GEM-dev

gem-dev@lists.iem.at

2417 discussions

Codecs in Gem (again!) & ffmpeg
by B. Bogart 31 Aug '05

31 Aug '05

Hello all, sorry for the cross-posting, Back to the old codec issue here. I was hapiliy working with photo-jpg compressed quicktime files for the last couple years. Now I'm working on an installation that I'm aiming to make using only open-source software. My "orginal" footage is a sequence of 300,000 jpeg frames (80% quality) And all I need is a video file that matches the jpg quality of the frames, and most importantly is seekable and loadable in Gem. So what I want to do is create a quicktime file, using photo-jpeg as the codec. Simple eh? hmmm. My first try was: ffmpeg -i %d.jpg -sameq -vcodec ljpeg output.mov Which created the file fine (where, FYI, quicktime pro died miserably) but was very big (which is ok) but I could not get vlc,mplayer,totem to load it. ffmpeg printed something about "MS style codec tag" when creating the file. so I also tried: ffmpeg -i %d.jpg -sameq -vcodec mjpeg output.mov which did about the same thing, so I gave up on quicktime and tried a couple AVIs: ffmpeg -i %d.jpg -sameq -vcodec mjpeg output.avi ffmpeg -i %d.jpg -sameq -vcodec ljpeg output.avi A little better, many of my Gem decoders loaded the file, but not a single one was able to make it seekable, I tried all the decoders... all return -1 frames. So right now I'm building a mpeg2 to try out, though I know decoding is going to be less efficient. So my final questions are: 1. What codec should I use for 300,000 frames (14GB or so) and is well suited to build from jpg frames? What container? 2. How can I create the [answer to #1] using ffmpeg or some other linux land app? Thank you! Oh and Johannes, I do have the v4l_compat module loaded, so I don't know why xawtv reads my v4l2 video stream but Gem does not. Any ideas? <B>

3 9

[Fwd: Bug#325567: gem: FTBFS (amd64): 'quicktime_supported_video' was not declared in this scope]
by IOhannes m zmoelnig 30 Aug '05

30 Aug '05

(oops, seems like i have sent this email originally to gem-dev-admin instead of gem-dev) Guenter Geiger wrote: > > Hi, > > Thanks for the patch. I am fighting with a changed API for ffmpeg on > unstable though. Did you compile on testing ? > for quicktime, neither "quicktime/quicktime.h" nor "lqt/quicktime.h" should be included but plain "quicktime.h" (and the full include-path returned by lqt-config (or "pkg-config libquicktime" (only available on unstable i think) should be added to the INCLUDES) as for PIC: is still have no real idea how to handle this upstream; currently i use/suggest something like "PKG_FFMPEG_LIBS=`ffmpeg-config --plugin-libs avformat ./configure" but i am not sure, whether this hack is suitable for a distribution. as for the ffmpeg-api changes, in upstream filmFFMPEG there are some adaptions to this. just my 2cents. mfg.asd.r IOhannes

1 0

vertex_tabread test patch?
by james tittle 26 Aug '05

26 Aug '05

heya, ...I was trying to play with this today and found out I don't know how to use it :-\ Am I right in reading that it will take multiple tables and interpret the first as vertices, color, normal, then texcoords? Basically I'm picking it up because I wanted to add some code I found on creating normals from "unstructured lists of polygons", and kinda finish up the shapeSynth... ...did you make a help/test patch? jamie

1 0

pix buffer improvements
by Georg Holzmann 26 Aug '05

26 Aug '05

Hallo gem devs! I need some fast changing images and so I added a open-message to pix_buffer_write, so that I can load an image into the buffer. I read the image like this: imageStruct *image = NULL; canvas_makefilename(getCanvas(), filename->s_name, buf, MAXPDSTRING); image = image2mem(buf); so my question is: do I need all the cache-stuff like in pix_image, or is a simple image2mem sufficient (like above) ? Now for the next: pix_buffer_read I just noticed, that pix_buffer_read needs much more cpu than pix_image, because stuff like this: Obj_header*ohead=(Obj_header*)pd_findbyclass(m_bindname, pix_buffer_class); [...] pix_buffer *buffer=(pix_buffer *)(ohead)->data; [..] img=buffer->getMess((int)m_frame); if (img && img->data) img->copy2ImageStruct(&state->image->image); is called every frame! So I'm thinking about, how to handle it in a more efficient way - maybe the best solution would be to check for a buffer and image only when the index changes ? Thanks, LG Georg -- Lust, ein paar Euro nebenbei zu verdienen? Ohne Kosten, ohne Risiko! Satte Provisionen für GMX Partner: http://www.gmx.net/de/go/partner

2 8

Re: [GEM-dev] pix buffer improvements
by Georg Holzmann 24 Aug '05

24 Aug '05

Hallo again! okay, I implemented now the stuff for pix_buffer_read and it performs much much better (as fast as pix_image). the changes: o) check for new buffer: at start rendering and if the buffer is changed with a message o) check for new image in the buffer: at start rendering, when buffer changes, when index changes, in auto-change-mode every frame of course The only situation, where the behaviour differs is, when a new image is loaded into a buffer - so you would have to retrigger the index now - is that okay? The new files are attached - if they are accepted I will also update the help patches ... LG Georg -- 5 GB Mailbox, 50 FreeSMS http://www.gmx.net/de/go/promail +++ GMX - die erste Adresse für Mail, Message, More +++

