Video Texture (bge.texture)

Introduction

The bge.texture module allows you to manipulate textures during the game. Several sources for texture are possible: video files, image files, video capture, memory buffer, camera render or a mix of that. The video and image files can be loaded from the Internet using a URL instead of a file name. In addition, you can apply filters on the images before sending them to the GPU, allowing video effect: blue screen, color band, gray, normal map. bge.texture uses FFmpeg to load images and videos. All the formats and codecs that FFmpeg supports are supported by bge.texture, including but not limited to:

  • AVI
  • Ogg
  • Xvid
  • Theora
  • dv1394 camera
  • video4linux capture card (this includes many webcams)
  • videoForWindows capture card (this includes many webcams)
  • JPG

How it works

The principle is simple: first you identify a texture on an existing object using the materialID function, then you create a new texture with dynamic content and swap the two textures in the GPU.

The game engine is not aware of the substitution and continues to display the object as always, except that you are now in control of the texture.

When the texture object is deleted, the new texture is deleted and the old texture restored.

Game Preparation

Before you can use the bge.texture module, you must have objects with textures applied appropriately.

Imagine you want to have a television showing live broadcast programs in the game. You will create a television object and UV-apply a different texture at the place of the screen, for example tv.png. What this texture looks like is not important; probably you want to make it dark gray to simulate power-off state. When the television must be turned on, you create a dynamic texture from a video capture card and use it instead of tv.png: the TV screen will come to life.

You have two ways to define textures that bge.texture can grab:

  • Simple UV texture.
  • Blender material with image texture channel.

Because bge.texture works at texture level, it is compatible with all the Blender Game Engine’s fancy texturing features: GLSL, multi-texture, custom shaders, etc.

Examples

Example of how to replace a texture in game with a video. It needs to run everyframe.

we will use ``GameLogic.expandPath()`` to build an absolute file name,
assuming the video file is in the same directory as the blend-file.
"""
import bge
from bge import texture
from bge import logic

cont = logic.getCurrentController()
obj = cont.owner

# the creation of the texture must be done once: save the
# texture object in an attribute of bge.logic module makes it persistent
if not hasattr(logic, 'video'):

    # identify a static texture by name
    matID = texture.materialID(obj, 'IMvideo.png')

    # create a dynamic texture that will replace the static texture
    logic.video = texture.Texture(obj, matID)

    # define a source of image for the texture, here a movie
    movie = logic.expandPath('//trailer_400p.ogg')
    logic.video.source = texture.VideoFFmpeg(movie)
    logic.video.source.scale = True

    # Note that we can change the ``Texture`` source at any time.
    # Suppose we want to switch between two movies during the game:
    logic.mySources[0] = texture.VideoFFmpeg('movie1.avi')
    logic.mySources[1] = texture.VideoFFmpeg('movie2.avi')

    # And then assign (and reassign) the source during the game
    logic.video.source = logic.mySources[movieSel]

    # quick off the movie, but it wont play in the background
    logic.video.source.play()


# Video playback is not a background process: it happens only when we refresh the texture.
# So you need to call this function every frame to ensure update of the texture.
logic.video.refresh(True)

Example of how to replace a texture in game with an external image. createTexture() and removeTexture() are to be called from a

"""
from bge import logic
from bge import texture


def createTexture(cont):
    """Create a new Dynamic Texture"""
    obj = cont.owner

    # get the reference pointer (ID) of the internal texture
    ID = texture.materialID(obj, 'IMoriginal.png')

    # create a texture object
    object_texture = texture.Texture(obj, ID)

    # create a new source with an external image
    url = logic.expandPath("//newtexture.jpg")
    new_source = texture.ImageFFmpeg(url)

    # the texture has to be stored in a permanent Python object
    logic.texture = object_texture

    # update/replace the texture
    logic.texture.source = new_source
    logic.texture.refresh(False)


def removeTexture(cont):
    """Delete the Dynamic Texture, reversing back the final to its original state."""
    try:
        del logic.texture
    except:
        pass

Video frames captured with DeckLink cards have pixel formats that are generally not directly usable by OpenGL, they must be processed by a shader. The three shaders presented here should cover all common video capture cases.

This file reflects the current video transfer method implemented in the Decklink module:
whenever possible the video images are transferred as float texture because this is more
compatible with GPUs. Of course, only the pixel formats that have a correspondant GL format
can be transferred as float. Look for fg_shaders in this file for an exhaustive list.

Other pixel formats will be transferred as 32 bits integer red-channel texture but this
won't work with certain GPU (Intel GMA); the corresponding shaders are not shown here.
However, it should not be necessary to use any of them as the list below covers all practical
cases of video capture with all types of Decklink product.

In other words, only use one of the pixel format below and you will be fine. Note that depending
on the video stream, only certain pixel formats will be allowed (others will throw an exception).
For example, to capture a PAL video stream, you must use one of the YUV formats.

To find which pixel format is suitable for a particular video stream, use the 'Media Express'
utility that comes with the Decklink software : if you see the video in the 'Log and Capture'
Window, you have selected the right pixel format and you can use the same in Blender.

