 6 + 5] |= (src[(c - ' ') * 8 + i] & 0b11) << 6;
+      dst[y_ix - i * 6 + 4] = src[(c - ' ') * 8 + i] >> 2; // 0b1111
+      break;
+    case 2:
+      dst[y_ix - i * 6 + 4] |= (src[(c - ' ') * 8 + i] & 0b1111) << 4;
+      dst[y_ix - i * 6 + 3] = src[(c - ' ') * 8 + i] >> 4; // 0b11
+      break;
+    case 3:
+      dst[y_ix - i * 6 + 3] |= src[(c - ' ') * 8 + i] << 2;
+      break;
+    case 4:
+      dst[y_ix - i * 6 + 2] = src[(c - ' ') * 8 + i];
+      break;
+    case 5:
+      dst[y_ix - i * 6 + 2] |= (src[(c - ' ') * 8 + i] & 0b11) << 6;
+      dst[y_ix - i * 6 + 1] = src[(c - ' ') * 8 + i] >> 2; // 0b1111
+      break;
+    case 6:
+      dst[y_ix - i * 6 + 1] |= (src[(c - ' ') * 8 + i] & 0b1111) << 4;
+      dst[y_ix - i * 6 + 0] = src[(c - ' ') * 8 + i] >> 4; // 0b11
+      break;
+    case 7:
+      dst[y_ix - i * 6 + 0] |= src[(c - ' ') * 8 + i] << 2;
+      break;
+    }
+  }
+}
+
+void make_vmu_framebuffer(uint32_t * buf)
+{
+  const uint8_t * src = reinterpret_cast<const uint8_t *>(&_binary_font_portfolio_6x8);
+  uint8_t * dst = reinterpret_cast<uint8_t *>(buf);
+
+  for (int i = 0; i < vmu_display::framebuffer_size; i++) {
+    dst[i] = 0;
+  }
+  const char * s =
+    " very   "
+    " funneh "
+    "        "
+    " :))    ";
+  for (int i = 0; i < 8 * 4; i++) {
+    int x = i % 8;
+    int y = i / 8;
+    render_glyph(dst, src, s[i], x, y);
+  }
+}
+
+template <typename T>
+inline void copy(T * dst, const T * src, const int32_t n) noexcept
+{
+  int32_t n_t = n / (sizeof (T));
+  while (n_t > 0) {
+    *dst++ = *src++;
+    n_t--;
+  }
+}
+
+void send_vmu_framebuffer(uint8_t port, uint8_t lm)
+{
+  uint32_t send_buf[1024] __attribute__((aligned(32)));
+  uint32_t recv_buf[1024] __attribute__((aligned(32)));
+
+  using command_type = maple::block_write<uint8_t[0]>;
+  using response_type = maple::device_reply;
+
+  auto writer = maple::host_command_writer(send_buf, recv_buf);
+
+  uint32_t host_port_select = host_instruction_port_select(port);
+  uint32_t destination_ap = ap_port_select(port) | ap::de::expansion_device | lm;
+
+  auto [host_command, host_response]
+    = writer.append_command<command_type, response_type>(host_port_select,
+							 destination_ap,
+							 true,      // end_flag
+							 vmu_display::framebuffer_size, // send_trailing
+							 0          // recv_trailing
+							 );
+  auto& data_fields = host_command->bus_data.data_fields;
+  data_fields.function_type = std::byteswap(function_type::bw_lcd);
+  data_fields.pt = 0;
+  data_fields.phase = 0;
+  data_fields.block_number = std::byteswap<uint16_t>(0x0000);
+
+  copy<uint8_t>(data_fields.written_data, reinterpret_cast<uint8_t *>(vmu_framebuffer), vmu_display::framebuffer_size);
+
+  maple::dma_start(send_buf, writer.send_offset,
+		   recv_buf, writer.recv_offset);
+
+  serial::integer<uint8_t>(host_response->bus_data.command_code);
+  serial::integer<uint8_t>(host_response->bus_data.destination_ap);
+  serial::integer<uint8_t>(host_response->bus_data.source_ap);
+  serial::integer<uint8_t>(host_response->bus_data.data_size);
+}
+
+void do_lm_request(uint8_t port, uint8_t lm)
+{
+  uint32_t send_buf[1024] __attribute__((aligned(32)));
+  uint32_t recv_buf[1024] __attribute__((aligned(32)));
+
+  auto writer = maple::host_command_writer(send_buf, recv_buf);
+
+  uint32_t host_port_select = host_instruction_port_select(port);
+  uint32_t destination_ap = ap_port_select(port) | ap::de::expansion_device | lm;
+
+  using command_type = maple::device_request;
+  using response_type = maple::device_status;
+
+  auto [host_command, host_response]
+    = writer.append_command<command_type, response_type>(host_port_select,
+							 destination_ap,
+							 true); // end_flag
+
+  maple::dma_start(send_buf, writer.send_offset,
