Julien Bayle

Hi there,
Still in the middle of a big project for which I need to capture 3 webcam simultaneously (swap the one I record (only one at a time).

I found the Decklink Quad HDMI Recorder : https://www.blackmagicdesign.com/products/decklink/techspecs/W-DLK-36 

It can drive 4 cam. We'd use 3 Panasonic GH 5 & 4.

We'd need a very tight sync etc.
The PC would be a very big one (i9, 128Gb RAM, BIG RTX for the texture process + replay)

Would this setup the best from the acquisition point of view ?
I read Blackmagic drivers behave well in this kind of setup, with max.


Hi there,
Still in the middle of a big project for which I need to capture 3 webcam simultaneously (swap the one I record (only one at a ... 

multiple-cams-simultaneous-real-time-acquisition

Thanks for sharing with us, best regards for the personal .

Hello there, I still didn't have any chance to get some answers :-)

We are on it.
jit.dx.grab seems a way to follow VS jit.grab

The issue here is the latency.
- cams => Decklink Quad HDMI Recorder => OBS ====> VERY FAST display (not recording this)
- cams => Decklink Quad HDMI Recorder => Max 8 =====> feeling latency

I can't understand.
We are using a basic patch just grabbing frames with jit .grab, or jit.dx.grab as mentioned.
1080p 25fps (from nice panasonic cams)

By latency, I mean: there is a delay (I guess about 250ms, 300ms) 

Hey, 
I capture cameras all the time.
Just to get this straight..
You are capturing into the Decklink to record, is this then going live to screen?
Trying to figure out where the latency is noticed?
There will always be latency, no matter what you do.
Can you not record with OBS?  If you run the cams at 50 fps, you'll halve the latency.
If running to a live screen, I run everything at 50 fps, capture with Resolume (surprisingly fast) and send via spout to max for processing?
Let me know if I can help further.
B

Hi there.
I'm talking about the delay between a movement in real life and the frames showing the movement on the computer.

Frame reading seems smooth.
Frame rate at reading seems ok too.

But that delay...

I'm not sure adding resolume + syphon in the chain could help (but I obviously don't know as I didn't test myself) VS jit.grab or even jit.dx.grab

And I don't talk about recording. Just... Displaying frame.

I tested with a basic patch. Same.


Decklink here are just the hub. It collects 2 flows. And potentially 4.
I was thinking about creating a switching system for not grabbing all flow at the same time (as only one would be recorded at a time... But I was afraid of small overload when switching a jit.grab on and off on the fly)... But even with one 1080p25i/s, it is the same delay. 

The computer is very powerful, I talk about a i9, 128Gb RAM, one BIG NVIDIA RTX. 

We trying with resolume (and obs) as frameservers through spout in the equation.

Current research is about recording a BIG file, constantly, and to trying to find a way to play some part of it (during the recording) in order to grab the correct moment in time and to "offset/remove" the delay.


Hello Julien,
I did a quick test here concerning frame latency.
I sent timecode @25fps from my laptop into my PC using a Blackmagic Decklink Duo 2.
I wasn't able to acquire an image with jit.dx.grab with max? I tried evrything. 
Might help..?
B

possibly helpful discussion - https://forum.openframeworks.cc/t/how-to-reduce-latency-when-projection-mapping-onto-moving-objects/42464 

Multiple Cams simultaneous real-time acquisition ?

Noah Mitchell

I'm trying to figure out how to sample only select pixels within jit.gl.pix

Say I want to use an snorm scaled up to have a repeat effect, but I only want the to happen with a certain part of the image not to take the whole part of the image.

Can I repeat only a select part of the image using a sample parameter...or is there another simpler way to do this?

Hello,
I'm trying to figure out how to sample only select pixels within jit.gl.pix
Say I want to use an snorm scaled up to have a repeat eff

repeat-only-selected-region-of-an-image-jit-gl-pix

that's a cool patch...not exactly what I'm looking for though.

Here's an easier way to explain. Say I'm using a mask on an image and only want the masked part of the image repeated. In your patch the entire image is still getting repeated

Matteo's example does exactly what you are describing, you just need to adjust the rsliders to define the masked part of the image.