Notes: * these shaders only decode the RGB channel and set the alpha channel to a fixed
value (look for color.a = ). It's up to you to add postprocessing to the color.
       * these shaders are compatible with 2D and 3D video stream
"""
import bge
from bge import logic
from bge import texture as vt

# The default vertex shader, because we need one
#
VertexShader = """
#version 130
   void main()
   {
      gl_Position = gl_ModelViewProjectionMatrix * gl_Vertex;
      gl_TexCoord[0] = gl_MultiTexCoord0;
   }

"""

# For use with RGB video stream: the pixel is directly usable
#
FragmentShader_R10l = """
    #version 130
    uniform sampler2D tex;
    // stereo = 1.0 if 2D image, =0.5 if 3D (left eye below, right eye above)
    uniform float stereo;
    // eye = 0.0 for the left eye, 0.5 for the right eye
    uniform float eye;

    void main(void)
    {
        vec4 color;
        float tx, ty;
        tx = gl_TexCoord[0].x;
        ty = eye+gl_TexCoord[0].y*stereo;
        color = texture(tex, vec2(tx,ty));
        color.a = 0.7;
        gl_FragColor = color;
    }
"""

# For use with YUV video stream
#
FragmentShader_2vuy = """
    #version 130
    uniform sampler2D tex;
    // stereo = 1.0 if 2D image, =0.5 if 3D (left eye below, right eye above)
    uniform float stereo;
    // eye = 0.0 for the left eye, 0.5 for the right eye
    uniform float eye;

    void main(void)
    {
        vec4 color;
        float tx, ty, width, Y, Cb, Cr;
        int px;
        tx = gl_TexCoord[0].x;
        ty = eye+gl_TexCoord[0].y*stereo;
        width = float(textureSize(tex, 0).x);
        color = texture(tex, vec2(tx, ty));
        px = int(floor(fract(tx*width)*2.0));
        switch (px) {
        case 0:
            Y = color.g;
            break;
        case 1:
            Y = color.a;
            break;
        }
        Y = (Y - 0.0625) * 1.168949772;
        Cb = (color.b - 0.0625) * 1.142857143 - 0.5;
        Cr = (color.r - 0.0625) * 1.142857143 - 0.5;
        color.r = Y + 1.5748 * Cr;
        color.g = Y - 0.1873 * Cb - 0.4681 * Cr;
        color.b = Y + 1.8556 * Cb;
        color.a = 0.7;
        gl_FragColor = color;
    }
"""

# For use with high resolution YUV
#
FragmentShader_v210 = """
    #version 130
    uniform sampler2D tex;
    // stereo = 1.0 if 2D image, =0.5 if 3D (left eye below, right eye above)
    uniform float stereo;
    // eye = 0.0 for the left eye, 0.5 for the right eye
    uniform float eye;

    void main(void)
    {
        vec4 color, color1, color2, color3;
        int px;
        float tx, ty, width, sx, dx, bx, Y, Cb, Cr;
        tx = gl_TexCoord[0].x;
        ty = eye+gl_TexCoord[0].y*stereo;
        width = float(textureSize(tex, 0).x);
        // to sample macro pixels (6 pixels in 4 words)
        sx = tx*width*0.25+0.01;
        // index of display pixel in the macro pixel 0..5
        px = int(floor(fract(sx)*6.0));
        // increment as we sample the macro pixel
        dx = 1.0/width;
        // base x coord of macro pixel
        bx = (floor(sx)+0.01)*dx*4.0;
        color = texture(tex, vec2(bx, ty));
        color1 = texture(tex, vec2(bx+dx, ty));
        color2 = texture(tex, vec2(bx+dx*2.0, ty));
        color3 = texture(tex, vec2(bx+dx*3.0, ty));
        switch (px) {
        case 0:
        case 1:
            Cb = color.b;
            Cr = color.r;
            break;
        case 2:
        case 3:
            Cb = color1.g;
            Cr = color2.b;
            break;
        default:
            Cb = color2.r;
            Cr = color3.g;
            break;
        }
        switch (px) {
        case 0:
            Y = color.g;
            break;
        case 1:
            Y = color1.b;
            break;
        case 2:
            Y = color1.r;
            break;
        case 3:
            Y = color2.g;
            break;
        case 4:
            Y = color3.b;
            break;
        default:
            Y = color3.r;
            break;
        }
        Y = (Y - 0.0625) * 1.168949772;
        Cb = (Cb - 0.0625) * 1.142857143 - 0.5;
        Cr = (Cr - 0.0625) * 1.142857143 - 0.5;
        color.r = Y + 1.5748 * Cr;
        color.g = Y - 0.1873 * Cb - 0.4681 * Cr;
        color.b = Y + 1.8556 * Cb;
        color.a = 0.7;
        gl_FragColor = color;
    }
"""

# The exhausitve list of pixel formats that are transferred as float texture
# Only use those for greater efficiency and compatiblity.
#
fg_shaders = {
    '2vuy': FragmentShader_2vuy,
    '8BitYUV': FragmentShader_2vuy,
    'v210': FragmentShader_v210,
    '10BitYUV': FragmentShader_v210,
    '8BitBGRA': FragmentShader_R10l,
    'BGRA': FragmentShader_R10l,
    '8BitARGB': FragmentShader_R10l,
    '10BitRGBXLE': FragmentShader_R10l,
    'R10l': FragmentShader_R10l
}


#
# Helper function to attach a pixel shader to the material that receives the video frame.
#

def config_video(obj, format, pixel, is3D=False, mat=0, card=0):
    if pixel not in fg_shaders:
        raise('Unsuported shader')
    shader = obj.meshes[0].materials[mat].getShader()
    if shader is not None and not shader.isValid():
        shader.setSource(VertexShader, fg_shaders[pixel], True)
        shader.setSampler('tex', 0)
        shader.setUniformEyef("eye")
        shader.setUniform1f("stereo", 0.5 if is3D else 1.0)
    tex = vt.Texture(obj, mat)
    tex.source = vt.VideoDeckLink(format + "/" + pixel + ("/3D" if is3D else ""), card)
    print("frame rate: ", tex.source.framerate)
    tex.source.play()
    obj["video"] = tex

