An audio file size calculator is one of the rare media calculations that is exact rather than approximate. Uncompressed audio has no variable content-dependent component: a second of 44.1 kHz, 16-bit stereo is 176,400 bytes whether it contains an orchestra or silence. Multiply by duration and you have the answer, to the byte, before you record anything.
That certainty is why this page separates the two cases so firmly. PCM size is arithmetic. Compressed size at a fixed bitrate is also arithmetic, on a different quantity. Everything else — variable bitrate, lossless compression ratios, container overhead — is estimation, and the page says which is which. Arb Digital publishes free tools for people who plan storage and bandwidth before committing to a format, and being exact about which numbers are exact is the point.
What This Audio File Size Calculator Does
Choose uncompressed or compressed, set the parameters, enter a duration, and it returns the file size along with the data rate in kilobits per second, the size of one minute, the size of one hour, and the same audio in the other format so the compression ratio is visible.
Three related tools sit next to this one on the site and the split between them is clean. The video file size calculator works from a video bitrate and duration, or solves the bitrate that fits a target size — video streams only, no per-sample arithmetic because video codecs have no equivalent of raw PCM in ordinary use. The image file size calculator works from resolution, colour channels, bit depth and a compression ratio for still images. This page owns the audio case: sample rate × bit depth × channels × duration, plus the common codec bitrates.
How to Use It
- Pick the format you are producing. Uncompressed for a recording master or an editing session, compressed for anything being delivered over a network.
- Set the sample rate. 44,100 Hz for music, 48,000 Hz for anything going alongside video. Higher rates are for capture and processing rather than delivery.
- Set the bit depth and channels. Both multiply the data rate directly. Mono is exactly half of stereo, and 24-bit is exactly 1.5 times 16-bit.
- Choose a bitrate for the comparison. Even in PCM mode this drives the compressed figure, which is what shows you the ratio.
- Enter the duration. Hours, minutes and seconds. Per-minute and per-hour figures update independently, which is useful for planning a series.
The Formula — How It's Calculated
Uncompressed PCM. Every sample of every channel occupies the bit depth. So the data rate in bits per second is sample rate × bit depth × channels, and the file size in bytes is that divided by eight, times the duration in seconds. Equivalently: bytes = sample rate × (bit depth ÷ 8) × channels × seconds.
Compressed at a constant bitrate. The bitrate already is the data rate, so bytes = bitrate in bits per second ÷ 8 × seconds. A 128 kbps file is 16,000 bytes per second regardless of sample rate, bit depth or channel count — those parameters are inputs to the encoder, not to the output size.
Megabytes here are decimal: one MB is 1,000,000 bytes. That is the convention used by storage vendors, hosting plans and most operating systems. Some file managers divide by 1,048,576 instead and call the result MB when it is properly a mebibyte, MiB. The subtitle shows both so the discrepancy is never a surprise.
Worked example, using the values the page loads with. 44,100 Hz, 16-bit, stereo, 3 minutes 30 seconds — 210 seconds. The data rate is 44,100 × 16 × 2 = 1,411,200 bits per second, which is 1,411.2 kbps, the familiar CD figure. The size is 44,100 × 2 bytes × 2 channels × 210 = 37,044,000 bytes, or 37.04 MB — which your file manager may display as 35.33 MiB. One minute of the same audio is 10,584,000 bytes, or 10.58 MB, and one hour is 635.04 MB. The same 210 seconds at 192 kbps is 24,000 bytes per second × 210 = 5.04 MB, a compression ratio of about 7.35 to 1.
Why 44.1 kHz Exists, and What Sample Rate Actually Buys
Sampling theory sets a hard limit: a sampled signal can only represent frequencies below half the sample rate. That half-rate boundary is the Nyquist frequency, and content above it does not merely get lost — it folds back down into the audible range as aliasing distortion, which is why every converter has a filter in front of it.
Human hearing tops out around 20 kHz, so a sample rate of at least 40 kHz is required. 44.1 kHz leaves 2.05 kHz of headroom for the anti-aliasing filter to roll off in, and the specific value comes from the video-tape-based digital recorders that early digital audio was stored on. 48 kHz became the standard for film and video, which is why an audio file at 44.1 kHz dropped into a video timeline needs resampling.
The practical consequence for file size is that doubling the sample rate doubles the file and adds frequency content nobody can hear. High rates are genuinely useful during capture and processing — filters behave better with more headroom, and pitch-shifting or time-stretching benefits — but for delivery they are pure size. The note frequency converter covers the frequency side of the same domain, and the sound level converter handles amplitude units.
Bit Depth Is Dynamic Range, Not Resolution
Bit depth is routinely described as audio "resolution", which invites the wrong mental picture. What it sets is the noise floor. Each additional bit roughly halves the quantisation noise, worth approximately 6 dB of dynamic range, so 16-bit provides about 96 dB and 24-bit about 144 dB.
144 dB exceeds the dynamic range of any playback system and most rooms. The reason 24-bit is standard for recording is not that the extra range is audible in the finished product; it is that it lets an engineer record with generous headroom, keeping peaks far from clipping, and still have a noise floor far below anything audible after the level is raised in the mix. That is a workflow benefit at capture, not a delivery benefit.
32-bit float is a different thing again. It is not primarily about range but about arithmetic: floating-point samples do not clip in the same way during processing, so intermediate calculations in a mixing chain stay clean. It is a working format, and delivering in it just doubles the file relative to 16-bit for no benefit at the listener's end.
What Compressed Bitrates Actually Deliver
Codec efficiency varies enormously, and a bitrate number means little without the codec attached. Opus, defined in IETF RFC 6716, "Definition of the Opus Audio Codec", supports every bitrate from 6 kbit/s to 510 kbit/s and the specification names its own sweet spots: 8–12 kbit/s for narrowband speech, 16–20 for wideband speech, 28–40 for fullband speech, 48–64 for fullband mono music and 64–128 for fullband stereo music.
