Bitrate is the amount of data used to represent one second of encoded video, usually expressed in kilobits or megabits per second (kbps or Mbps). In video codecs, it is the budget that an encoder spends on describing each frame, and it is the single most direct lever on file size, transmission bandwidth and visible quality.
Bitrate never describes quality on its own. The same number can look excellent with one codec and poor with another, because compression efficiency differs between formats such as MPEG-2, H.264, H.265 and AV1. Resolution, frame rate, motion and grain all change how many bits a given picture really needs.
How bitrate works inside video codecs
Bitrate is the result of every decision the encoder makes about how precisely to describe the picture. A video codec removes redundancy through prediction and transform coding: it predicts blocks from neighboring pixels or from earlier frames, transforms the remaining difference into frequency coefficients and then quantizes those coefficients. The quantization step is where most bits are saved and where most detail is lost.
The quantization parameter (QP) controls how coarse that rounding is. A lower QP keeps more detail and produces a higher bitrate; a higher QP discards fine texture and lowers the bitrate. Rate control is the part of the encoder that adjusts QP from frame to frame and block to block so that the output meets a target bitrate, a quality target or both.
Different frame types consume very different amounts of data. An I-frame or keyframe is coded without reference to other frames and is the most expensive. P-frames and B-frames reuse information from neighboring frames and can be many times smaller. The structure of the GOP therefore has a strong influence on the average bitrate and on how much it fluctuates.
| Frame type | Reference to other frames | Relative size | Role |
|---|---|---|---|
| I-frame | None | Largest | Random access, scene changes |
| P-frame | Earlier frames | Medium | Forward prediction |
| B-frame | Earlier and later frames | Smallest | Bidirectional prediction |
Rate control modes and bitrate behavior
Encoders distribute bits over time according to a rate control mode, and the choice of mode decides whether the bitrate stays flat or follows the content. Constant bitrate and variable bitrate are the two broad families, and each has several practical variants.
| Mode | What is held constant | Bitrate behavior | Typical use |
|---|---|---|---|
| Constant bitrate (CBR) | Data rate over a short window | Flat, padded if needed | Live broadcast, fixed channels |
| Variable bitrate (VBR) | Average data rate | Rises on complex scenes | Files, on-demand streaming |
| Capped VBR | Average with a ceiling | Variable below a peak limit | Adaptive streaming ladders |
| Constant quality (CRF or CQ) | Perceived quality level | Fully content-dependent | Archiving, offline encodes |
| Two-pass VBR | Average after analysis | Bits moved to hard scenes | Size-targeted delivery files |
Constant bitrate is predictable for networks and hardware decoders, but it wastes bits on simple scenes and starves complex ones. Variable bitrate spends more on action, water, foliage and film grain and less on static shots, which gives better quality at the same average size. Capped modes add a buffer model, often described as the VBV or HRD buffer, so that peaks never exceed what a decoder or connection can absorb.
Typical bitrate ranges by codec and resolution
Useful bitrates depend on the codec generation, resolution, frame rate and content complexity, so any table of values is a starting point rather than a rule. Newer codecs reach similar quality at lower rates: H.265 and VP9 often need roughly 30 to 50 percent less than H.264, and AV1 typically saves somewhat more, although results vary strongly with the encoder and its settings.
| Resolution | H.264 | H.265 or VP9 | AV1 |
|---|---|---|---|
| 720p | 2.5 to 5 Mbps | 1.5 to 3 Mbps | 1 to 2.5 Mbps |
| 1080p | 5 to 10 Mbps | 3 to 6 Mbps | 2 to 5 Mbps |
| 2160p (4K) | 25 to 45 Mbps | 12 to 25 Mbps | 10 to 20 Mbps |
Production formats sit in a completely different range. Intra-frame codecs such as ProRes and DNxHR code every frame as a full picture to keep editing responsive and generation loss low, so they run at hundreds of megabits per second, for example around 150 to 220 Mbps for ProRes 422 at 1080p and well over 700 Mbps for ProRes 422 HQ at 4K. These rates are meant for storage and editing, not for distribution.
