Project Threepio (Star Wars OOT subtitles) — Page 58

PM sent. You’re correct that the subtitles included with preservations are typically outdated and don’t include the README or the preferred SUP format subtitles.

I just lost my phone, in which it contained the files for the ANH Tagalog translation Project. Please wish me luck to help me find my phone again.

DeathStar314 said:

Can I also please have a temporary download link?

PM sent.

Dat_SW_Guy said:

I just lost my phone, in which it contained the files for the ANH Tagalog translation Project. Please wish me luck to help me find my phone again.

Oh crap. Good luck. And I’m off to make a backup of my own stuff…

Just continuing to go down the line of absolutely everything in this project to see if it could be improved for the next release, and I hit upon the font resource called Droid Sans Fallback which has been included with this project since we went Unicode. It’s basically a Unicode fallback font I kept in my back pocket for new languages. I never actually used it, but upon testing, it actually does a pretty dreadful job, so it’s going away.

So I’ve chosen a new Unicode fallback font for the next version, a customized version of the Noto Sans font (which is actually Google’s successor to Droid Sans) merged into a single file. The CJK flavor I’ve chosen is traditional Chinese, simply because I’ve already got all the major CJK languages covered with their own fonts, and any future CJK language support would have to be some regional Chinese dialect, which tend to favor traditional Chinese characters. Will I actually use this font? Probably not – but we could get a chance with the Tagalog subtitles, assuming there’s not a preferred font for Tagalog. But those who are on the lookout for a font covering a really wide range of the Unicode spec could do a lot worse.

Bah! Shows what I know, modern Tagalog uses Latin characters. Nevertheless, we have some very plausible scenarios where we could end up with a subtitle in a new script (Tamil, just by transcribing the dubs, would be pretty quick work). So we’ll have an improved backup font in 10.1 to cover any eventuality. Over 60,000 characters at this point, covering nearly all of the language groups in the BMP, and a smattering in the SMP. And I’ll be incorporating ConScript Unicode Registry glyphs so people can do subtitles in Sindarin or Klingon, because why not?

PM sent.

PM sent.

PM sent!

Any font geeks care to comment? I’m considering moving some subtitles to using Noto Sans Medium. At least on the Latin/Greek/Cyrillic side of things, two things that have always bugged me about Arial are that it’s a little on the thin side (but Bold is too bold, and the demibold variants I’ve found don’t have all the needed characters), and the damned square periods and dots on the I’s and J’s. But thicker letters and rounded dots may run up against some sort of more serious kerning issue, which is something I’ve always been pretty satisfied with on the Arial side. I’m also fine with the “even blander than Helvetica if that’s even possible” criticism of Arial, and Noto Sans is similar. I favor utilitarian subtitles that don’t draw attention to themselves in any way at all.

I may even use Noto Sans as a “one font to rule them all” option, but that’s unlikely at this point. I’d have to validate every script looks as good or better before switching. On the downside, Noto Sans isn’t available at medium weight in all scripts, and its normal weight is even thinner than Arial.

PM sent.

On the font front, this has taken a bit of a left turn. I’d originally been looking at the Noto Sans family of fonts as a means of creating some sort of single universal/fallback font. Well, that plan is dead. The font editing software I have (FontForge) can’t output the result of merged fonts without losing multiple anchor points (which destroys Thai and Hindi and makes me very nervous about the rest), so these separate Noto fonts need to stay separate in order to be usable.

Nevertheless I’m really growing to like these fonts – particularly the variant called “Noto Sans SemiCondensed SemiBold”, which really seems to have ideal proportions for subtitling IMO. Here is a very quick and dirty mockup showing Arial (current) on top, Noto Sans SemiCondensed Medium in the middle, and Noto Sans SemiCondensed SemiBold on the bottom. And no, that’s not even a real subtitle for what Han’s saying. You can click the image for a larger image.

The main differences are:

• Stroke thickness. I find thicker subtitles (within reason) to be more readable, but it can have side-effects, such as seeming bigger and brighter at the same point size.
• Letter width. Arial is a little wider than Noto SemiCondensed. I’m not sure it makes a difference readability-wise.
• Color. To counteract the fact that thicker fonts appear brighter – sometimes too bright – at the same color as a thinner font, I’ve chosen to make the new fonts slightly dimmer (still effectively white unless you break out the Photoshop, we’re not talking about going back to the days of yellow DVD subs). You’ll see that the middle Noto font seems approximately the same brightness as the top, but the bottom seems much brighter. In fact, the bottom two are the same brightness, both slightly dimmer than the top.
• Consistency between scripts. Right now we use a modernish utilitarian style for Latin/Cyrillic/Greek, but a more traditional style for Arabic and Thai. If we went all-in on Noto Sans, everything would be more stylistically similar.