#
# Attach this function to an object that has a material with texture
# and call it once to initialize the object
#


def init(cont):
    # config_video(cont.owner, 'HD720p5994', '8BitBGRA')
    # config_video(cont.owner, 'HD720p5994', '8BitYUV')
    # config_video(cont.owner, 'pal ', '10BitYUV')
    config_video(cont.owner, 'pal ', '8BitYUV')


#
# To be called on every frame
#
def play(cont):
    obj = cont.owner
    video = obj.get("video")
    if video is not None:
        video.refresh(True)

Video classes

class bge.texture.VideoFFmpeg(file, capture=-1, rate=25.0, width=0, height=0)

FFmpeg video source, used for video files, video captures, or video streams.

Parameters:
  • file (str) – Path to the video to load; if capture >= 0 on Windows, this parameter will not be used.
  • capture (int) – Capture device number; if >= 0, the corresponding webcam will be used. (optional)
  • rate (float) – Capture rate. (optional, used only if capture >= 0)
  • width (int) – Capture width. (optional, used only if capture >= 0)
  • height (int) – Capture height. (optional, used only if capture >= 0)
status

Video status. (readonly)

Type:int
Value:see `FFmpeg Video and Image Status`_.
range

The start and stop time of the video playback, expressed in seconds from beginning. By default the entire video.

Type:sequence of two floats
repeat

Number of times to replay the video, -1 for infinite repeat.

Type:int
framerate

Relative frame rate, <1.0 for slow, >1.0 for fast.

Type:float
valid

Tells if an image is available. (readonly)

Type:bool
image

Image data. (readonly)

Type:Buffer or None
size

Image size. (readonly)

Type:tuple of two ints
scale

Set to True to activate fast nearest neighbor scaling algorithm. Texture width and height must be a power of 2. If the video picture size is not a power of 2, rescaling is required. By default bge.texture uses the precise but slow gluScaleImage() function. Best is to rescale the video offline so that no scaling is necessary at runtime!

Type:bool
flip

If True the imaged will be flipped vertically. FFmpeg always delivers the image upside down, so this attribute is set to True by default.

Type:bool
filter

An additional filter that is applied on the video before sending it to the GPU.

Type:

one of:

preseek

Number of frames of preseek.

Type:int
deinterlace

Deinterlace image.

Type:bool
play()

Play (restart) video.

Returns:Whether the video was ready or stopped.
Return type:bool
pause()

Pause video.

Returns:Whether the video was playing.
Return type:bool
stop()

Stop video (play will replay it from start).

Returns:Whether the video was playing.
Return type:bool
refresh(buffer=None, format="RGBA", timestamp=-1.0)

Refresh video - get its status and optionally copy the frame to an external buffer.

Parameters:
  • buffer (any buffer type) – An optional object that implements the buffer protocol. If specified, the image is copied to the buffer, which must be big enough or an exception is thrown.
  • format (str) – An optional image format specifier for the image that will be copied to the buffer. Only valid values are “RGBA” or “BGRA”
  • timestamp (float) – An optional timestamp (in seconds from the start of the movie) of the frame to be copied to the buffer.
Returns:

see `FFmpeg Video and Image Status`_.

Return type:

int

Image classes

class bge.texture.ImageFFmpeg(file)

FFmpeg image source, used for image files and web based images.

Parameters:file (str) – Path to the image to load.
status

Image status. (readonly)

Type:int
Value:see `FFmpeg Video and Image Status`_.
valid

Tells if an image is available. (readonly)

Type:bool
image

Image data. (readonly)

Type:Buffer or None
size

Image size. (readonly)

Type:tuple of two ints
scale

Fast scale of image (near neighbour).

Type:bool
flip

Flip image vertically.

Type:bool
filter

Pixel filter.

Type:

one of…

refresh(buffer=None, format="RGBA")

Refresh image, get its status and optionally copy the frame to an external buffer.

Parameters:
  • buffer (any buffer type) – An optional object that implements the buffer protocol. If specified, the image is copied to the buffer, which must be big enough or an exception is thrown.
  • format (str) – An optional image format specifier for the image that will be copied to the buffer. Only valid values are “RGBA” or “BGRA”
Returns:

see `FFmpeg Video and Image Status`_.

Return type:

int

reload(newname=None)

Reload image, i.e. reopen it.

Parameters:newname (str) – Path to a new image. (optional)
class bge.texture.ImageBuff(width, height, color=0, scale=False)

Image from application memory. For computer generated images, drawing applications.

Parameters:
  • width (int) – Width of the image.
  • height (int) – Height of the image.
  • color (int in [0, 255]) – Value to initialize RGB channels with. The initialized buffer will have all pixels set to (color, color, color, 255). (optional)
  • scale (bool) – Image uses scaling. (optional)
filter

Pixel filter.

Type:

one of…

flip

Flip image vertically.

Type:bool
image

Image data. (readonly)

Type:Buffer or None
load(imageBuffer, width, height)

Load image from buffer.

Parameters:
  • imageBuffer (Buffer or Python object implementing the buffer protocol (f.ex. bytes)) – Buffer to load the image from.
  • width (int) – Width of the image to load.
  • height (int) – Height of the image to load.
plot(imageBuffer, width, height, positionX, positionY, mode=IMB_BLEND_COPY)

Update image buffer.