Doubling the frame rate does not double the required bitrate, because consecutive frames become more similar and prediction works better. An increase of about 30 to 60 percent is a common observation for 50 or 60 fps compared with 25 or 30 fps, but fast motion pushes it higher.
Bitrate compared with related measures
Bitrate is often confused with other quantities that describe image quality or data volume. Separating them helps when reading encoder settings, camera specifications and streaming reports.
| Concept | What it measures | Relationship to bitrate |
|---|---|---|
| Resolution | Pixels per frame | More pixels usually need more bits |
| Frame rate | Frames per second | More frames raise bitrate sublinearly |
| Bit depth | Bits per color sample | 10-bit adds modest overhead, reduces banding |
| Bits per pixel | Bitrate divided by pixel rate | Normalized comparison across formats |
| File size | Total stored data | Bitrate multiplied by duration |
| Bandwidth | Capacity of a connection | Must exceed peak bitrate plus overhead |
Bits per pixel is a helpful normalization: divide the bitrate by width, height and frame rate. Values around 0.05 to 0.1 bits per pixel are common for efficient H.264 delivery of ordinary content, while noisy or high-motion material needs more. File size follows directly from bitrate: 8 Mbps for 60 minutes produces about 3.6 gigabytes, before audio and container overhead.
Choosing a bitrate in practice
The right bitrate is the lowest rate at which the target audience sees no objectionable artifacts on its devices and connections. Engineers usually choose it by encoding test clips at several rates and comparing them visually and with objective metrics such as VMAF, PSNR or SSIM.
| Situation | Recommended approach | Reason |
|---|---|---|
| Live stream to a platform | CBR or tightly capped VBR | Stable upload and ingest |
| On-demand library | Two-pass or capped VBR ladder | Best quality per byte |
| Archive master | Constant quality or intra-frame codec | Quality first, size second |
| Grainy or high-motion film | Raise the rate or apply denoising | Grain is costly to encode |
| Screen recording | Lower rate, constant quality | Static content compresses well |
Encoder presets matter as much as the number itself. A slower preset searches more motion vectors and coding modes and can deliver the same quality at a noticeably lower bitrate, while hardware encoders such as NVENC trade some efficiency for speed.
Common bitrate mistakes and artifacts
Most visible compression problems come from a bitrate that is too low for the content. The typical symptoms are blocking in flat areas, banding in skies and gradients, mosquito noise around edges and smeared detail during fast motion.
- Comparing bitrates across codecs as if they were equal leads to wrong conclusions about quality.
- Re-encoding a low-bitrate file at a higher rate cannot restore lost detail; it only enlarges the file.
- Ignoring peak bitrate causes buffering even when the average fits the connection.
- Using 8-bit encoding for smooth gradients at low rates makes banding far more likely than 10-bit encoding.
A plain-language summary
Bitrate is how much information a video is allowed to use every second. Think of it as the number of words available to describe each picture. With plenty of words, the description can include every leaf and every strand of hair. With very few words, the description becomes vague, and the picture turns blocky or blurry.
A higher bitrate usually means better picture quality but bigger files and more internet data. Modern formats are simply better writers: they can say the same thing in fewer words, so a newer format at a lower bitrate can look as good as an older format at a higher one.
An example from practice
A small production company prepares a 20-minute documentary for an online video platform. The edit is finished in ProRes 422 HQ at 1080p and 25 fps, a master file of roughly 25 gigabytes. The team needs a delivery file that uploads quickly but keeps the dark interview scenes free of banding.
The editor encodes three test versions in H.264 using two-pass VBR at 6, 10 and 16 Mbps and compares them on a laptop and a large television. At 6 Mbps the shadows show blocking; at 10 Mbps the interviews look clean, but a handheld forest scene still smears. The final choice is 10 Mbps average with a 16 Mbps peak cap in 10-bit H.265. The result is a file of about 1.6 gigabytes that looks nearly identical to the master on both screens and uploads in minutes.