+                   recv_buf, writer.recv_offset);
+
+  auto& bus_data = host_response->bus_data;
+  auto& data_fields = bus_data.data_fields;
+  if (bus_data.command_code != maple::device_status::command_code) {
+    serial::string("lm did not reply: ");
+    serial::integer<uint8_t>(port, ' ');
+    serial::integer<uint8_t>(lm);
+  } else {
+    serial::string("    lm: ");
+    serial::integer<uint8_t>(lm);
+    serial::string("      ft:    ");
+    serial::integer<uint32_t>(std::byteswap(data_fields.device_id.ft));
+    serial::string("      fd[0]: ");
+    serial::integer<uint32_t>(std::byteswap(data_fields.device_id.fd[0]));
+    serial::string("      fd[1]: ");
+    serial::integer<uint32_t>(std::byteswap(data_fields.device_id.fd[1]));
+    serial::string("      fd[2]: ");
+    serial::integer<uint32_t>(std::byteswap(data_fields.device_id.fd[2]));
+    serial::string("      source_ap.lm_bus: ");
+    serial::integer<uint8_t>(bus_data.source_ap & ap::lm_bus::bit_mask);
+
+    if (std::byteswap(data_fields.device_id.ft) & function_type::bw_lcd) {
+      serial::string("send vmu_framebuffer\n");
+      send_vmu_framebuffer(port, lm);
+    }
+  }
+}
+
+void do_lm_requests(uint8_t port, uint8_t lm)
+{
+  if (lm & ap::lm_bus::_0)
+    do_lm_request(port, lm & ap::lm_bus::_0);
+  if (lm & ap::lm_bus::_1)
+    do_lm_request(port, lm & ap::lm_bus::_1);
+  if (lm & ap::lm_bus::_2)
+    do_lm_request(port, lm & ap::lm_bus::_2);
+  if (lm & ap::lm_bus::_3)
+    do_lm_request(port, lm & ap::lm_bus::_3);
+  if (lm & ap::lm_bus::_4)
+    do_lm_request(port, lm & ap::lm_bus::_4);
+}
+
+void do_device_request()
+{
+  uint32_t send_buf[1024] __attribute__((aligned(32)));
+  uint32_t recv_buf[1024] __attribute__((aligned(32)));
+
+  auto writer = maple::host_command_writer(send_buf, recv_buf);
+
+  using command_type = maple::device_request;
+  using response_type = maple::device_status;
+
+  auto [host_command, host_response]
+    = writer.append_command_all_ports<command_type, response_type>();
+
+  maple::dma_start(send_buf, writer.send_offset,
+                   recv_buf, writer.recv_offset);
+
+  for (uint8_t port = 0; port < 4; port++) {
+    auto& bus_data = host_response[port].bus_data;
+    auto& data_fields = bus_data.data_fields;
+    if (bus_data.command_code != response_type::command_code) {
+      serial::string("port: ");
+      serial::integer<uint8_t>(port);
+      serial::string("  disconnected\n");
+    } else {
+      serial::string("port: ");
+      serial::integer<uint8_t>(port);
+      serial::string("  ft:    ");
+      serial::integer<uint32_t>(std::byteswap(data_fields.device_id.ft));
+      serial::string("  fd[0]: ");
+      serial::integer<uint32_t>(std::byteswap(data_fields.device_id.fd[0]));
+      serial::string("  fd[1]: ");
+      serial::integer<uint32_t>(std::byteswap(data_fields.device_id.fd[1]));
+      serial::string("  fd[2]: ");
+      serial::integer<uint32_t>(std::byteswap(data_fields.device_id.fd[2]));
+      serial::string("  source_ap.lm_bus: ");
+      serial::integer<uint8_t>(bus_data.source_ap & ap::lm_bus::bit_mask);
+
+      do_lm_requests(port,
+		     bus_data.source_ap & ap::lm_bus::bit_mask);
+    }
+  }
+}
+
+void main()
+{
+  serial::init(4);
+
+  make_vmu_framebuffer(vmu_framebuffer);
+
+  do_device_request();
+
+  while (1);
+}
diff --git a/font/portfolio_6x8/Bm437_Portfolio_6x8.otb b/font/portfolio_6x8/Bm437_Portfolio_6x8.otb
new file mode 100644
index 0000000..d78f723
Binary files /dev/null and b/font/portfolio_6x8/Bm437_Portfolio_6x8.otb differ
diff --git a/font/portfolio_6x8/portfolio_6x8.data b/font/portfolio_6x8/portfolio_6x8.data
new file mode 100644
index 0000000..25618f5
Binary files /dev/null and b/font/portfolio_6x8/portfolio_6x8.data differ
diff --git a/tools/ttf_bitmap.cpp b/tools/ttf_bitmap.cpp