Parameters:
  • imageBuffer (Buffer, ImageBuff or Python object implementing the buffer protocol (f.ex. bytes)) – Buffer to load the new data from.
  • width (int) – Width of the data to load.
  • height (int) – Height of the data to load.
  • positionX (int) – Left boundary of the region to be drawn on.
  • positionY (int) – Upper boundary of the region to be drawn on.
  • mode (int) – Drawing mode, see `Image Blending Modes`_.
scale

Fast scale of image (near neighbour).

Type:bool
size

Image size. (readonly)

Type:tuple of two ints
valid

Tells if an image is available. (readonly)

Type:bool
class bge.texture.ImageMirror(scene, observer, mirror, material=0, width, height, samples, hdr)

Image source from mirror.

Parameters:
  • scene (KX_Scene) – Scene in which the image has to be taken.
  • observer (KX_GameObject) – Reference object for the mirror (the object from which the mirror has to be looked at, for example a camera).
  • mirror (KX_GameObject) – Object holding the mirror.
  • material (int) – ID of the mirror’s material to be used for mirroring. (optional)
  • width (integer) – Off-screen render buffer width (optional).
  • height (integer) – Off-screen render buffer height (optional).
  • samples (integer) – Off-screen render buffer samples (optional).
  • hdr (One of these constants) – Off-screen image format (optional).
alpha

Use alpha in texture.

Type:bool
horizon

Horizon color.

Type:float list [r, g, b, a] in [0.0, 1.0]
zenith

Zenith color.

Type:float list [r, g, b, a] in [0.0, 1.0]
background
Type:float list [r, g, b, a] in [0.0, 1.0]

Deprecated use bge.texture.ImageMirror.horizon() or bge.texture.ImageMirror.zenith() instead.

updateShadow

Choose to force shadow buffer update if there is a gap beetween image rendered and shadows.

Type:bool
colorBindCode

Off-screen color texture bind code.

Type:integer
capsize

Size of render area.

Type:sequence of two ints
clip

Clipping distance.

Type:float in [0.01, 5000.0]
filter

Pixel filter.

Type:

one of…

flip

Flip image vertically.

Type:bool
image

Image data. (readonly)

Type:Buffer or None
refresh(buffer=None, format="RGBA")

Refresh image - render and copy the image to an external buffer (optional) then invalidate its current content.

Parameters:
  • buffer (any buffer type) – An optional object that implements the buffer protocol. If specified, the image is rendered and copied to the buffer, which must be big enough or an exception is thrown.
  • format (str) – An optional image format specifier for the image that will be copied to the buffer. Only valid values are “RGBA” or “BGRA”
scale

Fast scale of image (near neighbour).

Type:bool
size

Image size (readonly).

Type:tuple of two ints
valid

Tells if an image is available. (readonly)

Type:bool
whole

Use whole viewport to render.

Type:bool
class bge.texture.ImageMix

Image mixer used to mix multiple image sources together.

filter

Pixel filter.

Type:

one of…

flip

Flip image vertically.

Type:bool
getSource(id)

Get image source.

Parameters:id (str) – Identifier of the source to get.
Returns:Image source.
Return type:one of…
getWeight(id)

Get image source weight.

Parameters:id (str) – Identifier of the source.
Returns:Weight of the source.
Return type:int
image

Image data. (readonly)

Type:Buffer or None
refresh(buffer=None, format="RGBA")

Refresh image - calculate and copy the image to an external buffer (optional) then invalidate its current content.

Parameters:
  • buffer (any buffer type) – An optional object that implements the buffer protocol. If specified, the image is calculated and copied to the buffer, which must be big enough or an exception is thrown.
  • format (str) – An optional image format specifier for the image that will be copied to the buffer. Only valid values are “RGBA” or “BGRA”
scale

Fast scale of image (near neighbour).

Type:bool
size

Image size. (readonly)

Type:tuple of two ints
setSource(id, image)

Set image source - all sources must have the same size.

Parameters:
setWeight(id, weight)

Set image source weight - the sum of the weights should be 256 to get full color intensity in the output.

Parameters:
  • id (str) – Identifier of the source.
  • weight (int) – Weight of the source.
valid

Tells if an image is available. (readonly)

Type:bool
class bge.texture.ImageRender(scene, camera, width, height, samples, hdr)

Image source from a render of a non active camera. The render is done on a custom framebuffer object if fbo is specified, otherwise on the default framebuffer.

Parameters:
  • scene (KX_Scene) – Scene in which the image has to be taken.
  • camera (KX_Camera) – Camera from which the image has to be taken.
  • width (integer) – Off-screen render buffer width (optional).
  • height (integer) – Off-screen render buffer height (optional).
  • samples (integer) – Off-screen render buffer samples (optional).
  • hdr (One of these constants) – Off-screen image format (optional).
alpha

Use alpha in texture.

Type:bool
horizon

Horizon color.

Type:float list [r, g, b, a] in [0.0, 1.0]
zenith

Zenith color.

Type:float list [r, g, b, a] in [0.0, 1.0]
background

Background color.

Type:float list [r, g, b, a] in [0.0, 1.0]

Deprecated use bge.texture.ImageRender.horizon() or bge.texture.ImageRender.zenith() instead.

updateShadow

Choose to force shadow buffer update if there is a gap beetween image rendered and shadows.

Type:bool
colorBindCode

Off-screen color texture bind code.

Type:integer
capsize

Size of render area.