Hi there. These patches have been a real help. I'm keen to enact a mirror repeat on the cropped & zoomed area but I'm struggling with how the right inlet of 'sample' works. Any tips greatly appreciated.

the right inlet of [sample] takes a 2-component vector representing the normalized position (aka between 0 and 1) at which the image coming into the first inlet is sampled.
The attribute @boundmode affects the sampling coordinates ONLY when they exceed the range [0, 1]. 
Some examples:
given input sampling coordinates x = -0.1 and y = 1.3
@boundmode wrap (default) -> x = 0.9; y = 0.3
@boundmode clamp -> x = 0; y = 1
@boundmode mirror -> x = 0.1; y = 0.7

Since your patch samples a subregion of the original video, the sampling coordinates never exceed the range [0, 1]. 
If you want to have mirrored sampling, you must implement it manually. This could be a way to go:

Aha. This is starting to make much more sense. The second output is is a great help. Thanks so much for the response.

Repeat only selected region of an image jit.gl.pix

Tarik

I'm trying to create a custom shader for rendering jit.gl.text objects (in 2d mode). But there seems to be no way for my GLSL code to access the texture containing the text? 

How can I find the texture that's used by this object? I'd be happy also to just know what the name is of the internal texture object so I could project that onto a videoplane with my shader attached to it, as a workaround

Hello!
I'm trying to create a custom shader for rendering jit.gl.text objects (in 2d mode). But there seems to be no way for my GLSL code to

how-does-one-obtain-the-jit-gl-text-texture

Hi TARIK, is jit.gl.node @capture 1 an option for you, something like here:
https://cycling74.com/forums/how-to-make-those-stunning-videogame-glitches-!/replies/1#reply-5ce70f7771e603431be84946  ?

yeah I think capturing to a texture via jit.gl.node is your best option. the internal texture atlas containing the font glyphs is not something we can expose to the object.

Hello Tarik, did you manage to make a texture out of jit.gl.text? 
I need this patch in order to send it to syphon. I'm trying but I'm fully lost... Can some body help me? thx

I ended up going 'ah whatever' and just rendering and drawing the text every single frame, rather than storing anything in textures. But if you're generally wondering how to get textures out of max msp into syphon, you can download the syphon package (file->show package manager->search for 'syphon') and look at the help file for 'jit.gl.syphonserver'.
If your question is more about how to render a scene to texture, take a look at the 'jit.gl.node' helpfile.
I believe a combination of whats in those two helpfile (and perhaps that of 'jit.gl.text') should give you the info you need


How does one obtain the jit.gl.text texture?

Pedro Santos

Hi. I need to build thumbnails of images imported through a jit.movie to a fixed jit.pwindow (1:1 ratio for instance).
As the images may differ in aspect ratio from the jit.pwindow, is there an easy solution using matrix processing to generate the thumbnails while keeping its proper aspect ratio and adding horizontal or vertical black bars accordingly?

I'm not talking about jit.gl.videoplane, jit.gl.cornerpin or jit.gl.render manipulation, just straight matrix processing. Has anyone done it before?

aspect-ratio-correction-through-matrix-processing-no-opengl

I suggest taking a look at Jitter Tutorial 14: Matrix Positioning. You can use 'dstdimstart' and 'dstdimend' messages to a jit.matrix object to place a video within a desired submatrix of a matrix  (in your case, a square matrix).

hey Christopher, would it be possible with a jit.gl.texture as well? I tried to recreate it with that, but haven't succeeded yet. 
Cheers

Hey there,

You can try something like this:

Aspect ratio correction through matrix processing (no OpenGL)

Deliverer

Hi,
It's easy to get audio from a matrix using jit.peek~. Is there a similar way to achieve same result from a texture? or do i have to first convert jit.gl.texture into a jit.matrix? ... and then seeing a big drop in performance.

Hi,
It's easy to get audio from a matrix using jit.peek~. 

audio-from-jitgltexture

I would first downsample via jit.gl.texture to whatever smallest workable dims, and then use jit.gl.asyncread for the readback (knowing there will be a 1 frame delay in the readback).

audio from jit.gl.texture

Matt Raftman

In the past I used jit.qt.movie to export small snippets from large video files using the commands: newmovie, append, flatten and timecodes for the selection

but now with the new max version, jit.qt.movie is replaced by jit.movie + the avf engine and this method this does not work any longer

is there a similar way to do this with jit.movie ? or do i have to fall back on jit.record ?