new file mode 100644
index 0000000..884ab8d
--- /dev/null
+++ b/tools/ttf_bitmap.cpp
@@ -0,0 +1,126 @@
+#include <sstream>
+#include <iostream>
+
+#include <assert.h>
+#include <stdint.h>
+
+#include <ft2build.h>
+#include FT_FREETYPE_H
+
+int32_t
+load_bitmap_char(FT_Face face,
+		 FT_ULong char_code,
+		 uint8_t * buf)
+{
+  FT_Error error;
+  FT_UInt glyph_index = FT_Get_Char_Index(face, char_code);
+
+  error = FT_Load_Glyph(face, glyph_index, FT_LOAD_DEFAULT);
+  if (error) {
+    std::cerr << "FT_Load_Glyph " << FT_Error_String(error) << '\n';
+    return -1;
+  }
+
+  assert(face->glyph->format == FT_GLYPH_FORMAT_BITMAP);
+  //printf("num_grays %d\n", face->glyph->bitmap.num_grays);
+  //printf("pitch %d\n", face->glyph->bitmap.pitch);
+  //printf("width %d\n", face->glyph->bitmap.width);
+  //printf("char_code %lx rows %d\n", char_code, face->glyph->bitmap.rows);
+  //assert((face->glyph->bitmap.rows % 8) == 0);
+  //assert(face->glyph->bitmap.width / face->glyph->bitmap.pitch == 8);
+
+  for (int y = 0; y < (int)face->glyph->bitmap.rows; y++) {
+    uint8_t * row = &face->glyph->bitmap.buffer[y * face->glyph->bitmap.pitch];
+    uint8_t row_out = 0;
+    for (unsigned int x = 0; x < face->glyph->bitmap.width; x++) {
+      if (x % 8 == 0) row_out = 0;
+      const uint8_t bit = (row[x / 8] >> (7 - (x % 8))) & 1;
+      std::cerr << (bit ? "█" : " ");
+      row_out |= (bit << (x % 8));
+      if (x % 8 == 7 || x == (face->glyph->bitmap.width - 1))
+        buf[(y * face->glyph->bitmap.pitch) + (x / 8)] = row_out;
+    }
+    std::cerr << "|\n";
+  }
+
+  // 'pitch' is bytes; 'width' is pixels
+  return face->glyph->bitmap.rows * face->glyph->bitmap.pitch;
+}
+
+template <typename T>
+constexpr inline T
+parse_num(decltype(std::hex)& format, const char * s)
+{
+  T n;
+  std::stringstream ss;
+  ss << format << s;
+  ss >> n;
+  return n;
+}
+
+int main(int argc, char *argv[])
+{
+  FT_Library library;
+  FT_Face face;
+  FT_Error error;
+
+  if (argc != 5) {
+    std::cerr << "usage: " << argv[0] << " [start-hex] [end-hex] [font-file-path] [output-file-path]\n\n";
+    std::cerr << "   ex: " << argv[0] << " 3000 30ff ipagp.ttf font.bin\n";
+    return -1;
+  }
+
+  auto start = parse_num<uint32_t>(std::hex, argv[1]);
+  auto end = parse_num<uint32_t>(std::hex, argv[2]);
+  auto font_file_path = argv[3];
+  auto output_file_path = argv[4];
+
+  error = FT_Init_FreeType(&library);
+  if (error) {
+    std::cerr << "FT_Init_FreeType\n";
+    return -1;
+  }
+
+  error = FT_New_Face(library, font_file_path, 0, &face);
+  if (error) {
+    std::cerr << "FT_New_Face\n";
+    return -1;
+  }
+
+  error = FT_Select_Size(face, 0);
+  if (error) {
+    std::cerr << "FT_Select_Size: " << FT_Error_String(error) << ' ' << error << '\n';
+    return -1;
+  }
+
+  assert(end > start);
+
+  uint32_t pixels_per_glyph = (face->size->metrics.height * face->size->metrics.max_advance);
+  assert(pixels_per_glyph % 8 == 0);
+  uint32_t bytes_per_glyph = pixels_per_glyph / 8;
+  uint32_t num_glyphs = (end - start) + 1;
+  uint8_t buf[bytes_per_glyph * num_glyphs];
+
+  uint32_t bitmap_offset = 0;
+  for (uint32_t char_code = start; char_code <= end; char_code++) {
+    int32_t bitmap_size = load_bitmap_char(face,
+					   char_code,
+					   &buf[bitmap_offset]);
+    if (bitmap_size < 0) {
+      std::cerr << "load_bitmap_char error\n";
+      return -1;
+    }
+
+    bitmap_offset += bitmap_size;
+    assert(bitmap_offset < (sizeof (buf)));
+  }
+  std::cerr << "bitmap_offset: 0x"  << std::dec << bitmap_offset << '\n';
+
+  FILE * out = fopen(output_file_path, "w");
+  if (out == NULL) {
+    perror("fopen(w)");
+    return -1;
+  }
+  fwrite(reinterpret_cast<void*>(&buf), bitmap_offset, 1, out);
+  fclose(out);
+}