Was just about to say that. Beyond that, it's pretty cool though so far.

The minimap is, in fact, the best idea anyone has given me today (and I think it addresses a concern 3 other people brought up in this thread). The underlying problem with all the suggestions is that you lose all the scaffolding, the paragraph breaks, the line spacing, the sense of where you are in the text that makes it easy to just jump into the text anywhere in the first place. I think this peripheral low fidelity map addresses it because you don't have to actively engage with it. Your concern about the control to not draw attention from the text is the entire point of this design. You can't make it too salient because that makes it steal the focus from the main text, which defeats the entire point.

I already have the info on paragraph and chapter breaks in the reader already, so this is more of a rendering issue rather than info display. The reason why I particularly like your first suggestion is that it lets you know something like a poem or different paragraph style is about to start so you can change your reading pattern accordingly rather than stumbling into it when you're already in prose mode reading at 400 wpm.

On the 2 modes thing, I feel like there isn't really a good reason to do it the other way. It was more that it happened that way rather than it needed to happen that way, and you're right that the main one should probably scroll and have bookmarks rather than making the user make a choice between 2 different interfaces.

Theme circles is definitely a solid suggestion, being a more practical implementation of visualising the underlying theme. Being able to see the actual colours of the background/text rather than an abstract blob is good, so that's a definite maybe.

Code isn't public yet, but extensions are something I've wanted to try for a while. In fact, someone sent me an email this morning asking precisely for that, to be able to paste a link and read articles. It can't be done directly in the page because of CORS policies, and doing it server side would involve routing the reading of articles through my server which is probably not ideal for everyone. An extension that pulls the article from whatever page you're on would be the most ideal version of this, from a privacy standpoint, so I'm considering it.

And finally, your last point is actually the first pushback I've had all day on the controls/interface. As far as 4 other people having difficulty finding the thing, I think you're right that something needs to be done, but I don't think hiding the options behind discovery is really the answer. There is a conflation in my brain between 'controlls are easily found and accessible' and 'controls are not hidden beneath layers of menus/bookmarks', and I think it's a false conflation, I had been treating them as one. Thanks for that.

That's the right answer, and it's pretty much what I’ve been considering since someone earlier in the thread asked about figures.

Epub makes it manageable because the images are just files in the zip. They are positioned at a set point in the spine, so I know exactly where in the word stream a figure belongs. Show it there, pause, and wait for a press to continue instead of auto advancing. You can pace text, but you can't pace a graph. There’s no sensible way to guess how long someone needs to view it.

The detail I keep reconsidering is captions. Right now, a caption flows into the stream as regular words, so you see "Figure 3 shows the relationship between" appearing one word at a time with no figure in sight. Those should be linked to the image and displayed alongside it instead.

Your question also made me realize that tables should probably be images too. Extracting text flattens a table into a series of values in reading order, which is worse than useless. Rendering the region as an image and pausing would be much better than what I do now.

Pdf is more challenging since I’d have to determine which images are real figures and which are rules and logos. Storage increases significantly too; everything is stored in IndexedDB on your device. So, a book full of plates is very different from a novel. I’ll probably need to downscale on import.

Where did you see it before? I’d really like to check out how they handled it.

Oh nice, the pre-processing hooks are a way cleaner way of solving the format issue than I was doing. When you are on the terminal you can shell out to pdftotext or pandoc and people use the format convertor that they already know and trust. But when you are writing stuff for the browser you have to include the format convertors in the download so I was including pdf.js, JSZip and mammoth in the build and still could not handle formats that I had not seen.

Your method also nicely solves that thing that I keep on getting asked for which is copy and paste a url and have it fetch the article and display the text. I had somebody email me about that today! On the terminal you just get something like: curl some_url | readable and you are pretty much done. On the browser it's a cross origin request so you will want to include a server that fetchies it and well that would kinda be missing the point of having local rendering.

Something that i was wondering is if monospaced text would have made it easier for you to find the alignment for the pivot? Fixed-width is simple, since your optimal offset is just a character offset. For proportional text i'm just taking my best effort pivot character and anchoring it, and then extending word fragments from that point in both directions. Otherwise the words can tend to wander as you eye flows from character to character.

Thanks, I'm glad the controls eventually clicked.

The saved toast is a reasonable hit and frankly is worse than annoying. The whole format is based on your eyes never having to leave the same area, so popping anything up next to the word display forces exactly the saccade you are designed to eliminate. It directly attacks the one function it is designed to fulfil. There’s no need for Auto-save to pop anything up at you either; it never received a request to save and there is no need to echo that action back at you.

Good call on Zen-mode, and I’m of the opinion that is what should be the default, once you start reading; when words begin moving make thechrome fade out, when words stop, make it reappear. I will consider both of these.

Was the control issue the Hold to Read option, out of interest? Just wondering what it was.