Type:sequence of two ints
filter

Pixel filter.

Type:

one of…

flip

Flip image vertically.

Type:bool
image

Image data. (readonly)

Type:Buffer or None
scale

Fast scale of image (near neighbour).

Type:bool
size

Image size. (readonly)

Type:tuple of two ints
valid

Tells if an image is available. (readonly)

Type:bool
whole

Use whole viewport to render.

Type:bool
depth

Use depth component of render as array of float - not suitable for texture source, should only be used with bge.texture.imageToArray(mode=’F’).

Type:bool
zbuff

Use depth component of render as grayscale color - suitable for texture source.

Type:bool
render()

Render the scene but do not extract the pixels yet. The function returns as soon as the render commands have been send to the GPU. The render will proceed asynchronously in the GPU while the host can perform other tasks. To complete the render, you can either call refresh() directly of refresh the texture of which this object is the source. This method is useful to implement asynchronous render for optimal performance: call render() on frame n and refresh() on frame n+1 to give as much as time as possible to the GPU to render the frame while the game engine can perform other tasks.

Returns:True if the render was initiated, False if the render cannot be performed (e.g. the camera is active)
Return type:bool
refresh()
refresh(buffer, format="RGBA")

Refresh video - render and optionally copy the image to an external buffer then invalidate its current content. The render may have been started earlier with the render() method, in which case this function simply waits for the render operations to complete. When called without argument, the pixels are not extracted but the render is guaranteed to be completed when the function returns. This only makes sense with offscreen render on texture target (see offScreenCreate()).

Parameters:
  • buffer (any buffer type of sufficient size) – An object that implements the buffer protocol. If specified, the image is copied to the buffer, which must be big enough or an exception is thrown. The transfer to the buffer is optimal if no processing of the image is needed. This is the case if flip=False, alpha=True, scale=False, whole=True, depth=False, zbuff=False and no filter is set.
  • format (str) – An optional image format specifier for the image that will be copied to the buffer. Only valid values are “RGBA” or “BGRA”
Returns:

True if the render is complete, False if the render cannot be performed (e.g. the camera is active)

Return type:

bool

class bge.texture.ImageViewport

Image source from viewport rendered by the active camera. To render from a non active camera see ImageRender.

alpha

Use alpha in texture.

Type:bool
capsize

Size of viewport area being captured.

Type:sequence of two ints
filter

Pixel filter.

Type:

one of…

flip

Flip image vertically.

Type:bool
image

Image data. (readonly)

Type:Buffer or None
position

Upper left corner of the captured area.

Type:sequence of two ints
refresh(buffer=None, format="RGBA")

Refresh video - copy the viewport to an external buffer (optional) then invalidate its current content.

Parameters:
  • buffer (any buffer type) – An optional object that implements the buffer protocol. If specified, the image is copied to the buffer, which must be big enough or an exception is thrown. The transfer to the buffer is optimal if no processing of the image is needed. This is the case if flip=False, alpha=True, scale=False, whole=True, depth=False, zbuff=False and no filter is set.
  • format (str) – An optional image format specifier for the image that will be copied to the buffer. Only valid values are “RGBA” or “BGRA”
scale

Fast scale of image (near neighbour).

Type:bool
size

Image size. (readonly)

Type:tuple of two ints
valid

Tells if an image is available. (readonly)

Type:bool
whole

Use whole viewport to capture.

Type:bool
depth

Use depth component of viewport as array of float - not suitable for texture source, should only be used with bge.texture.imageToArray(mode='F').

Type:bool
zbuff

Use depth component of viewport as grayscale color - suitable for texture source.

Type:bool

Image source from an external video stream captured with a DeckLink video card from Black Magic Design. Before this source can be used, a DeckLink hardware device must be installed, it can be a PCIe card or a USB device, and the ‘Desktop Video’ software package (version 10.4 or above must be installed) on the host as described in the DeckLink documentation. If in addition you have a recent nVideo Quadro card, you can benefit from the ‘GPUDirect’ technology to push the captured video frame very efficiently to the GPU. For this you need to install the ‘DeckLink SDK’ version 10.4 or above and copy the ‘dvp.dll’ runtime library to Blender’s installation directory or to any other place where Blender can load a DLL from.

Parameters:
  • format (str) – string describing the video format to be captured.
  • capture (int) – Card number from which the input video must be captured.

The format argument must be written as <displayMode>/<pixelFormat>[/3D][:<cacheSize>] where <displayMode> describes the frame size and rate and <pixelFormat> the encoding of the pixels. The optional /3D suffix is to be used if the video stream is stereo with a left and right eye feed. The optional :<cacheSize> suffix determines the number of the video frames kept in cache, by default 8. Some DeckLink cards won’t work below a certain cache size. The default value 8 should be sufficient for all cards. You may try to reduce the cache size to reduce the memory footprint. For example the The 4K Extreme is known to work with 3 frames only, the Extreme 2 needs 4 frames and the Intensity Shuttle needs 6 frames, etc. Reducing the cache size may be useful when Decklink is used in conjunction with GPUDirect: all frames must be locked in memory in that case and that puts a lot of pressure on memory. If you reduce the cache size too much, you’ll get no error but no video feed either.

