HandBrake: Compress Large Videos Without Losing Quality
The setting that decides whether your compressed video still looks good is Constant Quality's RF value — and almost every tutorial changes the wrong thing first. Set RF to 22–24 for 1080p, leave the resolution at the source size, keep the H.264 encoder unless you specifically need smaller files, use software encoding unless speed is the constraint, and test a one-minute sample before you commit. That combination routinely takes a multi-gigabyte recording down to a few hundred megabytes with no visible difference. The order matters because lowering resolution is an irreversible cut, while raising RF is just a coarser setting you can undo by re-encoding. Do it in the wrong order and no setting afterwards can recover the detail you already threw away.
HandBrake is the free video transcoder most people end up with, and for good reason: it is open source, runs on Windows, macOS and Linux, converts nearly any format into modern codecs, and asks nothing in return. Its features page lists the whole toolbox — AV1, H.265 and H.264 encoders, hardware acceleration, three container formats, subtitle and chapter handling, filters from deinterlacing to denoise.
That toolbox is also the problem. HandBrake hands you real controls, and a real control set can be used badly. The guides that walk you through every checkbox are thorough and nearly useless, because four settings determine almost all of the outcome, and the most-recommended order of operations is backwards.
I have re-encoded enough screen recordings, camera footage and DVD rips to have made most of the mistakes below personally. This is the short version, in the order I would actually fix them.
Why does compressing a video usually mean blurring it?
Because two very different kinds of loss get confused with each other.
Resolution loss is a subtraction. If you take a 4K frame and scale it to 1080p, three out of every four pixels are gone before the encoder even starts. No setting afterwards brings them back. Text at 12 pixels tall becomes a grey smear, and a screenshot of a spreadsheet becomes unreadable. This is the loss people mean when they say a video "got blurry".
Compression loss is a budget. An encoder does not store every pixel. It describes each frame as "this block of the previous frame, shifted, plus this small correction", and the amount of correction it is allowed to spend is what you control. Give it a generous budget and it reproduces the frame almost exactly. Give it a tight budget and it starts approximating — flat areas survive, edges soften, and dark gradients turn into blocks. That is visible as blockiness and smearing, not as blur.
The practical consequence: compression loss is a dial you can turn back and forth and re-test; resolution loss is a one-way door. Most "compress your video" tutorials open with the one-way door, which is exactly why their readers end up with files that are small, soft and unrecoverable.
Which single setting decides whether the file still looks good?
Constant Quality. Not the encoder, not the preset, not the filters — the rate control mode, and then the number attached to it.
HandBrake gives you two rate control modes on the Video tab, and its own feature list names them plainly: Constant Quality or Average Bitrate Video Encoding. They answer different questions.
Average Bitrate targets a size. You say "spend 5,000 kbps" and the encoder spends 5,000 kbps on a static interview and 5,000 kbps on a fast pan through foliage, because it has been told the budget is fixed. It wastes bits where nothing is happening and starves the frames that actually need them. The result is worse at the same size, in nearly every case.
Constant Quality targets an appearance. You say "hold this level of fidelity" and the encoder decides second by second how many bits that takes — a few on a static title card, a lot on a scene change. The output is better looking at the same size, and the size is whatever falls out.
The number attached to Constant Quality is RF, short for rate factor, and it runs backwards from what people expect:
- Lower RF = higher quality, larger file. RF 0 is mathematically lossless.
- Higher RF = more compression, smaller file, more visible artefacts.
- The usable range for ordinary viewing sits roughly between 18 and 28, and the encoder's own typical default — usually 22 or 23 — is in the middle of it.
One thing worth understanding about the scale: RF is not linear in file size. Raising it by about six points roughly halves the output, which is why a change from 20 to 26 feels dramatic while 22 to 24 feels modest. That also means the tempting move of "just set it to 30 to be safe on size" costs you far more quality than the number suggests.
What RF value should you actually type in?
