CBR vs VBR: Constant vs Variable Bitrate

Constant Bitrate allocates exactly the same number of bits to every second of audio. A 192 kbps CBR file uses 192 kilobits for every single second, whether that second is a dense orchestral climax or near silence. Its advantages are predictability and simplicity: the file size is known in advance (bitrate times duration), the data rate is steady, and any player — including very old or very simple ones — can handle it and report accurate duration and seek positions. Its disadvantage is inefficiency: on simple passages it spends bits it does not need, and on the most complex passages it may not have enough, so quality dips exactly where the music is hardest to encode. CBR is the older, more conservative approach, and it remains common in streaming and broadcast where a guaranteed, constant data rate matters more than squeezing out the last bit of quality-per-byte.

Variable Bitrate lets the encoder change the data rate moment to moment based on how hard the audio is to compress. A dense, transient-rich section might momentarily use 300+ kbps, while a sparse passage or silence drops to 100 kbps or less. The goal is consistent quality rather than consistent size: the encoder aims to keep perceived quality even by spending bits where the ear will notice and saving them where it will not. Because bits follow demand, VBR delivers better average sound quality at the same overall file size than CBR, or equivalently the same quality at a smaller size. Modern VBR encoders — LAME for MP3 being the classic example — have been refined over two decades of listening tests to make excellent allocation decisions, which is why VBR has been the recommended way to encode MP3 since the early 2000s. The trade-off is that final file size is not known until encoding finishes.

Average Bitrate is a hybrid that tries to capture VBR's efficiency with CBR's predictability. You set a target average bitrate, and the encoder varies the instantaneous rate around it — spending more on hard passages and less on easy ones — while steering the whole file toward your target average. Think of it as VBR working within a file-size budget. ABR gives more consistent, predictable sizes than pure quality-based VBR while still adapting to content, so it sits between the two in both quality and predictability. In practice it is the least-used of the three: when you care most about quality you pick VBR, and when you need exact size or rate control you pick CBR, leaving ABR as a compromise for situations that need a rough size target without the rigidity of CBR — for instance fitting an audiobook to a storage budget while keeping quality reasonable.

VBR's intelligence comes from the same psychoacoustic model that makes lossy compression work at all. As the encoder analyzes each short frame of audio, the model estimates how much quantization noise can be hidden — masked — by the sounds present, based on how human hearing works: loud tones conceal quieter nearby tones, transients mask sounds just before and after them, and the ear is insensitive at frequency extremes. A complex frame with lots of un-maskable detail (a cymbal crash, a busy mix) demands many bits to stay clean; a simple frame (a sustained note, a pause) can be encoded accurately with far fewer. VBR simply lets the bit count follow that demand frame by frame, whereas CBR forces a fixed budget regardless. This is why VBR is more efficient: it aligns spending with where the ear actually needs precision. The quality of a VBR file therefore depends heavily on the encoder's model — which is why a modern LAME encode sounds better than an old one at the same setting.

At the same average bitrate, VBR sounds as good as or better than CBR, a result established through decades of blind listening tests. VBR spends bits where they matter and saves them where they do not; CBR wastes bits on easy passages and can starve the hardest ones. The gap is most pronounced at low and medium bitrates — around 128 kbps and below — where every bit counts and CBR's fixed budget shows its limits. At very high bitrates like 320 kbps CBR there is so much headroom that even the hardest passages are well-served, so CBR and VBR both sound essentially transparent and the efficiency advantage of VBR mostly turns into smaller files rather than audibly better sound. The practical implication: if you want the best quality per megabyte, VBR wins clearly at moderate bitrates; if you are already encoding at maximum bitrate, the choice barely affects sound and comes down to whether you value predictable size (CBR) or a slightly smaller file (VBR).