The valid <displayMode> values are copied from the BMDDisplayMode enum in the DeckLink API without the ‘bmdMode’ prefix. In case a mode that is not in this list is added in a later version of the SDK, it is also possible to specify the 4 letters of the internal code for that mode. You will find the internal code in the DeckLinkAPIModes.h file that is part of the SDK. Here is for reference the full list of supported display modes with their equivalent internal code:

Internal Codes
  • NTSC ‘ntsc’
  • NTSC2398 ‘nt23’
  • PAL ‘pal ‘
  • NTSCp ‘ntsp’
  • PALp ‘palp’
HD 1080 Modes
  • HD1080p2398 ‘23ps’
  • HD1080p24 ‘24ps’
  • HD1080p25 ‘Hp25’
  • HD1080p2997 ‘Hp29’
  • HD1080p30 ‘Hp30’
  • HD1080i50 ‘Hi50’
  • HD1080i5994 ‘Hi59’
  • HD1080i6000 ‘Hi60’
  • HD1080p50 ‘Hp50’
  • HD1080p5994 ‘Hp59’
  • HD1080p6000 ‘Hp60’
HD 720 Modes
  • HD720p50 ‘hp50’
  • HD720p5994 ‘hp59’
  • HD720p60 ‘hp60’
2k Modes
  • 2k2398 ‘2k23’
  • 2k24 ‘2k24’
  • 2k25 ‘2k25’
4k Modes
  • 4K2160p2398 ‘4k23’
  • 4K2160p24 ‘4k24’
  • 4K2160p25 ‘4k25’
  • 4K2160p2997 ‘4k29’
  • 4K2160p30 ‘4k30’
  • 4K2160p50 ‘4k50’
  • 4K2160p5994 ‘4k59’
  • 4K2160p60 ‘4k60’

Most of names are self explanatory. If necessary refer to the DeckLink API documentation for more information.

Similarly, <pixelFormat> is copied from the BMDPixelFormat enum.

Here is for reference the full list of supported pixel format and their equivalent internal code:

Pixel Formats
  • 8BitYUV ‘2vuy’
  • 10BitYUV ‘v210’
  • 8BitARGB * no equivalent code *
  • 8BitBGRA ‘BGRA’
  • 10BitRGB ‘r210’
  • 12BitRGB ‘R12B’
  • 12BitRGBLE ‘R12L’
  • 10BitRGBXLE ‘R10l’
  • 10BitRGBX ‘R10b’

Refer to the DeckLink SDK documentation for a full description of these pixel format. It is important to understand them as the decoding of the pixels is NOT done in VideoTexture for performance reason. Instead a specific shader must be used to decode the pixel in the GPU. Only the ‘8BitARGB’, ‘8BitBGRA’ and ‘10BitRGBXLE’ pixel formats are mapped directly to OpenGL RGB float textures. The ‘8BitYUV’ and ‘10BitYUV’ pixel formats are mapped to openGL RGB float texture but require a shader to decode. The other pixel formats are sent as a GL_RED_INTEGER texture (i.e. a texture with only the red channel coded as an unsigned 32 bit integer) and are not recommended for use.

Example: HD1080p24/10BitYUV/3D:4 is equivalent to 24ps/v210/3D:4 and represents a full HD stereo feed at 24 frame per second and 4 frames cache size.

Although video format auto detection is possible with certain DeckLink devices, the corresponding API is NOT implemented in the BGE. Therefore it is important to specify the format string that matches exactly the video feed. If the format is wrong, no frame will be captured. It should be noted that the pixel format that you need to specify is not necessarily the actual format in the video feed. For example, the 4K Extreme card delivers 8bit RGBs pixels in the ‘10BitRGBXLE’ format. Use the ‘Media Express’ application included in ‘Desktop Video’ to discover which pixel format works for a particular video stream.

status

Status of the capture: 1=ready to use, 2=capturing, 3=stopped

Type:int
framerate

Capture frame rate as computed from the video format.

Type:float
valid

Tells if the image attribute can be used to retrieve the image. Always False in this implementation (the image is not available at python level)

Type:bool
image

The image data. Always None in this implementation.

Type:Buffer or None
size

The size of the frame in pixel. Stereo frames have double the height of the video frame, i.e. 3D is delivered to the GPU as a single image in top-bottom order, left eye on top.

Type:(int,int)
scale

Not used in this object.

Type:bool
flip

Not used in this object.

Type:bool
filter

Not used in this object.

play()

Kick-off the capture after creation of the object.

Returns:True if the capture could be started, False otherwise.
Return type:bool
pause()

Temporary stops the capture. Use play() to restart it.

Returns:True if the capture could be paused, False otherwise.
Return type:bool
stop()

Stops the capture.

Returns:True if the capture could be stopped, False otherwise.
Return type:bool

Texture classes

class bge.texture.Texture(gameObj, materialID=0, textureID=0, textureObj=None)

Class that creates the Texture object that loads the dynamic texture on the GPU.

Parameters:
  • gameObj (KX_GameObject) – Game object to be created a video texture on.
  • materialID (int) – Material ID default, 0 is the first material. (optional)
  • textureID (int) – Texture index in case of multi-texture channel, 0 = first channel by default. In case of UV texture, this parameter should always be 0. (optional)
  • textureObj (Texture) – Reference to another Texture object with shared bindId which he user might want to reuse the texture. If this argument is used, you should not create any source on this texture and there is no need to refresh it either: the other Texture object will provide the texture for both materials/textures.(optional)
bindId

OpenGL Bind Name. (readonly)

Type:int
close()

Close dynamic texture and restore original.

mipmap

Mipmap texture.

Type:bool
refresh(refresh_source, timestamp=-1.0)

Refresh texture from source.

