Framerate

If you want to retain the same frame rate as the source media, then you do not need to change the frame rate. If you want to reduce or increase the FPS, then enter the desired field via our web UI or our API. 

Since users often will be uploading a wide variety of videos, I generally like to break them down into two types:

Static/Low Action - stable tripod shots, very little background movement, actors standing still 
Active/High Action - panning/jerky camera, lots of action, sports-like movement

A good starting point is to choose a variable bitrate setting equal to the width of the video. So, for example: 640x480 SD at 640 kilobits per sec (kbps), or 1280x720 HD at 1280 kbps. Higher action video may require a slightly higher bitrate to prevent blocking artifacts.

Noise Reduction

Applying some noise reduction is useful to save bits for high detail regions, but be careful not to overdo it. I've seen video samples where whole areas of ocean and grassy fields disappear to achieve a lower bitrate. This of course, falls under artistic preference, but generally I'd rather see a smaller framesize and more detail. At low bitrates, it is increasing important to improve the quality of noisy video sources, such as film containing lots of grain or video shot in low light. The 3D noise reduction in ffmpeg allows control over both luma and chroma values for fine-tuning your output image quality.

	<noise_reduction>4:3:6</noise_reduction> 

luma_spatial – Spatial Luma Strength. Allowed values: [0,255]
chroma_spatial – Spatial Chroma Strength. Allowed values: [0,255]
luma_temp – Temporal Luma Strength. Allowed values: [0,255]

General recommended starting value is 4:3:6. [luma_spatial:chroma_spatial:luma_temp] Noise reduction is also available via our web interface as the High Quality 3D Denoiser option.

Single Pass vs. Two Pass

For most purposes 2-pass encoding achieves very good results. It's a tradeoff of diminishing returns, 2-pass gaining perhaps 10% quality bit-for-bit but doubling the encoding time. Do not lower qcomp, CBR is horrible on quality. I'd experiment with values floating between 0.60 and 0.80 if you want more VBR. if qcomp = 1.00 then quantizer is constant for second pass. Real variable bitrate with constant quality. if qcomp = 0.00 then bitrate is constant for second pass. Real constant bitrate with variable quality.

<two_pass>yes</two_pass>

I'd recommend having two sample videos, tell your users to choose Low or High Action content setting, experiment a bit with your B-frames then define two "baseline" settings for each bitrate. For web video it's best to narrow your targets to four different bitrates at most, especially if you are going to be processing thousands of users uploading. Most folks have a slow (up to 240kbits), good (~700kbits), or fast (2mbits and higher) connection. I'd say H.264 over 2mbits is generally overkill for website content. For general purposes, I'd recommend 2-pass and we push a 10 second keyframe interval (300 frames) which may not be appropriate for "high-action" source video.

For more detail on H.264 controls for scenecut thresholds, B-frames, and more, please refer to:

Advanced H.264 Guide http://sites.google.com/site/linuxencoding/x264-ffmpeg-mapping

H.264 parameters for our API http://www.encoding.com/help/article/advanced_configuration_options_for_the_libx264_video_codec

An excellent collection of HD videos at 2mbits/sec can be found at http://californiaisaplace.com/cali/ 

16x16 Macroblocks

H.264/AVC does a much more efficient job when the horizontal and vertical framesize dimensions are multiples of 16. Good examples include:

SD (4:3) aspect ratios: 320x240, 432x320, 480x360, 544x400, 640x480, 768x576
HD (16:9) aspect ratios: 432x240, 576x320, 640x360, 720x400, 848x480, 1024x576, 1280x720, 1920x1080

In 4:2:0 H.264/AVC coding, each block contains 4 luminance samples (Y), 1 blue sample (Cb), and 1 red sample (Cr). Modern video decoding chips (GPUs) are optimized for playback of 16x16 macroblocking.

Keyframes and GOPs

Low action scenes generally handle more bidirectional (B-frames) better since they don't have to track interframe motion as aggressively. Higher action content will require more keyframes (I-frames) to keep the picture from breaking apart. Longer GOPs with more B-frames also require more buffering by the playback GPU to recursively track the motion for each macroblock. Fortunately, x264 offers very good scene detection, which is why for most applications, we set keyframes to 300.

H.264/AVC sample for modern mobiles (30 fps with a 10 second GOP)

	<framerate>30</framerate>
	<keyframe>300</keyframe>

For older computers, and early generations of iPod and Blackberry phones, the chips might not have enough processing power and memory to successfully buffer longer GOPs. Keep your bitrates low, try lower framerates, and shorter GOPs.

H.264/AVC sample for older mobiles (15 fps with a 4 second GOP)

	<framerate>15</framerate>
	<keyframe>60</keyframe>

More information about GOPs available on wiki http://en.wikipedia.org/wiki/Group_of_pictures

Turbo Mode

NOTE: For bigger or longer HD encoding jobs, turbo mode is absolutely recommended since you will see speed gains in the neighborhood of 3x faster vs. normal mode. Please be aware turbo is running on more powerful encoders, so it costs an extra $1 per gigabyte.

<turbo>yes</turbo>