The Only Formula You Need
Uncompressed audio size is pure arithmetic: sample rate × (bit depth ÷ 8) × channels × seconds = bytes. A CD-quality stereo track — 44,100 samples per second, 16 bits per sample, 2 channels — eats 176,400 bytes per second, which is why a 3-minute song is about 30 MB and a 74-minute CD holds 650 MB. Every number on this page falls out of that one formula; the format buttons just apply each codec's compression on top.
Note the units: storage is usually quoted in megabytes (MB), while streaming and upload speeds are megabits (Mbps) — eight times smaller. A 30 MB file on a 25 Mbps connection takes about 10 seconds to upload, not one. The uploader estimate above does this conversion for you.
Sample Rate: How High Is High Enough?
Sample rate sets the highest frequency you can capture — the Nyquist limit is exactly half the rate, so 44.1 kHz captures up to 22.05 kHz, just past human hearing. 44.1 kHz remains the music-delivery standard (CDs, most streaming masters). 48 kHz is the film, TV, and video standard — if your audio will ever touch video, record at 48 to avoid a sample-rate conversion later.
88.2/96 kHz double the data for benefits most listeners cannot hear on playback, but they genuinely help processing: pitch-shifting, time-stretching, and heavy plugin chains accumulate fewer digital artifacts at high rates. Many engineers track at 96 kHz and deliver at 44.1/48. 192 kHz is essentially archival or sound-design territory — four times the storage of 48 kHz for inaudible gains in normal music work. Your hard drive will notice before your ears do.
Bit Depth: The Dynamic Range Dial
Bit depth sets the dynamic range — roughly 6 dB per bit, so 16-bit gives ~96 dB and 24-bit gives ~144 dB. 16-bit is the CD standard and perfectly fine for finished, properly dithered masters. 24-bit is the recording standard because its huge headroom forgives imperfect gain staging: record a little quiet and boost later with no noise penalty.
32-bit float — common inside DAWs — effectively cannot clip; you can record absurdly hot or quiet and normalize afterward. It costs 50% more storage than 24-bit, so it is a working format, not a delivery one. Never deliver 32-bit float to a distributor unless asked; 24-bit WAV (or 16-bit with dither) is the universal handoff.
Compressed Formats: What You Actually Lose
FLAC (and ALAC) are lossless: typically 50–60% of WAV size with bit-identical audio. Archive everything in FLAC, not MP3 — future you, with better speakers, will be grateful. MP3 320 kbps is the "transparent enough" lossy standard: a 3-minute song shrinks from 30 MB to ~7 MB, and most listeners cannot reliably tell it from the WAV in blind tests. AAC 256k (Apple/YouTube/Spotify's family) beats MP3 at equal bitrates, which is why streamers standardized on it.
The real cost of lossy is generational: every re-encode of an MP3 degrades further, while WAV→FLAC→WAV is a perfect round trip. Rule: keep a lossless master, encode lossy copies from it, never from each other.
Multitrack Sessions: Where Storage Really Goes
A finished mix is megabytes; the session is gigabytes. Eight tracks of 24-bit/48 kHz for an hour is nearly 4 GB — and that is before takes, playlists, and backups multiply it. The session stacker above totals your tracks so the number stops being a surprise. Two practical rules fall out: record at the lowest rate/depth your work justifies (24/48 for most music), and commit and archive — bounce virtual instruments to audio and move dead takes to cold storage.
SSDs changed the game for active sessions (fast seeks, silent), but spinning drives or cloud cold storage remain the economical archive. Whatever the medium: 3-2-1 backup — three copies, two media types, one offsite. Sessions are irreplaceable; drives are not.
Streaming Bitrates and Delivery Specs
Know your destination's spec before you bounce. Spotify normalizes loudness to −14 LUFS and transcodes uploads to Ogg Vorbis/AAC; Apple Music prefers 24-bit masters for "Apple Digital Masters". YouTube re-encodes everything to AAC. Podcast hosts typically want MP3 128k mono for speech (half the size of stereo, and nobody misses stereo on a voice). Audiobooks (ACX) demand specific RMS and peak specs — check before you master.
For client delivery, 24-bit/48 kHz WAV is the safest universal master: every video editor, mastering engineer, and distributor accepts it. Include a 16-bit/44.1 kHz dithered copy for CD replication if needed.
Upload Time: Plan the Handoff
Big files meet slow connections at the worst moment — deadline night. A 4 GB session on a 25 Mbps upload takes over 20 minutes; on 5 Mbps it is nearly two hours. The estimator above tells you before you start. If the math is ugly, options include: delivering stems instead of full sessions, using FLAC to halve the payload losslessly, or shipping a drive for truly massive projects. And always verify the upload — checksums or at minimum a file-size comparison — because a truncated WAV fails silently until mastering day.
Common Mistakes
Mistake one: recording at 192 kHz "for quality" on a laptop with 256 GB of storage — the session fills the drive mid-take. Mistake two: converting MP3 to WAV to "restore quality" — the lost data is gone; the WAV is just a bigger MP3. Mistake three: normalizing to 0 dBFS for streaming, where lossy transcoders then clip — leave a dB of headroom. Mistake four: archiving only the mix and deleting the session, then the client asks for "the vocal up 1 dB" two years later.
Dither: The Final Step Nobody Sees
When you reduce bit depth — bouncing a 24-bit mix to 16-bit for CD, for example — you must dither. Truncating bits without dither creates quantization distortion: harsh, correlated artifacts most audible in quiet fades and reverb tails. Dither adds a tiny amount of shaped noise that randomizes the quantization error, replacing distortion with a smooth, inaudible noise floor. It feels backwards — adding noise to improve quality — but the math is settled and every professional master uses it.
The rules are simple: dither once, at the very end, when going down in bit depth — never when staying at the same depth, and never twice. Going 24-bit to 16-bit gets dither; 24-bit to 24-bit does not. Most DAWs and limiters include POW-r or MBIT+ options; any of them beats truncation. And critically, dither after the final limiter in the chain — it is literally the last process before the file is written. Get this right and your quiet passages stay silky; skip it and fades get grainy.
Putting It All Together
The workflow: track at 24-bit/48 kHz (96 if you process heavily), keep lossless masters in FLAC, deliver 24/48 WAV to clients, encode MP3/AAC copies from the master for listeners, and run the session stacker before the drive fills up. Audio storage is cheap until it is not — usually at 2 a.m. before a deadline. A minute with this calculator beats an hour of cleanup.
Archiving Sessions: The 3-2-1 Rule
Knowing file sizes is only half the battle; keeping the files safe is the other half. Professional studios follow the 3-2-1 rule: three copies of every session, on two different media types, with one copy offsite. A finished album project with multitracks, mixes and masters can easily run 50–100 GB, which means a single 1 TB drive holds roughly ten to twenty projects — fewer than most producers expect. Use the session stacker above to estimate each project's footprint, then size your backup drives with headroom to spare.
For archiving, keep the uncompressed masters (WAV or AIFF) and the original multitrack stems; you can always re-encode smaller delivery formats later, but you can never recover quality discarded at export. FLAC is an excellent compromise for archiving mixes — typically 50–60% smaller than WAV with zero quality loss. Label everything with dates and sample rates, because a folder named "final_mix_v2_REAL.wav" helps nobody in three years.