Things to look for: the new subs might seem bigger and brighter–but I don’t want them to seem too big or bright. i.e. Would we need to back off the boldness and go back to medium weight? Reduce the brightness a bit more? Something else? Is there a better type of mockup which would be more useful?

A little more info on the font issue:

One of the things I’d been thinking about going into this was how in one of my many PM discussions over the years, someone said they preferred to use Project Threepio’s “matching” English subtitles. That’s fine and all, but the reason they preferred them was that they felt they were easier to read–they weren’t really that interested in the matching aspect. And he was right–the matching subs were both larger and thicker than the standard English subs.

With that in mind, looking up at the bottom of the three examples in the image I just posted, that font starts looking… familiar. Sure, it’s not an exact match, but proportion-wise, wow, I pretty much just picked a close match for the Greedo font, didn’t I? Of course I did just ditch the square periods for round ones, but nobody’s perfect.

I don’t plan on changing the height of the fonts very much from what they’ve been. I tested that original size on various real-world (read: small, crappy) TV setups, and it worked, and it also seems to roughly match what you see on commercial disks. As for nice wall-sized home theatre setups, that’s what the scale_subs script is for. The newer font is narrower, so the letters seem smaller at approximately the same height, but it’s also thicker, so I’d call it a wash (other opinions welcome).

As for things I’m still wrestling with, there’s the CJK fonts. Those are not condensed fonts, so they’ll look wider, and you’ll see that especially with Cantonese, where there are lots of Latin characters mixed in. The sans-serif fonts are also not currently available in semibold, so I’ve opted for medium – semibold was about as thick as I was willing to go. And for some crazy reason, Pango isn’t able to render any of these CJK fonts on Windows. Which is fine – for the first few years, I couldn’t render subtitles on anything but Windows, so I guess it’s only fair for Linux to get an exclusive feature. Unless I figure out how to fix it.

Then there’s SDH. If I make the text easier to read, does that obviate the need for the black background? Very few commercial disks use it. But I don’t want to do anything to compromise intelligibility for our deaf viewers, so maybe not.

I think that’s it. I’m increasingly sold on the concept of using Noto Sans for every single subtitle. Medium for CJK, SemiCondensed SemiBold for everything else.

Another mockup of the plans for the new fonts, a little closer to the final product:

TOP: Current.
BOTTOM: New (in progress)

A couple things to notice here. The new font is definitely your more modern style of Arabic text, Arial had a very calligraphic appearance. Because that traditional font looked harder to read (IMO, but what do I know about reading Arabic?), I also made it bigger. The new subtitles will be more-or-less the same size for all languages, so some languages (particularly Arabic) will see noticeably smaller subs. English will stay about the same height as before, just a little narrower.

Also look at the color. Pure white subtitles can be blinding when letters are thick enough. If you really look at the subtitles in the top frame, they really seem to glow. In the bottom frame, I took the brightness down a touch. It’s still effectively white, just without that intensity of pure 0xFFFFFF. The new subtitles will actually be a little less bright across the board.

So far, I haven’t tried Inkscape. Should ImageMagick+Pango hit a limit, I just may.

Well, I’ve been messing with this so long I can’t tell if this is a nice compromise or if I split the proverbial baby. Oh well, here’s a mockup of the new SDH look, current on top, new on bottom:

There’s no doubt I like the new font. In fact, the more I see the new font, the crappier and more amateurish the Arial font looks IMO. Dang.

Something more than the normal contrast seems likely helpful for SDH. If you’re deaf and can’t read every single subtitle quickly and easily, that’s pretty bad, so I want to avoid that. You’ll see that in addition to the semitransparent block, there’s still a thin opaque black border around the text which is useful against very bright backgrounds. I think it’s an improvement. Does it meet high readability standards? Is it overkill? At least I’ve seen subtitles that look like this before, so I know I’m not the only one. I’m inclined to think it’s a little too transparent right now, but that may be only because I’ve been staring at it too long.

DVD downscales will have to use solid black, however, as DVD subtitles have 1-bit transparency.

EDIT: They’re clickable images now so you can see more detail. And yes, the new subtitles are technically dimmer than the originals in terms of RGB values – same as the Arabic example above. That shows how much letter thickness enters into perceived brightness.

Well, I’m one who did prefer the ‘Matching’ English subtitles in my copies
of the films. What you are doing here looks good to me, I like the wider stroke
and the less bright look of the last few screens you’ve posted here.
Keep it up, and I’ll have to grab the latest version when you are done!

That’s good to hear. And see? Three years later or whatever, I do listen 😉

Really like the more transparent look for the subtitles - nice one.

Plus, as a UK’er, nice to not see the old arial type face - too many very poor subtitle companies use it for their news bulletins and etc - it’s become a bit of a standard bearer for bad / comedy subtitling! 😃

I edited the SDH mockup above with a transparency level I prefer (slightly more opaque).