Read that last figure against MP3 practice, where 192 to 320 kbps is the normal range for stereo music, and the generational gap is obvious. A modern codec at 96 kbps is doing work that an older one needed two or three times the data for. If you are choosing a delivery format today and compatibility permits it, the bitrate you need is substantially lower than the numbers that became habitual in the MP3 era.
For uncompressed transport the parameters are standardised too. RFC 3551, the RTP Profile for Audio and Video Conferences with Minimal Control, defines static payload types with their clock rates and channel counts — L16 linear PCM at 44,100 Hz appears in both one-channel and two-channel forms, which is exactly the calculation this page performs.
Where the Estimate Stops Being Exact
Three things push a real file off the calculated number, all upward.
Container overhead. A WAV file carries a RIFF header of at least 44 bytes, and often more once metadata chunks are added. On a three-minute file that is invisible. On ten thousand short interface sounds it is not, and it is the reason very short clips are proportionally larger than the arithmetic suggests.
Variable bitrate. Most modern encoding is VBR, which spends more data on complex passages and less on simple ones. The target bitrate is an average across the file, so the result lands near the calculated size but rarely on it. Quiet spoken-word material typically comes in under target; dense music can exceed it.
Lossless compression. FLAC and ALAC reconstruct the original samples exactly, so their output is content-dependent and cannot be calculated in advance. Typical results are 50 to 70 per cent of the PCM size, but a recording of near-silence compresses far further and a dense, loud master far less. Use the PCM figure as the ceiling and treat any lossless number as an estimate. The data storage converter handles the MB-versus-MiB conversion in isolation.
Turning Size Into Bandwidth and Time
File size and streaming bandwidth are the same quantity viewed differently. A file that streams in real time needs sustained throughput equal to its data rate — that CD-quality 1,411.2 kbps stream needs about 1.4 Mbps continuously, which is a substantial fraction of many mobile connections, whereas a 96 kbps Opus podcast needs a hundredth of a megabit and will hold up on almost anything.
For downloads, divide file size in bits by connection speed in bits per second. The 37.04 MB PCM example is 296.4 megabits, so on a 25 Mbps connection it takes just under twelve seconds at full speed. The 5.04 MB compressed version takes about 1.6 seconds. For a podcast archive of a hundred hour-long episodes, that difference is 63.5 GB against 0.86 GB, and it decides whether you are paying for a CDN tier or not. The bandwidth requirement calculator and the streaming bitrate calculator take that further, and the audiobook time calculator handles the listening-duration side.
Arb Digital's web design team builds sites where media is sized, encoded and delivered properly, so pages load fast without sacrificing quality.
See Web Design Services Talk to Arb DigitalCommon Mistakes to Avoid
- Assuming sample rate affects a compressed file's size — at a fixed bitrate the output size depends on bitrate and duration only.
- Mixing MB and MiB — the same file is 37.04 MB or 35.33 MiB, and the 4.9% gap is enough to break a storage estimate at scale.
- Expecting a VBR file to match the calculation — the target bitrate is an average, so the result lands nearby rather than exactly.
- Delivering at capture settings — 96 kHz 24-bit is a working format; shipping it to listeners multiplies size for content they cannot hear.
- Ignoring header overhead on very short clips — a 44-byte WAV header is nothing on a song and significant across thousands of interface sounds.
Related Free Tools From Arb Digital
Use the video file size calculator for video streams and target-size bitrates, the image file size calculator for stills, the streaming bitrate calculator for live upload requirements, the data storage converter for byte-unit conversions, and the bandwidth requirement calculator when you are sizing a connection rather than a file. Everything else is in the free online tools hub.
Frequently Asked Questions
For uncompressed PCM, multiply sample rate by bit depth by channel count to get bits per second, divide by eight for bytes per second, then multiply by the duration in seconds. For a fixed-bitrate compressed file, divide the bitrate by eight and multiply by the duration.
44,100 Hz, 16-bit stereo runs at 1,411.2 kbps, which is 176,400 bytes per second. One minute is 10,584,000 bytes, or 10.58 MB. One hour is 635.04 MB.
No. At a fixed bitrate the output size depends only on the bitrate and the duration. Sample rate, bit depth and channel count are inputs to the encoder and affect what it can represent, not how many bytes it writes.
Because it is probably dividing by 1,048,576 rather than 1,000,000 and labelling the result MB when it is properly MiB. The same file is 37.04 MB or 35.33 MiB. The subtitle on this page shows both figures.
The Opus specification names 28 to 40 kbit/s as a sweet spot for fullband speech and 48 to 64 kbit/s for fullband mono music. Older codecs such as MP3 need substantially more data for comparable results, which is why 96 to 128 kbps remains common where compatibility matters.
Not exactly. Lossless compression is content-dependent, so the ratio varies with the material — typically 50 to 70 per cent of the PCM size, but much better on quiet content and much worse on dense loud masters. Use the PCM figure as an upper bound.
Bit depth sets the noise floor rather than detail: about 96 dB of dynamic range at 16-bit and about 144 dB at 24-bit. The extra range is a recording and mixing benefit — it allows generous headroom during capture — rather than something a listener gains from in a finished file.
The same as its data rate. CD-quality PCM needs about 1.4 Mbps sustained, while a 96 kbps compressed stream needs about a hundredth of a megabit per second, which is why compressed delivery is the norm for anything going over a network.
Uncompressed PCM sizes on this page are exact arithmetic. Compressed figures assume a constant bitrate, and real variable-bitrate and lossless encodes will differ, so treat those as planning estimates rather than guarantees.