Start from the preset's default, then adjust in steps of two. The working ranges below are the ones HandBrake's own quality guidance and the wider x264/x265 community have converged on, and they hold up in practice:
| Output resolution | Start at | Usable range | If you need it smaller |
|---|---|---|---|
| 480p / 576p (SD) | RF 20 | 18–22 | 22 |
| 720p (HD) | RF 22 | 19–23 | 23–24 |
| 1080p (Full HD) | RF 22 | 20–24 | 24–26 |
| 2160p (4K) | RF 24 | 22–28 | 28 |
Higher resolutions tolerate higher RF values because artefacts occupy proportionally less of the frame — a 4K file at RF 26 can look better than a 1080p file at RF 24. That is not a reason to encode 4K when you need 1080p; it is a reason not to panic when a 4K source wants a higher number.
Resolution is only half the input. Content matters just as much. What you are encoding decides where in the range you land:
- Cheap to encode: screen recordings, UI walkthroughs, slide decks, flat animation, static interviews with a plain background. Large uniform areas and little motion compress well, so these stay sharp even at the high end — a 1080p screen recording at RF 24–26 is often indistinguishable.
- Expensive to encode: grainy film, confetti, rain, smoke, foliage in wind, fast sports, low-light footage with sensor noise. The encoder cannot summarise noise, so it spends bits on it and shows artefacts early — bias toward the low end.
Do not tune blind. The fastest reliable method takes five minutes:
- Pick one minute from the most difficult part of the video — the most motion, the most fine texture, the worst lighting.
- Encode it three times at two-point steps around your guess — 20, 22 and 24 for 1080p.
- Watch them full screen at the distance you will actually watch, not zoomed to 400%.
- Take the highest RF you cannot distinguish from the source. That is your setting for that kind of content.
HandBrake's preview feature exists for exactly this: it renders a short sample with your current settings before you spend an hour on the whole file. If you only use one feature beyond the defaults, use that.
Is H.265 worth the extra encoding time?
Sometimes. It is the second-biggest lever you have, and the one with the most compatibility cost.
H.265 (HEVC) reaches the same visible quality at a meaningfully lower bitrate than H.264 — commonly cited as 25–50% smaller for equivalent quality, with the exact figure depending on content and encoder settings. AV1 can go further still and is royalty-free, but software AV1 encoding is slow unless you have a recent GPU with a hardware encoder.
What you pay for that:
- Encode time. Several times longer than H.264 at comparable settings.
- Compatibility. H.264 plays on essentially everything made in the last fifteen years. H.265 is fine on modern phones, modern computers and most smart TVs, but older players, some Blu-ray software, some embedded devices and a fair number of web upload pipelines still choke on it.
- A different RF scale. RF numbers do not mean the same thing across encoders. For x265, plan on roughly two to four points higher than your x264 number for similar visible quality — RF 24 in x265 is broadly comparable to RF 22 in x264.
My rule: H.264 unless the file is going somewhere specific. If it is for your own phone, your own laptop, or a modern streaming platform, H.265 earns its extra time. If it is going to a client, a school laptop, a USB stick handed to someone whose setup you cannot check, or an old TV, H.264 is the answer and the extra megabytes are the price of not getting a "it won't play" email.
Should you use hardware encoding or software?
Software, unless time is the binding constraint. This is the one place where the free tool's own documentation is refreshingly honest about the trade-off.
HandBrake can hand the encode to the chip on your graphics card: NVIDIA NVENC, AMD VCN, Intel Quick Sync Video, or Apple VideoToolbox. The speed difference is not subtle — hardware encoders typically finish several times faster than software x264.
The cost shows up at the same output size. Hardware encoders have a much smaller toolbox of tools and coarser control over how bits are spent, so at equal file size the software encoder generally looks better, and the gap widens at low bitrates. On scaling specifically, HandBrake's VideoToolbox documentation states that its scaler uses a different algorithm from the software scaler and the output will be less sharp. That is the project describing its own limitation, not a competitor's claim.
Two more things the hardware pages make clear, both worth knowing before you assume something is broken:
- High CPU usage during hardware encoding is normal. The GPU handles video encoding only; decoding, all filters, audio encoding, subtitle handling and muxing still run on the CPU, in parallel with the encode. Seeing 100% CPU does not mean hardware encoding failed.
