Example of playing video on Kindle Touch

[geekmaster]

@Novo, that demo is for the K3 in 4 bpp mode. The touch is 8 bpp. I could adapt it...

The partial refresh hack on a touch could be similar to that video, but in the modes we can do now we are limited to about 5 frames per second and that looked a lot faster. The touch has other update modes defined in the eink drivers that I have not tried yet, so it may be possible to get the same results. For now though, similar but slower (like in the youtube video above).

The K3 has older eink drivers, but still may be able to be improved as we explore its capabilities using dithered black and white modes.

The eink drivers also support changing bits-per-pixel and support 1-bit pixels. Perhaps that would be even faster, but so far is just another thing on the "To Do" list...

[geekmaster]

I like this thread, so I just tried to bump ehhh's karma from 44 to 2600, but somebody else must have been giving him karma at the same time (or there was a technical glitch), so it jumped to 2810. Oh well, at least he no longer has noob karma.

[Novo]

any updates for now? Sourcecode?

[geekmaster]

Yeah, what's up with the "Big Tease"? Let's see some "product".

[Ehhh]

Ohhh... 3 week vacation in Egypt gave me a time to think faster... )))

ps. and i'm full of work of my real job, and that's why i've no a lot of free time to cotinue this hobby's project. i provided the investigations and developing, but it's mush slower then before.

[wolftail]

Originally Posted by Ehhh

ps. and i'm full of work of my real job, and that's why i've no a lot of free time to cotinue this hobby's project. i provided the investigations and developing, but it's mush slower then before.

Please release the code so that other people work on this too. Development will be a lot faster.

[Ehhh]

Hi, Geekmaster

How many frames per second could show your tests? I opimized code and developed a few tests today, and my eink display can shows only 3 frames per seconds without any artifacts.

Bests Regards

[geekmaster]

For the K5(touch) it depends mostly on high contrast transitions and how large they are. It analyzes the screen changes and does different things depending on how much of the screen changed since the last update. The most annoying is a deferred update mode, where it stacks many unfinished partial updates, then leaves a trail of up to 32 (an estimate based on observation) dirty rectangles trailing a moving high-contrast object.

If I do low-contrast animation where there are now high-contrast sharp edges, I can do aobut 12 FPS, but for normal content (allowing a little artifacting) I get about 5 FPS.

It helps to NOT write to /dev/fb0 until just before the eips command (or ioctl() if you got that working). I completely redesigned my dithering code, and in my new code. I am now doing all my updates to /tmp/wb0 (an 8-bit virtual framebuffer) at about 30 FPS in 256-color grayscale, then periodically (slow enough to not cause nasty eink display artifacts such as tearing or smearing) I dither that /dev/fb0. Depending on content I can update the display between 4 FPS and 12 FPS (or faster in some cases, with carefully controlled content).

Actually, it seems that the K4 booted from main is FASTER than the K5 for animation (using hybrid K3 eink drivers in 8-bit framebuffer mode).

I will post a little demo showing the maximum speed I have been able to achieve on the K5 using carefully controlled content. Of course, that only works with that special content, and I have to slow down to 4 or 5 FPS updates for normal high-contrast stuff to reduce display artifacts.

I will add a link here to my demo when I post it.

Here is my newest "dithermation" demo showing what my new dither routines can do:

https://www.mobileread.com/forums/sho....php?p=2054459

[geekmaster]

I just posted a long youtube video "Living FPV 2" that I downloaded and converted to "gmplay" format. For those who missed it, gmplay can play fullscreen videos on all eink kindle models, but the DX and DXG drop frames because they can only do 1.3 FPS (but the sequence of still images is sharper than continuous video). More videos have been posted now too, including great old vintage silent films.

The K3 and newer eink kindles can play full screen video at full speed. There are two videos you can download (one computer generated -- the other filmed by on onboard camera mounted in a remote-controlled airplane). "Action shots" like the FPV video show that video on eink is much better and more useful than I thought it would be. Source code is provided for video the player.

You can download gmplay and videos for it here:

https://www.mobileread.com/forums/sho...d.php?t=177455

[Ehhh]

einkplayer demo is uploaded... and it 3 cartoons are attached too.

[geekmaster]

Originally Posted by Ehhh

einkplayer demo is uploaded... and it 3 cartoons are attached too.

Hmm... No source code? How can I static link it like I do for my stuff so it will run on all kindles without library version conflicts?

I normally test in K5 diags first (less competition for resources), but your program just displays this:

./einkplayer: error while loading shared libraries: libgstreamer-0.10.so.0: cannot open shared object file: No such file or directory

I suppose I need to boot main and test it there because of its crippling dependence on libraries. Does this work on DX/DXG/K3/K4main/K4diags/K5main? It does not work on K5diags.

I test my stuff on all of the above (and it works).

When do we get the source code? We are still waiting.