Parameters:
  • refresh_source (bool) – Whether to also refresh the image source of the texture.
  • timestamp (float) – If the texture controls a VideoFFmpeg object: timestamp (in seconds from the start of the movie) of the frame to be loaded; this can be used for video-sound synchonization by passing time to it. (optional)
source

Source of texture.

Type:

one of…

Certain DeckLink devices can be used to playback video: the host sends video frames regularly for immediate or scheduled playback. The video feed is outputted on HDMI or SDI interfaces. This class supports the immediate playback mode: it has a source attribute that is assigned one of the source object in the bge.texture module. Refreshing the DeckLink object causes the image source to be computed and sent to the DeckLink device for immediate transmission on the output interfaces. Keying is supported: it allows to composite the frame with an input video feed that transits through the DeckLink card.

Parameters:
  • cardIdx (int) – Number of the card to be used for output (0=first card). It should be noted that DeckLink devices are usually half duplex: they can either be used for capture or playback but not both at the same time.
  • format (str) – String representing the display mode of the output feed.

The default value of the format argument is reserved for auto detection but it is currently not supported (it will generate a runtime error) and thus the video format must be explicitly specified. If keying is the goal (see keying attributes), the format must match exactly the input video feed, otherwise it can be any format supported by the device (there will be a runtime error if not). The format of the string is <displayMode>[/3D].

Refer to VideoDeckLink to get the list of acceptable <displayMode>. The optional /3D suffix is used to create a stereo 3D feed. In that case the ‘right’ attribute must also be set to specify the image source for the right eye.

Note: The pixel format is not specified here because it is always BGRA. The alpha channel is used in keying to mix the source with the input video feed, otherwise it is not used. If a conversion is needed to match the native video format, it is done inside the DeckLink driver or device.

source

This attribute must be set to one of the image sources. If the image size does not fit exactly the frame size, the extend attribute determines what to do.

For best performance, the source image should match exactly the size of the output frame. A further optimization is achieved if the image source object is ImageViewport or ImageRender set for whole viewport, flip disabled and no filter: the GL frame buffer is copied directly to the image buffer and directly from there to the DeckLink card (hence no buffer to buffer copy inside VideoTexture).

Type:one of… - VideoFFmpeg - VideoDeckLink - ImageFFmpeg - ImageBuff - ImageMirror - ImageMix - ImageRender - ImageViewport
right

If the video format is stereo 3D, this attribute should be set to an image source object that will produce the right eye images. If the goal is to render the BGE scene in 3D, it can be achieved with 2 cameras, one for each eye, used by 2 ImageRender with an offscreen render buffer that is just the size of the video frame.

Type:one of… - VideoFFmpeg - VideoDeckLink - ImageFFmpeg - ImageBuff - ImageMirror - ImageMix - ImageRender - ImageViewport
keying

Specify if keying is enabled. False (default): the output frame is sent unmodified on the output interface (in that case no input video is required). True: the output frame is mixed with the input video, using the alpha channel to blend the two images and the combination is sent on the output interface.

Type:bool
level

If keying is enabled, sets the keying level from 0 to 255. This value is a global alpha value that multiplies the alpha channel of the image source. Use 255 (the default) to keep the alpha channel unmodified, 0 to make the output frame totally transparent.

Type:int
extend

Determines how the image source should be mapped if the size does not fit the video frame size. * False (the default): map the image pixel by pixel. If the image size is smaller than the frame size, extra space around the image is filled with 0-alpha black. If it is larger, the image is cropped to fit the frame size. * True: the image is scaled by the nearest neighbor algorithm to fit the frame size. The scaling is fast but poor quality. For best results, always adjust the image source to match the size of the output video.

Type:bool
close()

Close the DeckLink device and release all resources. After calling this method, the object cannot be reactivated, it must be destroyed and a new DeckLink object created from fresh to restart the output.

refresh(refresh_source, ts)

This method must be called frequently to update the output frame in the DeckLink device.

Parameters:
  • refresh_source (bool) – True if the source objects image buffer should be invalidated after being used to compute the output frame. This triggers the recomputing of the source image on next refresh, which is normally the desired effect. False if the image source buffer should stay valid and reused on next refresh. Note that the DeckLink device stores the output frame and replays until a new frame is sent from the host. Thus, it is not necessary to refresh the DeckLink object if it is known that the image source has not changed.
  • ts (float) – The timestamp value passed to the image source object to compute the image. If unspecified, the BGE clock is used.

Filter classes

class bge.texture.FilterBGR24

Source filter BGR24.

class bge.texture.FilterBlueScreen

Filter for Blue Screen. The RGB channels of the color are left unchanged, while the output alpha is obtained as follows:

  • if the square of the euclidian distance between the RGB color and the filter’s reference color is smaller than the filter’s lower limit, the output alpha is set to 0;
  • if that square is bigger than the filter’s upper limit, the output alpha is set to 255;
  • otherwise the output alpha is linarly extrapoled between 0 and 255 in the interval of the limits.
color

Reference color.

Type:sequence of three ints
Default:(0, 0, 255)
limits

Reference color limits.

Type:sequence of two ints
Default:(64, 64)
previous

Previous pixel filter.

Type:

one of…

class bge.texture.FilterColor

Filter for color calculations. The output color is obtained by multiplying the reduced 4x4 matrix with the input color and adding the remaining column to the result.

matrix

Matrix [4][5] for color calculation.

Type:sequence of four sequences of five ints
Default:((256, 0, 0, 0, 0), (0, 256, 0, 0, 0), (0, 0, 256, 0, 0), (0, 0, 0, 256, 0))
previous

Previous pixel filter.