I used jit.(qt).record earlier before i used the newmovie/append/flatten method, but i preferred the latter because it’s very elegant since it doesn’t recompress so its a lot quicker and has better image quality

Hi I have a problem with jit.movie
In the past I used jit.qt.movie to export small snippets from large video files using the commands: ... 

export-video-using-jitmovie-jitqtmovie

yeah all that functionality got chopped many years ago with the deprecation of quicktime and removal of 32 bit application support in macos.

My recommendation is to learn how to use ffmpeg and call into it from shell. chatgpt can be quite useful in this.

Ah yes I know ffmpeg, i also use it. Didn't make that connection, was thinking about it a bit too much inside of the box ;)
So u mean i can make Max generate a .sh batch script for ffmpeg right?  a very good tip, Thank you very much! 
(although its a bit saddening that on this point max functionality has become less rather than more.)


Export video using Jit.movie / Jit.qt.movie

Stefano Scarani

Well, I've just discovered that jit.playlist can not contain more than 14 FullHD files and play them; if  I exceed this number, Max console does not give any errors and  jit.playlist seems to work, but it does not output any video signal.

Max 7.3.4 @64bit / MacOS 10.12 / i7cuad 16 Gb

Well, I've just discovered that jit.playlist can not contain more than 14 FullHD files and play them; if  I exceed this number, Max console 

jitplaylist-number-of-file-limit

I know this is an old post - but is there no work around for this by running everything from gpu in any way? I'm encountering this issue now and it would make life easy if I could bypass it.

Jit.playlist number of file limit

Ullstein

Hi I need to mix more then 2 videos and so I thought I can use the Vizzie 4MIXR. But strange behaviour. Set up is simple at this point 2 x jit.movie go into 4MIXR IN 1 and 2, 4MIXR out connected to PROJECTR:
When 4MIXR is ON, I see an image but the movie is stopped and the fader has no effect, when 4MIXR is OFF, I see movie 1 running, as it is supposed to. Movie 2 is never visible.
What do I have to do to see a running movie 
MAC M1 Ultra OS 13.4.1

Hi I need to mix more then 2 videos and so I thought I can use the Vizzie 4MIXR. But strange behaviour. 

vizzie-4mixr-behaves-weird

vizzie keeps one bizzy or buzzy ....
this is reduced form of what that 4MIXR thing is doing


Many Thanks Source Audio, that works indeed. 
Have a nice summer

vizzie 4MIXR behaves weird

Riccardo Santalucia

I am having some trouble with the @parallax attribute of the [jit.gl.pbr] object.The problem I am facing is visible in the video trailer about this object published on youtube by Cycling '74, at 0:22 (link at the end)

Explaining it, it occurs that the edges of the object mantain a linear profile instead of showing, as they should, different heights and contours as a result of the heigtmap process.

The correct behaviour is visible in the second part of the video, from 1:15 on, where you clearly see that the shape do not mantain it's spheric nature.

Here the video: https://www.youtube.com/watch?v=YKNwqvuW8lQ

Hi there.
I am having some trouble with the @parallax attribute of the [jit.gl.pbr] object.The problem I am facing is visible in the video ... 

jitglpbr-unexpected-behaviour-of-the-parallax-attribute

this feature has not been implemented as it requires more drastic architectural changes than we initially anticipated.

I understand, I will then go on with [jit.gl.material]. 
Keep up the great work , I am looking forward for future implementations.

[jit.gl.pbr]: unexpected behaviour of the @parallax attribute

Joshua Knapp

I'd like to create a patch that takes a video and displays it on a video grid consisting of n items.
Each item, showing the initial video, should be slightly time delayed to its left item. If we think of it in terms of a matrix of dimension 2 x 2, the item in cell 0 0 would show the original video without delay. The item in cell 1 0 would be delayed by e.g. 50 ms. Relative to cell 0 0, cell 0 1 would be delayed by 100 ms and cell 1 1 by 150 ms.

I've tried using jit.glue, but it doesn't generate textures and is thus not able to perform well. Another disadvantage is that I was only able to create 7 or 8 rows and columns, if I go higher the video just disappears (even after resizing my jit.world render). Also, this approach requires me to patch the video matrix into n amounts of time delays and n amounts of jit.glue row and column inlets, which is way to much work for a high dimension matrix.