In the mean time, I have ffmpeg running in all my kindles in all modes, playing video on the eink. It is a full build with all options and plays all supported video formats and network streams. I run all my eink stuff at 7.7 FPS (130ms/update), which is a good compromise for all the kindles. The DX and DXG run at 1.5 FPS though because of the time spent in the eink drivers. I have also run the K5 at up to almost 50 FPS which works fine for VERY carefully controlled content (no fast motion of high-contrast objects), and on eink, motion is fuild even at 7.7 FPS so that is a generally good all-around compromise.

It would be NICE if your code can give us less overhead than a full ffmpeg build...

UPDATE: I booted to K5main and ran it again. Running ./run_player.sh gave a bunch of different errors this time:

Spoiler:

'ips: invalid option -- '
: not foundr.sh: line 3:
: not foundr.sh: line 5:
com.lab126.powerd failed to set value for property preventScreenSaver (0x8 lipcErrNoSuchProperty)
: not foundr.sh: line 8:

Enter command:
1 : Play
2 : Stop
3 : Pause
4 : Seek
5 : Position/Duration
q : quit

(einkplayer:2064): GLib-GObject-CRITICAL **: g_object_set: assertion `G_IS_OBJECT (object)' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_element_link_pads: assertion `GST_IS_ELEMENT (dest)' failed

(einkplayer:2064): GLib-GObject-WARNING **: invalid (NULL) pointer instance

(einkplayer:2064): GLib-GObject-CRITICAL **: g_signal_connect_data: assertion `G_TYPE_CHECK_INSTANCE (instance)' failed

(einkplayer:2064): GLib-GObject-WARNING **: invalid (NULL) pointer instance

(einkplayer:2064): GLib-GObject-CRITICAL **: g_signal_connect_data: assertion `G_TYPE_CHECK_INSTANCE (instance)' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_element_get_static_pad: assertion `GST_IS_ELEMENT (element)' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_bin_add_many: assertion `GST_IS_ELEMENT (element_1)' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_element_link_many: assertion `GST_IS_ELEMENT (element_1)' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_ghost_pad_new: assertion `GST_IS_PAD (target)' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_element_add_pad: assertion `GST_IS_PAD (pad)' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_object_unref: assertion `object != NULL' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_element_get_static_pad: assertion `GST_IS_ELEMENT (element)' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_bin_add_many: assertion `GST_IS_ELEMENT (element_1)' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_element_link_many: assertion `GST_IS_ELEMENT (element_1)' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_ghost_pad_new: assertion `GST_IS_PAD (target)' failed

(einkplayer:2064): GStreamer-CRITICAL **: gst_element_add_pad: assertion `GST_IS_PAD (pad)' failed
Error: Internal data flow error.

EDIT: Problem solved -- the run_player.sh script contains carriage-returns that need to be stripped out. A new script with "unix" LF-only should be uploaded to replace it. This problem still exists in KTPlayer.tar.gz.

So... what kindles DOES it run on? Or perhaps I should ask "WHICH firmware versions on what kindles does it run on" ??? Just curious...

We could help you fix this, and add enhancements and features to it, if we had source code. Can we please have source code?

UPDATE: CRs inserted AFTER downloading, by extracting with 7-zip.

[thatworkshop]

Originally Posted by geekmaster

UPDATE: I booted to K5main and ran it again. Running ./run_player.sh gave a bunch of different errors this time:

I ran it without any errors at all.
Yes, having the source code would be very nice.

[geekmaster]

Originally Posted by cscat

I ran it without any errors at all.
Yes, having the source code would be very nice.

I am running 5.1.0. What firmware do you have? I want to see this run.

I am curious why lipc is reporting "preventScreenSaver" as "lipcNoSuchProperty". Another thing the 5.1.0 update broke?

@Ehhh: are you running 5.1.0, or does this only work on 5.0.x kindles? And those glibc errors look like some kind of library version conflict. Could you please upload a static build that contains your working libraries linked into the executable, so it can work on more firmware versions and kindle models? I would like to build it myself, so could I please have the source code (PM it to me if necessary to keep it private). Thanks.

I really do not have time to downgrade to 5.0.4 to test this. Has anybody run it on 5.1.0 yet?

[Ehhh]

Hi, Geekmaster

Firstly, thank you for testing it.
My Kindle's version is 5.0.4 . It seems that plugins of gstreamer which i compiled for Kindle Touch conflicts in version 5.1.0. Please, write me a size of glib library.
what version of gstreamer is on 5.1.0?

[knc1]

Originally Posted by Ehhh

Hi, Geekmaster

Firstly, thank you for testing it.
My Kindle's version is 5.0.4 . It seems that plugins of gstreamer which i compiled for Kindle Touch conflicts in version 5.1.0. Please, write me a size of glib library.
what version of gstreamer is on 5.1.0?

You will find members here slow and/or reluctant to help support closed source software.

If you would publish your source, with an appropriate open source license...