Compression
VBR, CBR and Constant Quality: Which Bitrate Mode to Actually Use
A constant bitrate spends the same amount on a still shot and a fight scene. A variable one does not. Here is what each mode is for, and why the third option is usually the right answer.
4 min read · Published 16 August 2026
Bitrate is how much data a video spends on each second of picture, and it is the single largest influence on both quality and file size. The mode is the rule for how that spending is distributed across the video, and it is where most of the confusion lives.
There are three modes worth knowing. Two of them are usually presented as a pair, and the third, which is what most people actually want, often is not offered by name at all.
Constant bitrate
CBR spends the same number of bits on every second regardless of what is happening on screen.
Consider what that means for a twenty minute interview. Two people sit still against a plain wall. Almost nothing changes between frames, so the codec has very little to describe, and it spends the full allowance describing it anyway. Bits go into encoding a wall that has not moved.
Now the same video at the same setting during a handheld shot in the rain. Every pixel changes every frame, the codec needs far more than its allowance to describe that honestly, and it does not get more. Quality collapses exactly where the picture is most demanding.
So CBR wastes data where the picture is simple and starves it where the picture is complex, which is the opposite of what you would choose. It survives because it buys one thing nothing else can: total predictability.
Variable bitrate
VBR lets the rate move. Simple passages get less, complex passages get more, and the average lands near the target you set.
That is the sensible allocation, and for anything watched from a file rather than streamed live it produces a visibly better result at the same average rate. The interview stops wasting bits on a static wall and the rain scene stops falling apart.
There is a distinction inside VBR worth knowing. Single pass VBR encodes once and decides as it goes, which means it is guessing about material it has not seen yet. Two pass VBR analyses the whole video first, works out where the difficult passages are, then encodes with that knowledge. It takes roughly twice as long and distributes the budget considerably better, which matters most when the target rate is tight.
Constant quality, the one that is usually right
The third mode inverts the question. Instead of naming a rate and accepting whatever quality results, you name a quality and accept whatever file size results.
In most encoders this is a number called CRF, for constant rate factor. Lower means better quality and a larger file. The scale is not a percentage and the useful range is narrower than it looks: for H.264, most material sits between 18 and 28, with 23 as a common default. A change of about 6 roughly halves or doubles the file size.
The reason to prefer this is that it asks a question you can actually answer. Almost nobody knows what bitrate a particular clip needs. Everyone knows whether they want it to look excellent or merely acceptable. Constant quality lets you say the second and works out the first per scene.
Set a quality rather than a rate whenever you are not constrained to a specific file size. A five minute clip and a two hour film do not need the same bitrate, and constant quality is the mode that understands that without being told.
Which to use
| Situation | Mode | Why |
|---|---|---|
| Live streaming | CBR | The connection has a fixed ceiling and cannot absorb spikes |
| Uploading to a platform | Constant quality | It will re-encode anyway, so give it a clean source |
| Fitting a hard size limit | Two pass VBR | The only mode that reliably hits a target size |
| Archiving your own footage | Constant quality, low number | Quality is the point and storage is cheap |
| Sending to someone on a slow connection | Two pass VBR | Predictable size matters more than the last of the detail |
| Broadcast or hardware playback | CBR | Some decoders genuinely require a fixed rate |
Why live streaming is the exception
A live stream is limited by the connection carrying it, and that limit is a ceiling rather than an average. A variable encoder that averages 6 Mbps will happily produce a 15 Mbps burst during a difficult scene, and a connection that can sustain 8 Mbps will drop frames when it arrives.
The viewer sees stuttering or a stall. CBR never exceeds its rate, so nothing bursts, so nothing stalls. The wasted bits during quiet moments are the price of that, and for live delivery it is worth paying.
The trade you are making
Every mode gives up one of three things: predictable size, consistent quality, or efficient use of data. You cannot have all three.
- CBR gives up efficiency and quality consistency to guarantee the rate.
- VBR gives up an exact size to spend the data where it counts.
- Constant quality gives up any size prediction at all to keep the quality even throughout.
Knowing which of the three you actually need decides the mode in a sentence. If a specific number of megabytes matters, use two pass VBR. If the connection has a hard ceiling, use CBR. Otherwise use constant quality, because the thing you care about is how it looks.
Common questions
Is VBR always better quality than CBR?
At the same average bitrate, yes, for anything played back from a file. VBR spends more where the picture is complex and less where it is simple. CBR only wins where a fixed rate is a hard requirement, which in practice means live streaming and some broadcast or hardware decoders.
What CRF number should I use?
For H.264, 18 is close to visually lossless, 23 is a sensible default and 28 is noticeably compressed but usable. A change of about 6 roughly halves or doubles the file size. H.265 and AV1 use the same idea on a different scale, so their numbers are not directly comparable.
Why can constant quality not tell me the file size in advance?
Because the size depends entirely on how difficult the material turns out to be. A static presentation and a snowstorm at the same CRF can differ by a factor of ten. If you need a specific size, that is exactly what two pass VBR exists for.
Is two pass worth double the encoding time?
When you are aiming at a specific file size, yes, because the first pass is what makes the budget land where it should. When you are using constant quality, no, since there is no budget to distribute.