I'm looking for a solution that takes as little wiring as possible and runs efficiently with textures.
The number of grid rows and columns need to be actively adjustable and the rest of the patch should do the following calculations automatically. I'm sure this is somehow possible, so that the original video is resized to fit into "the first cell of a matrix", then create time delayed copies of that cell to the neighboring cells. 

Can somebody please guide me in the right direction?

Hello,
I'd like to create a patch that takes a video and displays it on a video grid consisting of n items.
Each item, showing the initial ... 

generating-a-video-grid

What you need is a video delay. 
The simplest way to accomplish your goal IMO is using a 3D texture accessed via shader.

At the moment, there's no other way to access and use 3D textures.
The main advantage of this approach is that the computational cost is independent of the grid size (try to crank up the "grid_size" value).

Another possibility is to use jit.gl.textureset, which is great for implementing video delay. However, since you need a video grid, you have to access the video buffer multiple times and at different locations, and probably jit.gl.textureset would be slow and/or cumbersome to use in the context.

Another possible approach could be using the object jit.gl.polymovie (see "multihead" tab in the help file), but again, it's hard to assign a different frame to each grid cell in a flexible manner.

I hope this helps,
Cheers!

I am stunned by how efficient your patch is and how you managed to do that with so little patching. Your solution is basically exactly how I imagined the program to work. Thank you so much for your time!

This leaves my brain a little less flooded with endorphins, as I didn't go through days of pain, trying to find a solution myself 😂. Nevertheless, I would be happy to use your patch as a working base to implement other functions like rotation. And also use it to render videos for non-commercial and maybe commercial purposes. Do I have your permission to do so?

Btw you are a beast (positively speaking), I've been following your patreon for some time now and your content is always educational and exciting. 

Thanks so much, Joshua!
I'm glad you enjoy my educational content.

Sure, feel free to use the patch however you prefer

Generating a Video Grid

Rob Ramirez

Happy post RNBO day everyone, I've got some details and context on what the Jitter team has cooked up for the 8.5 update.

Top of the list, GLCore (or glcore) and what it's all about. With 8.5 we've done 2 things:

Renamed the formerly beta gl3 engine to glcore

Transitioned our default graphics engine from the legacy gl2 to the modern glcore.

The reason for the renaming is to better reflect the reality that most graphics cards will support OpenGL versions beyond version 3. The engine provides support for modern core profile OpenGL at whatever version your graphics card supports.


For existing gl3 users not much has changed in functionality (other than the new juicy bits described below). However the consequences extend beyond the technology. We now have a solid foundation for continuing our engine modernization and growth, and can fully document the new behavior and features. To that end there are a handful of new examples to check out in the glcore Examples Extras patch.

For gl2 users this is a significant but hopefully seamless change. We will publish more details on how best to utilize modern graphics engines and other ins and outs, and we strongly encourage contacting support with any issues. However if you wish to revert back to gl2 simply change your Graphics Engine (gfxengine) preference back to gl2. You first might want to check out the afore mentioned glcore Examples patch to see what you're forgoing.


A new entry to our Jitter Tools suite is the jit.gl.textmult object, which allows for manipulating letters and words of text using a jit.gl.multiple like interface and highly efficient mesh rendering. As an added bonus for text lovers we’ve removed the 1024 character limit on jit.gl.text input strings.

Also included with 8.5 are some lovely new jit.gl.pass post processing effects. The atmospheric effect adds global fog and atmospheric scattering to any 3D scene with ease, and for experimenting with global illumination check out the temporal screen-space ambient occlusion (tssao-gi) and screen-space reflections (tssao-gi-ssr and ssr) effects.

Speaking of global illumination, the new jit.gl.environment object handles the loading and sharing of environment map resources for image based lighting (IBL), cubemap, and skybox rendering for supported objects. This currently includes jit.gl.pbr, jit.gl.material, and jit.gl.skybox. The object functions like jit.gl.light, where if a context has an environment attached, the supported objects will use it for rendering.

We've also added glTF PBR material support to jit.gl.model, giving much better results when rendering these model files. In the video below we see non-PBR rendering on the left, and PBR rendering on the right.