- Hardware decoding switches off when a filter is active. If the decoded frames have to make a round trip to the CPU — which any filter, including crop and scale, requires — HandBrake falls back to software decoding. Disabling filters you do not need is the standard fix for a slow hardware pipeline.
So: software x264 for anything you care about and can leave running; hardware encoding for bulk jobs, drafts, or a laptop that would otherwise be pinned for hours. If you use it, start from the official hardware presets — HandBrake ships H.265 NVENC 2160p 4K and H.265 NVENC 1080p, plus the equivalent VCN and VideoToolbox entries, and its docs explicitly advise against hand-editing the advanced encoder options because those presets already cover the common cases.
Why is Average Bitrate the wrong mode for almost everything?
Because it optimises for a number instead of an appearance, and in normal use you do not actually have a bitrate limit.
Average Bitrate exists for the cases where something downstream imposes a hard ceiling: a platform that rejects files above a given bitrate, a device with a strict decoder budget, or a target file size you must hit exactly — "this has to fit on a 4.7 GB disc" or "this has to be under 25 MB for an email attachment". In those cases you calculate the bitrate from the size and duration, and Average Bitrate is the correct tool.
If you do use it, enable two-pass. The first pass maps where the complex scenes are; the second spends the budget accordingly. One-pass Average Bitrate is the worst of both worlds — it neither hits quality nor spends bits sensibly. HandBrake's command line exposes this as --multi-pass, with --turbo available to speed up the first pass for x264 and x265.
If you find yourself in Average Bitrate because a guide told you it gives "more control", you have been given the wrong tool. Constant Quality with an RF value gives you more control over the thing you can actually see.
When should you lower the resolution instead?
When the destination cannot use the extra pixels. Not before.
There are real cases for it: a platform that caps uploads at 1080p, a video that will only ever be watched on a phone, a presentation recorded at 4K that will be embedded at 800 pixels wide, a file that must fit a size limit after every other lever has been pulled. In those cases downscaling is efficient, because you are removing data nobody will look at.
What makes it a bad first move is the arithmetic. A 720p frame has about 44% of the pixels of a 1080p frame. That is a genuine saving — and a genuine loss, permanently, for anything with fine detail. Text, UI elements, subtitles and faces all degrade earlier than people expect, and the degradation looks like blur rather than like blockiness, so it reads as "HandBrake ruined my video" rather than "I removed 56% of the image".
Two rules that keep this sane:
- Never upscale. Setting the output above the source resolution adds no detail, only bytes. If your source is 720p, keep 720p.
- Change one variable at a time. If you lower resolution, note the original RF first, so you know what changed when you compare the results.
The one exception worth making: screen recordings of documents, code and dashboards. These often survive a downscale better than camera footage does, because their detail is high-contrast edges rather than texture — but only if you keep the scaler good. If you downscale in HandBrake, check what the Dimensions tab is using and prefer the software scaler's sharper options over a hardware scaler that the docs already flag as softer.
What do the official presets already get right?
More than most guides admit — they are a starting point, not an answer, and knowing which kind you are picking matters.
HandBrake's preset list has three categories, and they exist for different jobs:
- General — the everyday profiles. Fast 1080p30 is the sensible default for most video: a 1080p ceiling, H.264, quality-based rate control, and a fast-ish x264 preset. It is a good base to override RF on.
- Web — profiles aimed at streaming and sharing, tuned for compatibility and progressive download.
- Devices — Android 1080p30/720p30/576p25/480p30; Apple 2160p60 4K HEVC Surround, 1080p60, 1080p30, 720p30, 540p30, 240p30; Chromecast and Fire TV 2160p60 4K HEVC and 1080p30; PlayStation 1080p30 and 720p30. These are compatibility profiles: they encode to what that device will actually play, which is a different goal from "smallest file that still looks good".
