Comment #1 and Comment #2 were entered on the wrong bug.

This can be fixed by an update to TTS. See b/73660996 internally.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromiumos/platform/assets/+/a9d303a3cc498799f16050f97b7a27b5cb8d2fea

commit a9d303a3cc498799f16050f97b7a27b5cb8d2fea
Author: Katie D <katie@chromium.org>
Date: Wed Mar 07 04:18:59 2018

Updates the Chrome OS TTS engine.

This fixes a bug where unknown unicode start characters caused
the 'start' events not to be fired.

See b/73660996 for updates to the .nexe.zip files and more
background.

TEST=Tried TTS Debug extension and some manual tests with Select-
to-Speak on Chrome OS on linux, x86 architecture (snappy), and
arm architecture (peach pit). Other than a large delay before
starting speech (2-3 seconds), everything seemed fine. I believe
the delay was in the last version as well, nothing new.

BUG=chromium:813870

Change-Id: I22f0b714ba17f7c2aa70cea5b7190d2f1d8e193d
Reviewed-on: https://chromium-review.googlesource.com/950314
Commit-Ready: Katie Dektar <katie@chromium.org>
Tested-by: Katie Dektar <katie@chromium.org>
Reviewed-by: Dominic Mazzoni <dmazzoni@chromium.org>

[modify] https://crrev.com/a9d303a3cc498799f16050f97b7a27b5cb8d2fea/speech_synthesis/patts/tts_service_x86_64.nexe.zip
[modify] https://crrev.com/a9d303a3cc498799f16050f97b7a27b5cb8d2fea/speech_synthesis/patts/tts_controller.js
[modify] https://crrev.com/a9d303a3cc498799f16050f97b7a27b5cb8d2fea/speech_synthesis/patts/tts_service_x86_32.nexe.zip
[modify] https://crrev.com/a9d303a3cc498799f16050f97b7a27b5cb8d2fea/speech_synthesis/patts/manifest.json
[modify] https://crrev.com/a9d303a3cc498799f16050f97b7a27b5cb8d2fea/speech_synthesis/patts/tts_service_arm.nexe.zip

Next step is that bullets are nodes in the tree which need to be skipped if speech continues after them. As a test, use a simple unordered list.

 Issue 832533  has been merged into this issue.

To get this to work, we'd really want a word end character index from the TTS engine in addition to the start character.

Alternate: move the startChar to the first spoken character:
(hi) had a startChar of 1 instead of 0
((((hi)))) had a startChar of 4 instead of 0

Moving back to "available", we need additional TTS work to fix this completely.

This bug requires a change to TTS to be completely fixed.