For links to the example patches mentioned (and loads more) check out the glcore Examples and Jitter Tools Extras patches, and the jit.gl.pass reference page.

glcore-extras.png

textmult.png

fog-fx.png

Happy post RNBO day everyone, I've got some details and context on what the Jitter team has cooked up for the 8.5 update.
Top of the list, ... 

jitter-8-5-and-glcore

Hey everyone, we've got an update for glcore included in the Jitter Tools 1.0.6 version now live in the Package Manager. This update addresses a handful of reports we've received since the release of 8.5 and is recommended for all 8.5.x glcore users.

fix move texture output when skybox enabled in same context

jit.gl.mesh - fix crash on certain adjacency draw modes

jit.gl.path - evalout equals zero drawing fix

jit.gl.text2d - fix screenmode 0 rendering with high dpi display

jit.gl.environment - reflection map edge length attribute

jit.gl.meshwarp - external input matrix for initial mesh geometry via mesh_matrix message

jit.gl.meshwarp - output_mesh_matrix message to output the mesh vertex data

Rob, congrats to you and c74 on glcore. I get the effort your team put into it. :-)
Are there C/++ examples for the updated interfacing possible HOST to/from GPU for textures, buffers, mapping between max,opencl,opengl, etc.

Unfortunately I don't have a great answer for you. The SDK has not been updated with examples for developing for glcore, and given the current state of flux in our backend graphics engine work, I don't have a good estimate for when that might change. 

That being said, both Syphon and Spout are open source and demonstrate how an external can support both legacy gl2 and modern glcore using some wrapper functions we expose in the jit.gl.draw.h header file. If you're only concerned with core profile support you can ignore these wrappers and link to the gl3 library headers instead (which may be easiest by defining GL3_VERSION=1 in your project preprocessor definitions). Additionally, the basic principles of jit.gl external development outlined in our SDK documentation still hold true.

I'm also more than happy to provide whatever insight I can for any brave explorers in this space.

Fantastic, thank you Rob and the team!

Do you have do you have any idea when TXAA / TAA is coming to Jitter!? Ref here at 16m20s https://youtu.be/S7xPbjICNsw 

I just posted the TAA pass fx here:
https://cycling74.com/forums/sharing-temporal-anti-aliasing-taa-fx-for-jit-gl-pass 

At the moment, it's quite complex to make it available as a built-in tool, but if you're interested in it, check out the patches at the link above

Thanks for the response; will check out the patches!

I am excited about the great jitter upgrade!

I'm working on some samples, so check out my twitter as well. 

https://twitter.com/mirrorboy/status/1662831859439644674?s=20
https://twitter.com/mirrorboy/status/1662699971462979585?s=20

I have a few questions as I use it.

Will the pbr in jit.gl.model be reflected without jit.gl.pbr connected? (I loaded the glTF and it seems to reflect the pbr textures without specifically connecting jit.gl.pbr or jit.gl.material). and, I could not change pbr parameter on jit.gl.model.

Also, if the texture loaded in jit.gl.model references a multi-texture by albedo, can I use the multi-texture in jit.gl.pbr when drawing with matrixoutput or jit.gl.mesh?

Jitter's material has many inheritance relations and parameters, and jit.gl.model is very difficult due to some invisible parts, so it would be great to have contents like 1st step.

Cycling '74 Forums

I am excited about the great jitter upgrade! I'm working on some samples, so check out my twitter as well.

Interesting read, thx Rob. B

Aha. This is starting to make much more sense. The second output is is a great help. Thanks so much for the response.

I ended up going 'ah whatever' and just rendering and drawing the text every single frame, rather than storing anything in textures.

Hey there, You can try something like this:

I would first downsample via jit.gl.texture to whatever smallest workable dims, and then use jit.gl.asyncread for the readback (knowing ...

Ah yes I know ffmpeg, i also use it.

Hi Stefano, I know this is an old post - but is there no work around for this by running everything from gpu in any way? I'm encountering ...

Many Thanks Source Audio, that works indeed. Have a nice summer HP

I understand, I will then go on with [jit.gl.material]. Keep up the great work , I am looking forward for future implementations.

Thanks so much, Joshua! I'm glad you enjoy my educational content. Sure, feel free to use the patch however you prefer