The core practical difference between the modes is knowing the file size in advance. CBR size is simple arithmetic: bitrate times duration, so a 4-minute track at 192 kbps CBR is almost exactly 5.76 MB every time. That predictability matters for storage budgets, broadcast slots, and streaming buffers. VBR size cannot be known until encoding completes, because it depends on the music's complexity — a sparse acoustic track and a dense metal track encoded at the same VBR quality setting will differ in size. This unpredictability is usually a non-issue for personal libraries, where you care about quality and total disk usage rather than per-file size, but it can complicate systems that pre-allocate space or bill by exact size. ABR splits the difference by targeting an average, giving roughly predictable sizes with some content adaptation. For most people encoding music to keep, VBR's variable size is a fair price for better quality-per-byte; for pipelines that must know sizes ahead of time, CBR or ABR is safer.

VBR once had a real compatibility pitfall around seeking and duration, and understanding it explains lingering advice to 'use CBR for compatibility.' Because a VBR file's data rate varies, a player cannot calculate a timestamp's byte position by simple arithmetic the way it can with CBR. To seek accurately and display the correct total duration, the player relies on a small header at the start of the file — the Xing or Info header for MP3 (VBRI for some encoders) — that maps time to file position. Well-made VBR files include this header and behave perfectly. But older or poorly-written players that ignored the header would show wrong durations, seek to the wrong place, or fail to scrub through VBR files. Modern software and hardware handle VBR correctly, and every reputable encoder writes the header, so this is largely a solved historical problem — but it is why VBR earned a reputation for flakiness, and why CBR is still occasionally specified for maximum compatibility with legacy or minimal players.

For MP3, the LAME encoder's quality-based VBR presets are the gold standard, selected with the -V switch from V0 (highest quality) to V9 (lowest). Each targets a quality level and lets the bitrate float to achieve it. V0 averages roughly 245 kbps and is transparent for nearly all listeners on nearly all material — the choice when you want top MP3 quality without going to 320 CBR. V2 averages about 190 kbps and is transparent for most listeners, widely considered the sweet spot of quality versus size. V4 averages about 165 kbps for good casual quality, and V6 around 115 kbps for space-saving encodes with audible compromises on demanding music. Because these are quality targets, a simple track encoded at V2 comes out smaller than a complex one at the same setting, yet both aim for the same perceived quality. For most people ripping or converting music to MP3, V0 or V2 gives better results per megabyte than any CBR setting short of 320 kbps.

The CBR/VBR distinction plays out differently across codecs. MP3 supports CBR, ABR, and VBR, with LAME's VBR presets being the recommended path and the Xing header handling seeking. AAC likewise supports CBR and true VBR; Apple's encoder offers a 'True VBR' mode alongside constant and constrained options, and AAC VBR is well-supported on modern devices. Opus is variable-bitrate by design — its default is VBR, with a constrained-VBR mode for when a rough rate ceiling is needed and CBR available mainly for fixed-bandwidth links; for Opus you normally just set a target quality and let it vary. Vorbis is similarly VBR-first, encoded by a quality level rather than a fixed rate. Lossless formats like FLAC are inherently variable (compression follows content) and the CBR/VBR terminology does not really apply. The general trend is that newer codecs assume VBR because it is simply more efficient, and reserve constant-rate modes for streaming and transmission scenarios that need predictable bandwidth.

For music files and personal libraries, use VBR — it delivers the best quality per megabyte, and modern players handle it flawlessly. For MP3 specifically, LAME -V0 or -V2 is the recommended setting. Use CBR when you need predictable file size or a guaranteed constant data rate: internet radio and live streaming, broadcast, some podcast hosts that specify a fixed bitrate, and the rare legacy or minimal player that mishandles VBR seeking. Use ABR when you want content-adaptive quality but still need to hit a rough size target, such as fitting long audio to a storage limit. When in doubt for a file you intend to keep and listen to, choose VBR; when in doubt for something you must stream or transmit at a fixed rate, choose CBR. For modern codecs like Opus and AAC, simply pick a quality or bitrate target and let their native variable-rate engines do the rest.