H.264 Vs H.265 Vs AV1: The Same Video At Half The Size

The Question In The Creators Group

A few weeks after my own drive incident, a course creator in a local creators group posted the problem in one line: his 4 TB archive drive had filled in eight months, he could not delete a single lesson, and a second drive felt like admitting defeat. I knew that arithmetic intimately, because I had just lived the mirror image of it. His lessons were fine. His codec was the heavy part. He archived everything in H.264 at full upload bitrate, the default his editor offered, and defaults are where drives go to fill.

The short answer I gave him, and then tested on my own client's course the following month: at the same picture quality, H.265 needs roughly half the H.264 bitrate, and AV1 lands between the two in the model this page uses, at about 65 percent.

Since file size is bitrate times duration, the sizes follow the same ratios. A ten minute 1080p lesson that weighs 624 MB in H.264 weighs about 324 MB in H.265 and about 414 MB in AV1. The rest of this article is the test I ran, and the two costs that the smaller numbers hide.

One Lesson, Exported Three Ways

I took one finished lesson from the course, ten minutes of slides and presenter at 1080p, and exported it three times, changing only the codec. The delivery audio stayed at 320 kbps in all three files. The calculator's codec model applies a factor to the video bitrate: H.264 keeps the entered 8 Mbps, H.265 scales it to 4 Mbps, AV1 scales it to 5.2 Mbps. Those factors are approximations, and I treat them that way, but they matched my exports closely enough to plan storage with.

The Fraunhofer HHI institute, which helped build the HEVC standard, states the design result plainly: compared with H.264, HEVC achieves about a 50 percent bit rate reduction at the same subjective video quality. My export test is that sentence with a file attached. AV1's 35 percent saving in this model is smaller, and that surprised me until I remembered that the factor is a planning approximation, not a promise; on some content AV1 pulls level with HEVC, and on talking-head lessons like mine it trails a little.

A Month Of Lessons In Three Codecs

One lesson is a rounding error. A course is not. The client published twenty lessons that month, each about ten minutes, and the archive choice priced out like this: H.264 kept 12,480 MB, about 12.48 GB. AV1 kept 8,280 MB, about 8.28 GB. H.265 kept 6,480 MB, about 6.48 GB. The H.265 archive saved 6 GB against the default in a single month, on a single course, and the saving compounds with every month the library grows. Per hour of finished video, the same figures read 3.74 GB, 2.48 GB and 1.94 GB.

This is also where I point readers who arrive with the opposite problem, a single file that will not upload. Codec is the second lever, after exporting a delivery file instead of a master. I set out that first lever, with the 624 MB or 11 GB comparison, in The Article On Why A 1080p File Can Be 600 MB Or 11 GB. Codec choice decides what happens after the master is out of the delivery folder: how fast the archive grows, and what it costs to keep.

Half the storage, twice the export time, and one angry email from a student with an old tablet. That is the real codec comparison.

Run the same multiplication across a year and the codec stops being a settings detail. That client publishes about four courses a year at the pace of that month. Four months of lessons at these rates total 49,920 MB in H.264, about 49.92 GB, against 25,920 MB, about 25.92 GB, in H.265. AV1 lands between, at 33,120 MB.

One course library is a rounding error on a modern drive; four years of course libraries, with masters and project files beside them, is where a creator ends up posting that question in a group chat. The codec decision is really a decision about how fast a drive fills, made once and paid monthly.

Anyone can run the manual version of this check on a single file before committing a library to it. Take the file's size, and for a planning estimate of its H.265 twin, halve it; for AV1, take about two thirds. A 900 MB H.264 lecture plans at roughly 450 MB in H.265 and roughly 600 MB in AV1. The estimate will not match a real export to the megabyte, because encoders respond to content, but it is close enough to choose a drive, a plan tier or an upload window, which is what planning numbers are for.

One more wrinkle from that month of exports deserves a place here, because it changes what delivery codec even means. The file I hand to a platform is not the file a viewer streams.

Platforms re-encode uploads into their own ladders of versions, and the big video platforms serve newer codecs, VP9 and AV1 among them, to devices that support them. So the delivery file's job is to be a clean, compatible source at a sane bitrate, not to be the final word in compression. The archive is mine to squeeze. The stream is the platform's problem, and it has far bigger computers for it.

What The Saving Actually Costs

The angry email in that quote is a real one. A student tried to watch a lesson on an old tablet, and the H.265 file would not play. H.264 decodes on effectively every device made in the last fifteen years.

H.265 needs newer hardware for smooth playback, and plenty of budget Android tablets in Pakistan still struggle with it or refuse outright. AV1 sits one step further along the same road: browser and platform support is strong and growing, and the codec is royalty-free, developed by the Alliance for Open Media, but hardware decoding lives on recent devices, so older phones decode it in software and pay in battery and heat.

The export time row deserves honesty. On my older laptop, the month's H.265 re-exports ran overnight, roughly twice the H.264 time, and AV1 encodes slower still. That trade is worth making for an archive I touch rarely. It is a poor trade for a file I will reopen and trim next week, because editing a heavily compressed codec makes the machine work for every frame. Storage is cheap once. Time and compatibility bill me every month.

Upload time compounds the same way, and for a creator on a metered or shared connection it is the half of the codec saving that gets felt weekly. Sending that month of lessons at a 20 Mbps uplink takes about 83 minutes in H.264, about 55 minutes in AV1 and about 43 minutes in H.265.

The archive saving is storage; this one is evenings. When the group's creator told me his uploads ran during dinner because they took so long, the codec change was the whole prescription, and it cost him nothing but an overnight re-export and one compatibility decision about delivery.

The Split I Use Now

My working rule, and the one I gave the creator in that group: archive in H.265, deliver in H.264, and keep the camera masters on their own drive under their own rule. The archive copy is the one that grows forever, so it gets the efficient codec. The delivery copy is the one that meets old tablets and borrowed phones, so it gets the codec that plays everywhere. AV1 earns a place when a platform asks for it or when I control the playback device; it is not my default for files that travel to strangers.

For a creator choosing an archive and delivery codec for a whole library, that split is the use case in one paragraph, and the video file size calculator prices any version of it before a single export starts: pick the preset, set the codec, read the size per hour, multiply by the library. Anyone planning physical drives around the result should then read The Wedding Season Storage Plan, where the same multiplication decides how many terabytes a season of shooting really needs. Codecs decide the rate. Duration does the rest.