Yeah, the skimming thing surprised me too when I started looking into it. In regular reading, people skip about a quarter to a third of the words, mainly small ones like "the" and "of." Your eyes grab the next word or two out of the corner of your eye before you actually see them.

With RSVP, you save the eye movement, but you also lose the skipping. You have to look at every word, no matter if it’s important or not. Those two factors might balance each other out, which I think is a big reason the actual speed gain is much less than what most speed reading products promise.

The formatting issue you brought up is something I haven’t figured out. You miss line breaks, the shape of paragraphs, punctuation that sits in your peripheral vision, and even just keeping track of where you are on the page. I add longer pauses at punctuation to restore some of the sentence rhythm, but the structure of paragraphs and dialogue is still missing. That’s why it doesn’t work for me with fiction that has a lot of dialogue.

Thanks. Calibre’s crossed my mind a few times, the obvious hook is the content server, but the browser integration makes it clunky. Calibre-server doesn’t spit out CORS headers that you can use from another origin, so a web app can’t just read your library over HTTP unless you do a bit of reconfiguring it, and making people set CORS flags just for my little tool seems like a bad bargain.

The avenue that actually looks viable is directory import. Calibre keeps everything as normal files in Author/Title folders with a metadata.opf next to each book, so if I point a directory picker at your library directory, I can traverse it, grab the opf file to get the actual title/author rather than having to guess based on the filename, and then suck the epub files directly in. Would be a properly slick bulk import.

Catch is the File System Access API only works in Chromium at present, so anyone using Firefox/Safari would still be forced to pick their files one by one. Would be a worthwhile trade-off though as it gracefully degrades.

Thanks. The parsing cost is the entire one-time cost, the rest of the operation involves simply walking an array of words, a 900 page novel and a short story will perform the exact same. There is essentially zero cost in rendering a word at a time no matter the size of the book.

PDF is significantly the worst. We use pdf.js, and it has to page by page to retrieve the text layer, so a large book takes a little while and the tab feels 'busy' while doing so. EPUBs are considerably faster, it's mostly a case of unzipping and parsing xhtml.

The main thing that was the difference maker was separating storage into two IndexedDB stores, one for metadata, and one for the full text. When we open the library list we only read from the metadata store, not pulling the entire content of all books you own into memory. We only pull in the one you actually want to read.

The biggest edge case by far are scanned PDFs that have no text layer (there is no OCR, so nothing is returned) that are currently not failing audibly enough to make it clear.

Thanks. To be frank, the images and graphs are the weak spot at the moment. The parsers extract just the text layer and that’s the extent of it, so images are just discarded and you don’t even get told there was one.

Tables are worse than dropped - PDF extraction de-flattens them into a stream of values in reading order, so you just get a torrent of disconnected numbers flitting past which don’t mean a great deal.

I would read figures in the normal way for this type of book and then use the tool for the prose. It’s not really fixable at the format level too, there is no good way of showing a person a graph word by word. I suppose what I really should add is some kind of flag to say there was a figure here instead of just letting the process go ahead and drop everything. Otherwise, you can’t really know that you’re missing things.

Thanks, that's the detail I was most hoping someone would pick up on.

It started as a flat interval and felt terrible. Every word got the same slot, so sentences never landed and it read like a stock ticker.

Now there's a base duration from the WPM setting, with per word multipliers on top: 3x on sentence-ending punctuation, 2x on commas and semicolons, and an extra 0.7x if the word is 8 characters or longer.

The punctuation weights came from reading passages out loud and noticing where I actually stopped. Those are the breath points, so giving them proportionally more time puts the sentence structure back in. The long word bump is closer to the eye tracking literature, where fixation duration scales with word length. It felt wrong for a 12 letter word to get the same slot as "the".

The actual numbers were tuned by ear over a lot of real reading. I tried syllable estimates and word frequency weighting too, but neither was noticeably better than punctuation plus length, and they made the rhythm less predictable, which turned out to matter more than being clever.

E-ink is the frustrating part because it's where people actually read books, yet it's the display least capable of doing this. Partial refresh allows for a handful of updates each second, but it comes with ghosting. At 300 words per minute, which is a word every 200 milliseconds, you'd be reading through the residue of the last two or three words. Full refresh clears that, but it flashes the entire panel and takes long enough that you would only read about a word per second.

Your idea about sentences does affect the experience, but I don’t believe that makes it worse; it’s just a different approach. Once a whole sentence appears on screen, your eyes move across it again, so the no eye movement part is lost. What remains is the return sweep, the long jump back to the start of the next line, and the moment when you lose your place on a dense page. Those are real challenges in normal reading, and chunking by sentence removes both. You would end up closer to guided pacing than RSVP, which might suit a book better anyway.

Let me know how the procrastinated book turns out.