TextReadsFast for YouTube 作者： ravando

Read YouTube subtitles one word at a time, at a fixed focal point, so you can
follow along at 2x or 3x.

Reading is slow mostly because your eyes move. Roughly every few characters
they jump to a new fixation point, and see almost nothing during the jump.
RSVP (Rapid Serial Visual Presentation) removes those jumps by holding the
text still and moving the words instead.

TextReadsFast pins each word so its ORP (Optimal Recognition Point), the letter
slightly left of centre that your brain uses to identify a word, sits at the
same screen position every time. Your eye finds it once and never has to move
again.

The pace comes from YouTube's own playback rate. Set the video to 2x and the
reader tracks it exactly, because it reads from the video clock rather than
from a timer of its own; so pausing, seeking and changing speed all work.

SIX MODES
- RSVP: one word, pinned, with the pivot letter picked out
- RSVP + Bionic: the same, with the words either side bolded so you can read ahead without moving your eye off the pivot
- Bionic: a line at once, leading letters emboldened
- Plain: a line with nothing marked at all, the quietest of them
- Highlighter: a block behind the word being spoken
- Focus line: one word, plainly centred

Any of the line modes can slide, advancing a word at a time, or hold still
showing a whole line and swapping once at a sentence boundary. That is around
twenty changes a minute instead of three hundred, and it is far easier on the
eyes over a long video.

READ MODE
A study view for the video you are already watching: chapters on the left,
your timestamped notes under the video, and an optional AI assistant that
summarises the transcript and answers questions about it. Export everything to
a single Markdown file: an Obsidian vault, for instance.

PRIVACY & AI ASSISTANT
The reader sends nothing anywhere. Captions are read from YouTube, on pages
you already opened, and settings stay in your browser.

The AI assistant is opt-in and uses YOUR OWN API key. Until you add one,
nothing is transmitted. Once you do, transcripts are sent to the provider you
chose (OpenRouter by default) and to nobody else, there is no server behind
this extension. Your key is stored locally and is never synced.

No analytics. No telemetry. No account. No ads. Open source (MIT).

REQUIRES CAPTIONS
Videos without captions do nothing — there is no transcription here, so there
is nothing to fall back on. Auto-generated captions work best, because they
carry per-word timing.

Not affiliated with, endorsed by, or sponsored by YouTube or Google.