Back Fall; or, Dodge in Hell
Chapters (77)
  1. Title Page
  2. Contents
  3. Dedication
  4. Book 1
  5. Part 1
  6. Chapter 1
  7. Chapter 2
  8. Chapter 3
  9. Chapter 4
  10. Chapter 5
  11. Chapter 6
  12. Chapter 7
  13. Chapter 8
  14. Part 2
  15. Chapter 9
  16. Chapter 10
  17. Chapter 11
  18. Part 3
  19. Chapter 12
  20. Chapter 13
  21. Chapter 14
  22. Chapter 15
  23. Chapter 16
  24. Part 4
  25. Chapter 17
  26. Chapter 18
  27. Chapter 19
  28. Chapter 20
  29. Chapter 21
  30. Chapter 22
  31. Chapter 23
  32. Chapter 24
  33. Part 5
  34. Chapter 25
  35. Chapter 26
  36. Chapter 27
  37. Chapter 28
  38. Chapter 29
  39. Chapter 30
  40. Chapter 31
  41. Chapter 32
  42. Part 6
  43. Chapter 33
  44. Chapter 34
  45. Chapter 35
  46. Chapter 36
  47. Chapter 37
  48. Chapter 38
  49. Chapter 39
  50. Chapter 40
  51. Chapter 41
  52. Part 7
  53. Chapter 42
  54. Book 2
  55. Part 8
  56. Chapter 43
  57. Chapter 44
  58. Part 9
  59. Chapter 45
  60. Chapter 46
  61. Chapter 47
  62. Part 10
  63. Chapter 48
  64. Chapter 49
  65. Part 11
  66. Chapter 50
  67. Chapter 51
  68. Chapter 52
  69. Chapter 53
  70. Chapter 54
  71. Chapter 55
  72. Chapter 56
  73. Map of The Land
  74. Acknowledgments
  75. About The Author
  76. Also By Neal Stephenson
  77. Copyright

Corvallis had come to know that voice. Since their last face-to-face meeting in Zelrijk-Aalberg, it was the only voice in which Corvallis had heard Elmo Shepherd speak.

People had long since got in the habit of representing themselves, in virtual spaces, with avatars. Audio representation—the voice your avatar spoke in—had lagged somewhat behind. It had become important in T’Rain, the game that Richard and Pluto and Corvallis and others at Corporation 9592 had pioneered. It was a devilish problem that no programming team in its right mind would want to be saddled with, and so it had fallen into, or been elevated to the status of, weird stuff. It was all well and good to imagine how cool it would be if your dungeon-raiding team of wizards, dwarves, elves, and the like could all communicate freely in voices and accents that were movie-quality, as finely realized in their own way as the avatars, weapons, and environments. But when one of the players was Chinese and didn’t speak much English; and another was from rural Arkansas and making no effort to hide his accent; and another was from Boston but trying to fake the Scottish accent that he fancied a dwarf would speak in; and another was in London and acutely sensitive to how badly the guy from Boston was getting it wrong; and one of them was a woman playing a male character; and one of them was carrying on a running argument with his mom, who wanted him to take out the garbage; when all of that was happening at once, in real time, making those characters sound as good as they did in the movies was impossible. The Weird Stuff team had made some headway using a divide-and-conquer approach. They had figured out ways to make male voices sound female and vice versa. They had used speech-to-text systems to decode what the players were saying, then used AI algorithms to filter out the “Take out the garbage now or you are grounded” chatter, then used text-to-speech technology to re-render them with correct accents, and so on. That technology had made its way into applications beyond gaming.