Type:

one of…

class bge.texture.FilterGray

Filter for grayscale effect. Proportions of R, G and B contributions in the output grayscale are 28:151:77.

previous

Previous pixel filter.

Type:

one of…

class bge.texture.FilterLevel

Filter for levels calculations. Each output color component is obtained as follows:

  • if it is smaller than its corresponding min value, it is set to 0;
  • if it is bigger than its corresponding max value, it is set to 255;
  • Otherwise it is linearly extrapoled between 0 and 255 in the (min, max) interval.
levels

Levels matrix [4] (min, max).

Type:sequence of four sequences of two ints
Default:((0, 255), (0, 255), (0, 255), (0, 255))
previous

Previous pixel filter.

Type:

one of…

class bge.texture.FilterNormal

Normal map filter.

colorIdx

Index of color used to calculate normal (0 - red, 1 - green, 2 - blue, 3 - alpha).

Type:int in [0, 3]
Default:0
depth

Depth of relief.

Type:float
Default:4.0
previous

Previous pixel filter.

Type:

one of…

class bge.texture.FilterRGB24

Returns a new input filter object to be used with ImageBuff object when the image passed to the ImageBuff.load() function has the 3-bytes pixel format BGR.

class bge.texture.FilterRGBA32

Source filter RGBA32.

Functions

bge.texture.getLastError()

Last error that occurred in a bge.texture function.

Returns:The description of the last error occurred in a bge.texture function.
Return type:str
bge.texture.imageToArray(image, mode)

Returns a Buffer corresponding to the current image stored in a texture source object.

Parameters:
  • image

    Image source object of type:

  • mode (str) –

    Optional argument representing the pixel format.

    • You can use the characters R, G, B for the 3 color channels, A for the alpha channel, 0 to force a fixed 0 color channel and 1 to force a fixed 255 color channel.

      Examples:

      • ”BGR” will return 3 bytes per pixel with the Blue, Green and Red channels in that order.
      • ”RGB1” will return 4 bytes per pixel with the Red, Green, Blue channels in that order and the alpha channel forced to 255.
    • A special mode “F” allows to return the image as an array of float. This mode should only be used to retrieve the depth buffer of the class:ImageViewport and ImageRender objects. The default mode is “RGBA”.
Returns:

An object representing the image as one dimensional array of bytes of size (pixel_size*width*height), line by line starting from the bottom of the image. The pixel size and format is determined by the mode parameter. For mode ‘F’, the array is a one dimensional array of float of size (width*height).

Return type:

Buffer

bge.texture.materialID(object, name)

Returns a numeric value that can be used in Texture to create a dynamic texture.

The value corresponds to an internal material number that uses the texture identified by name. name is a string representing a texture name with IM prefix if you want to identify the texture directly. This method works for basic tex face and for material, provided the material has a texture channel using that particular texture in first position of the texture stack. name can also have MA prefix if you want to identify the texture by material. In that case the material must have a texture channel in first position.

If the object has no material that matches name, it generates a runtime error. Use try/except to catch the exception.

Ex: bge.texture.materialID(obj, 'IMvideo.png')

Parameters:
  • object (KX_GameObject) – The game object that uses the texture you want to make dynamic.
  • name (str) – Name of the texture/material you want to make dynamic.
Returns:

The internal material number.

Return type:

int

bge.texture.setLogFile(filename)

Sets the name of a text file in which runtime error messages will be written, in addition to the printing of the messages on the Python console. Only the runtime errors specific to the VideoTexture module are written in that file, ordinary runtime time errors are not written.

Parameters:filename (str) – Name of the error log file.
Returns:-1 if the parameter name is invalid (not of type string), else 0.
Return type:int

Constants

bge.texture.SOURCE_ERROR
bge.texture.SOURCE_EMPTY
bge.texture.SOURCE_READY
bge.texture.SOURCE_PLAYING
bge.texture.SOURCE_STOPPED

See Wikipedia’s Blend Modes for reference.

bge.texture.IMB_BLEND_MIX
bge.texture.IMB_BLEND_ADD
bge.texture.IMB_BLEND_SUB
bge.texture.IMB_BLEND_MUL
bge.texture.IMB_BLEND_LIGHTEN
bge.texture.IMB_BLEND_DARKEN
bge.texture.IMB_BLEND_ERASE_ALPHA
bge.texture.IMB_BLEND_ADD_ALPHA
bge.texture.IMB_BLEND_OVERLAY
bge.texture.IMB_BLEND_HARDLIGHT
bge.texture.IMB_BLEND_COLORBURN
bge.texture.IMB_BLEND_LINEARBURN
bge.texture.IMB_BLEND_COLORDODGE
bge.texture.IMB_BLEND_SCREEN
bge.texture.IMB_BLEND_SOFTLIGHT
bge.texture.IMB_BLEND_PINLIGHT
bge.texture.IMB_BLEND_VIVIDLIGHT
bge.texture.IMB_BLEND_LINEARLIGHT
bge.texture.IMB_BLEND_DIFFERENCE
bge.texture.IMB_BLEND_EXCLUSION
bge.texture.IMB_BLEND_HUE
bge.texture.IMB_BLEND_SATURATION
bge.texture.IMB_BLEND_LUMINOSITY
bge.texture.IMB_BLEND_COLOR
bge.texture.IMB_BLEND_COPY
bge.texture.IMB_BLEND_COPY_RGB
bge.texture.IMB_BLEND_COPY_ALPHA