1 0

Funny crash in pd/Gem
by B. Bogart 24 Aug '05

24 Aug '05

Hey all, After running my patch for about 2-3 hours it will suddently exit. This could be related to some off issues I've been having with CPU usage, PD using up 90% when it should be using more like 20%. What confuses me is that all gdb returns is: pd_gui: pd process exited Couldn't get registers: No such process. (gdb) where Cannot fetch general-purpose registers for thread 1076267520: generic error Is this a watchdog kill? Anyhow I'll try and pinpoint the cpu usage issue and see if that is the cause. I'm running pd 0.38-4 on debian/sarge using the demudi packages. Thanks, B>

3 3

pix_video on linux revisited
by B. Bogart 24 Aug '05

24 Aug '05

Hey Johannes, Ok, so my video capture cards arrived today. They are leadtek winfast tv2000 xp expert cards. They are found by my 2.6.8 kernel and I can see video fine using xawtv. Oh and this kernel seems to be creating v4l2 devices, rather than v4l... is v4l2 backwards compatible? All the good news, onto the bad... tested on both gem 0.90 and gem CVS I am unable to see video. A simple pix_video patch loads fine but when I press "create, 1" to the gemwin I just get a black gem-window and PD freezes, no crash, no high cpu usage, the GUI just sits there unresponsive (as if the pd engine has crashed). In gem CVS I get this in the console when opening the patch: video4linux ieee1394 for linux Then I hit render and I get: pix_videoNEW: starting transfer GEM: pix_video: Opened video connection F Then PD is frozen. In Gem 0.90 I get this message when opening the patch: video4linux Then I hit render and I get: pix_videoNEW: starting transfer cap: name Leadtek Winfast 2000XP Expert type 7 channels 3 maxw 640 maxh 480 minw 48 minh 32 picture: brightness 32896 depth 26222 palette 4 channel 0 name Television type 1 flags 1 channel 1 name Composite1 type 2 flags 0 channel 2 name S-Video type 2 flags 0 setting to channel 1 buffer size 1867776, frames 8, offset 0 233472 wanted format is 0x1908 setting cmcapture to 64x64 5 frame 0 0, format 5, width 64, height 64 GEM: pix_video: Opened video connection F vmmap 8371BA8 Then the same PD freeze happens. As a reminder this is debian/sarge running demudi packages and kernel 2.6.8 PD 0.38-4 Any ideas Johannes? Sometimes I wish PD would crash so at least I can see what is happening! I'm starting to feel uneasy about my plan to use linux for this installation for "stability" and freedom. *sigh* I'd love to make it work... B>

1 0

texture coordinate differ on linux and OSX
by B. Bogart 23 Aug '05

23 Aug '05

Hey all, I spent a good afternoon figuring out how to create the illusion of one texture spead over multiple geos. (have not looked at the vertex thing, since I have been unable to compile vertex's on linux yet) So we figured it out on the laptop, all well and good, rectangular texture (video), "mode 1" where the coords range from 0 to the size of the texture in pixels. so we figured out all that was need to create the geo tile, moved over the linux (the machine the installation will run on), and bam, the pattern does not fit... :( Looks like textures range from 0-1 even in rectangular mode on nvidia/linux. The 0-1 coordes are "smart" in that "1" does actually stand for the width or height and so the message of "0 1 1 1 1 0 0 0" works to perfectly texture a 640x480 texture on on square. Is this due to openGL implimentation? Its pretty anyoying the complex pix_coordinate patches are not cross-platform... Any way to make it consistant? At first it was much easier to figure it out with 0-pixel size cordinates, but that is relative to the texture size, whereas the way its working on linux/nvidia the texture size does not come into play, not even the aspect ratio.. The anoyance of dipping way into small floats to spread one texture over multiple geos is a little ugly.. 1/320 is the width of the texture spead over 320 goes? Ok, Thanks all for bearing with me. You'll probably hear from me again this week. b>

4 3

Just a simple question (for a change)
by B. Bogart 23 Aug '05

23 Aug '05

Hi all, So I have a prototype system with a dynamic array of squares that clip/offset the texture to create the illustion of one texture on all squares. I tried making many squares (40x30) and that crashed pd. (my dynamic patcher did that is). So in order to simulate some kind of softness with fewer squares I was thinking about using an alpha channel for each square? So the question is: Can I apply the RGB texture using certain texture coordinates, and then apply an Alpha texture, using different pix_coordinate settings? Since I would want the alpha channel to clamp to the edges of the geo and not slide off as the RGB texture does... I think its impossible since the RGB and A are considered the same "texture". Also I realized I think the video source will be YUV anyhow... Say, does v4l provide YUV or RGB? I'll make a vertex array version also, but that would have the same issues with hard-edges and textures... Any brilliant ideas about making the geos soft? Thanks, B.

2 1

pix_video CVS broken?? (gem 0.90 build fails)
by B. Bogart 22 Aug '05

22 Aug '05

Hey all, I thought I would try a little experiment with the computer controlled cmaera today, but alas so many issues. The plan was to use Gem to capture single jpeg frames at each state of camera movement. The problem is I can't get pix_video to show anything! There are 4 video4l devices on this machine, I'm using /dev/video0 xawtv sees the video fine if I choose NTSC, and composite. No matter what I send to pix_video I can't get it to texture anything. I spent the most time with the "mode ntsc $1" messages w/out luck. Is anyone using pix_video with a video4linux device (this is a bt848) with gem CVS? Also I could not compile Gem 0.90 from gem.iem.at because it fails to find GL on this system! (a fedora core2 I did not set up myself..) I just installed the nvidia driver today. I'll try copying the pix_video src from 0.90 into my CVS src I guess. Thanks B.

2 4

Jump to page:

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

GEM-dev