Exporting

-> DepthFlow supports realtime previews and video exporting.

After creating your Animation (or using a #preset), and sending Inputs images to a scene instance, it's time to run the animation and export them to a shareable video!

Video¶

Simply send an output="video.mp4" argument to the scene.main method/command:

scene = MyScene(backend="headless")
scene.main(output="video.mp4")

depthflow main --output video.mp4

Note: See all available options with depthflow main --help, most you already expect:

scene.main(
    output="video.mkv",
    fps=60,
    time=10,
    width=1920,
    height=1080,
    ssaa=1.0,
)

Batch¶

For exporting multiple videos, you can follow something similar to the batch.py example script, where a scene is reused on many renders over globbing image directories - some notes:

• Always initialize the scene with a backend=headless for compatibility.
• Always reset the scene's state before rendering again, as the previous animation ending could leave changes in the camera parameters a second animation doesn't enforce.

Codec¶

Very large topic, until ShaderFlow documentation is written, you can:

scene = MyScene()
scene.ffmpeg.h264(preset="slow", crf=23)
scene.ffmpeg.h264_nvenc(...)
scene.ffmpeg.h265(...)
scene.main(output="video.mp4")

# See all available codecs
depthflow --help

# See a codec options with
depthflow h264 --help

# Use a codec with settings
depthflow h264 --preset slow --crf 23 main -o video.mp4 (...)

Quality¶

The main things that affect the final video quality are:

1. Depthmap precision: DepthFlow is highly sensitive to good relative depths between any two points on the scene, and not object sillhouette precision. Learn more at Inputs/#depth.
2. SSAA Value: Render at a higher resolution and downscales to output, default being 1.0 (pure). Must only go as high¹ as 2 * subsample (default 2), the downscale kernel size, hurts quality otherwise. A value of 2.0 is plenty for final exports.
3. Resolution: No gains in videos larger than input sources, should at least match it. When using SSAA, zooming, or with large #heights or #offsets, a larger input image has benefits.
4. Quality parameter: Explained in Camera/#quality.
5. Image contents: Explained in Inputs/#image.
6. Encoder settings: Explained in #codec.