File size calculator

Every answer on this page comes from one relationship: bits per second times seconds gives bits. All the mistakes come from the units on either side of it.

The factor of eight, and the other one underneath it

Bitrate is quoted in bits and file size in bytes, so eight is where most wrong answers start. An 8,000 kbps stream is a thousand kilobytes a second, not eight thousand, and 45 minutes of it comes to about 2.51 GB instead of 20. It is the same confusion that makes a "100 megabit" line download at roughly 12 megabytes a second.

Underneath that sits a smaller ambiguity: whether a megabyte is 1,000,000 bytes or 1,048,576. Storage manufacturers use the decimal figure and operating systems have historically shown the binary one, and the gap is about 5% at the megabyte and 7% at the gigabyte. This page uses binary megabytes, so the answer matches what a file manager reports; a file showing 2.5 GB on disk is about 2.68 decimal gigabytes on a drive label.

Hitting a size target

A single encoding pass has to guess how to spend its bits before it has seen the footage, and typically overshoots or undershoots by several per cent. Two-pass analyses the whole file first and then distributes the bits toward the complex sections, and for an upload ceiling like a mail attachment or a Discord limit that difference is the difference between a file that fits and one you encode twice.

The figure this page gives is the total, so take your audio off it to get the video allowance, typically 128 to 320 kbps for stereo. Leave one or two per cent of margin as well, because container overhead and encoder variance both land on the wrong side of a hard limit.

Cards hold less than the arithmetic says

Recording adds a figure that is not in the equation at all. A 100 Mbps camera needs 12.5 MB/s of sustained write, and card speeds are quoted in megabytes while camera bitrates are quoted in megabits, so the comparison is not the one on the packaging. That is what the V-ratings are for: V30 and V60 guarantee 30 and 60 MB/s sustained, and the sustained figure is the one to read.

The capacity figure needs reading carefully too. A card sold as 128 GB holds 128 decimal gigabytes, which an operating system shows as about 119; that is the same bytes counted differently, not less card, and the answer here counts the full 128. What genuinely costs you space is the filesystem, a per cent or two. What costs you a take is write performance, which degrades as a card fills and as it ages, and on a marginal card can stop a recording partway through. Professional practice is to swap cards at about 80% instead of filling them, which gives up a little capacity and removes a whole class of failure.

Uncompressed audio

This is the one case where the bitrate is a consequence, not a setting: sample rate times bit depth times channels. CD audio is 44,100 × 16 × 2, which is 1,411 kbps, and that single figure is why 74 minutes of it takes about 750 MB.

Nyquist sets the ceiling at half the sample rate, but not the practical one. A real converter needs an anti-aliasing filter below that half, and no filter is infinitely steep, so 44.1 kHz gives clean response to about 20 kHz rather than a hard 22.05. That margin is exactly why 44.1 was chosen instead of 40. Video uses 48 kHz for a separate reason: it has a clean relationship with common frame rates, which simplifies synchronisation.

Higher sample rates cost storage linearly and deliver much less than the numbers suggest for finished audio, since 44.1 already captures everything above the range of human hearing. Where they genuinely help is during processing, so studios often work at 96 kHz and deliver at 48. Bit depth is the more consequential choice when recording, because 24-bit leaves far more room for a conservative input level.