there is a missing subtitle on "00:21.12", the last image of the file.
It can help you to know that
on the black frames following this image the count becomes continuous,
no more void subtitle on frames mm:ss.01.

Could you send me your source file?

I am sending you 55MB by Wetransfer

• Sarah-iPhone30-to-Handbrake24fps.mp4
• All frames from 00m00s00i to 00m22s02i.srt

if at 00:03.06 (or any other TC) you click on the scrollbar
the TC and the sub is reset to 00:00.00, but the scroll button stays in place (see pict.)

Screen-Shot-2017-07-13-at-06.49.47.jpg (47,871 bytes)   

Screen-Shot-2017-07-13-at-06.49.47.jpg (47,871 bytes)   

Non-moving scroll button should be fixed by f0a9735731740bf9da24847294d9781f1bc3a331

Dynamic screen capture of 2.11.16 in PREVIEW : impossible to seek and fine tune a sub position.
(macOS 10.12.5)

Is this a unique bug or is it expressed by others that are open?

Carl, i am under time pressure right now,
I will answer your bug 'closing' qestions all in one
by making a summary with '42 next week;

First of all, do you think my srt frame indexing file is a valid one
in terms of ms in/out TCs ?
--jp

Hi JP, there is no rush so take your time. The subtitle file you sent me does have a fault: "00:00.00i start" has a in and out time of 0, and 00:00.01i has a start time of 0 and a end time of 1 (frame). So I think something is wrong with the creation of subtitles.

something is unclear, for me the millisec duration is OK,
proof is they work for the next second and all others.

do i have to start at 01.041 ?

1
00:00:00,000 --> 00:00:00,041
00:00.00i start

2
00:00:00,041 --> 00:00:00,083
00:00.01i

3
00:00:00,083 --> 00:00:00,124
00:00.02i

4
00:00:00,124 --> 00:00:00,166
00:00.03i

I attributed a TC lenght of 41 or 42 ms to each srt sub (1000/24= 41.66ms.
it works in Preview and in exported mp4 or Prores files.
but no way to make it work on the standalone Player ; so I tested 1ms TC gaps (or overlaps)
between subs, the result is NG.

These tests let me think that subtitle duration vs playback frame period is very tricky.
Could it be that a 24fps .srt file becomes not acceptable to the Standalone Player
because a minute frequency discrepency progressively makes the frontier be trespassed
and subtitles become impossible to sync ?

Subtitlewise, 2.11.49 is quite close to 2.12.0 :)
INTEROP metasubs are accurate and continuous*
While in pause, clicking on the timeline no longer sends back the player to the 00:00:00 start frame
but it makes the sub unpleasantly disappear ; you must click --> or <-- to see it back.

SMPTE metasubs still show a +1 frame error and disappear right after the 00:01.23 mark.

• there is no more void on hh:mm:ss.00 frames, provided there is a 2ms overlap
  between the subtitle TC and the frame count of these frames.

DoM-Player 11-49.jpg (57,620 bytes)   

DoM-Player 11-49.jpg (57,620 bytes)   

2.11.50 test
SMPTE metasubs are now accurate and show to the end of the file...
but only on mm:ss.00 frames (on which the TC overlaps 1ms ahead of the frame border),
so tomorrow I will make another test with a DCP built with a modified .srt in which I will
apply this 1ms offset to all frames.

BurnBlu-SoftOrange-Smpte-50.jpg (28,744 bytes)   

BurnBlu-SoftOrange-Smpte-50.jpg (28,744 bytes)   

• countingwise, 2.11.51 meta subs are perfect in both smpte and interop flavors.
• the volatiliry of the subs which needed a + or - 1 frame nudge to reappear after a jump
  on the timeline is now solved in smpte but not in interop.

I understand it is complicated to cater for all aspects of subtitle rendering including cueing. Ideally, DCP-o-matic would need to carry a state of all titles (search backwards and recue for upcoming subs). I guess Carl has good understanding of what a player should do to enable this, but the question is, is it compatible with the way the player works currently. For me, it would be sufficient if the player renders subs correctly during a normal playout. I understand that may be not enough for quality checks. However, subtitle files/timing should probably be qc'd in the subtitle application?

• Carsten

2.11.52 (Main & Player) : same interop subtitle volatiliy : needs [->] or [<-] for sub to reappear
after a time jump on the scrollbar.
Btw, would it be possible to also use the keyboard real arrows for +/- 1 frame scrubing?

I just filed a feature request for simple keyboard actions like play/pause, frame left/right.

I fear that constantly monitoring subtitle display status (including all display aspects like fade, etc.) could slow down the player considerably?

11a6decd652158b2288de17dc1aeafcd860fdfef should fix disappearing subs on seek.

is this "d860fdfef should fix disappearing subs on seek."
also works for DpM preview?

wink wink, say no more

Should do, it's the same code.

should do but doesn't, a great expectation is now down to earth ;-)

no worry, we will patiently wait for 2.13.x, your 2.12 standalone player is now
such a jewel to play with !

Are you looking at DCP content in the preview, or is it your video/srt combination?

It works for me in 2.11.55 three ways:

• in the standalone player - softsubs show up even while dragging the slider (if they are set long enough, of course ;-)
• in DCP-o-matic with the project loaded (e.g. with a subless DCP and an added srt file)
• in DCP-o-matic when I load the subtitled DCP

They also appear/disappear properly when nudging around their entry point.
It is not 100%, they don't always reappear when cueing backwards over their exit point. But most of the time ;-) The philosophical question is, is an exit point backwards still an exit point, or an entry point ;-)

That's in 2.11.55/Sierra. And no, luckily, there seems to be no performance penalty as I had feared.

• Carsten

If it's your Sarah-iPhone the problem is caused by I-frames being 12 seconds apart, so a seek to 11s ends up being to 0 and so there is no subtitle. I have an idea for how to fix this (0001204) but I think it's too disruptive to include in 2.12.0 at this late stage. I'll bump this to 2.13.x unless I see a nice way to do 0001204 in the mean time.

I'm fairly sure this stuff is resolved and/or moved to other bugs; do let me know if not.