Frequently asked questions
These are the questions that arrive most often in editorial correspondence. Each answer links onward to the module that treats the subject in full.
The questions below cover the recurring decisions in a compression workflow: which codec to choose, how to set quality, what container to use, and how to know when an encode is finished. Where an answer depends on your own material, it says so rather than offering a universal number.
Which codec should I use for general delivery?
Use the most efficient codec that your playback targets can decode. In practice that often means a baseline tier in a widely supported codec and a premium tier in a more efficient one, with the manifest offering both. See the codec architecture module.
What CRF value gives the best quality?
The question has no fixed answer, because the scale is per-codec and the visible result depends on resolution, frame rate and viewing distance. Start from the ranges in the CRF and RF module, then validate on a representative sample.
Is a lower CRF always better?
Lower means higher quality and a larger file at the same content. Past a certain point the additional bits describe detail that cannot be seen at the intended viewing distance, so the extra storage and bandwidth buy nothing perceptible.
What is the difference between a container and a codec?
A container is an envelope that indexes and times streams; a codec is the compression method applied to the content. Converting the container alone can be lossless when the payload is untouched. See container formats explained.
Why does the file play on my computer but not on my television?
Almost always a declaration mismatch. Compare profile, level, bit depth and colour metadata against what the television advertises. Decoders reject streams outside their advertised envelope rather than degrading gracefully.
How long should a GOP be?
Long enough to exploit temporal redundancy and short enough to give a useful seek grid. One to two seconds between keyframes is a common compromise, and it should be aligned across every representation in a ladder.
Are hardware encoders good enough?
For a large share of volume delivery they are, and they win on total cost. For premium masters and archival work, a software encoder at a slow preset still leads on rate-quality. See GPU acceleration.
How do I know when an encode is finished well?
Measure with a metric that correlates with perception, confirm by eye on a calibrated display, and test playback on a target device. All three checks together are the minimum for a defensible delivery.
Does denoising help or hurt the result?
Light temporal denoising usually reduces the bitrate at a fixed quality target with little visible cost, because sensor noise is expensive to code. Over-denoising removes texture the viewer expects to see, so measure rather than guess.
What is a bitrate ladder and do I need one?
A ladder is a set of encodes at graduated quality so a player can adapt to network conditions. Any service serving more than a trivial audience benefits from one. See bitrate ladders.
If your question is not answered here, the glossary defines every acronym used across the site, and the contact page lists the editorial desk.