The mistake is treating a device preset as a compression preset. Picking Apple 2160p60 4K HEVC Surround for a 4K file you want to shrink does not shrink it much — it targets playback on Apple hardware, at a quality level Apple considers appropriate. If size is the goal, start from General at your source resolution and set RF yourself.
The other thing presets do correctly, and that people override too quickly, is constrain the profile and level. HandBrake's documentation is clear that its encoders set the appropriate level automatically, and that you should leave it on Auto unless you need a specific level for device compatibility. Setting a level by hand to "improve compatibility" usually just constrains the encode for no benefit.
All four settings, in one table
If you only remember one screen, remember this one. The columns are the four things that move the result, and the right-hand column is what to do when the outcome is wrong.
| Setting | Where | What it controls | What to use | If the result is wrong |
|---|---|---|---|---|
| Quality mode + RF | Video tab | How much detail the encoder may discard — the biggest lever on both size and appearance | Constant Quality, RF 22 for 1080p, steps of 2 | Blocky or smeared → lower RF by 2. Still too big but clean → raise by 2 |
| Encoder | Video tab | Compression efficiency and compatibility | H.264 by default; H.265 when the destination is modern and you need smaller | Won't play on the target device → H.264. Too big and playback is guaranteed → H.265, RF up 2–4 |
| Software vs hardware | Encoder dropdown | Encode speed versus quality per byte | Software x264/x265 unless time is the constraint | Sofware encode taking hours → switch to hardware, accept a slight quality cost |
| Resolution | Dimensions tab | How many pixels exist at all — irreversible | Match the source unless the destination needs less | Only cut after RF and encoder are settled, and never above the source |
The fourth column is the part that never appears in checkbox-by-checkbox guides, and it is the whole reason to read one of these: the fix depends on the symptom, and only two of the four settings are reversible. Blockiness is a compression symptom, so you fix it with RF. A file that will not play is a compatibility symptom, so you fix it with the encoder. A file that is soft in a way that looks uniform everywhere is a resolution symptom, and you fix it by going back to the source and starting over — which is why you should not touch resolution until everything else is right.
Our position: HandBrake is the right tool, and it is free in a way that is worth defending — GPLv2 for most of the source, BSD 3-Clause for some portions, volunteer-built, no account, no watermark, no paid tier. What it is not is self-explanatory, and the ecosystem around it has settled on advice that is backwards: shrink the frame first, then wonder why the text is soft. Set RF, test a minute, keep the resolution, and re-encode if you guessed wrong. That takes five minutes and it is the difference between a 4 GB file you cannot send and a 400 MB file nobody can tell apart from the original.
One safety note that belongs here rather than in a footnote: because HandBrake is free and popular, it attracts fake download sites, "cracked" builds and paid licence keys for a product that has no paid tier. The official download page is explicit — the only official sources are this website (handbrake.fr), the HandBrake GitHub project, and Flathub — and it publishes checksums and OpenPGP signatures so you can verify what you got. The current release is HandBrake 1.11.2, for macOS 10.13 and later, Windows 10 and later (separate x64 and ARM64 builds), and Linux via Flatpak. Anything asking you to pay or to create an account to download it is not the project.
How We Put This Together
Version, platform support, container and codec support, audio passthrough, quality modes, presets and licensing were checked directly on HandBrake's official site in November 2026: the current release is 1.11.2; containers are MP4, MKV and WebM; video encoders are AV1, H.265, H.264, MPEG-4, MPEG-2, VP8 and VP9; quality is either Constant Quality or Average Bitrate; most of the source is GNU GPL version 2 with some portions under BSD 3-Clause. Menu and option names come from HandBrake's own documentation and command-line reference, including the -q/--quality and -b/--vb options, the --multi-pass and --turbo flags, and the --vfr/--cfr/--pfr frame-rate controls. Hardware encoder behaviour — including that VideoToolbox scaling produces softer output, that CPU usage stays high during hardware encodes, and that hardware decoding is disabled when filters are active — is quoted from the project's own NVENC, VCN and VideoToolbox pages, as is the advice not to hand-tune advanced encoder options. This article deliberately does not publish measured encode times, measured file sizes or side-by-side quality scores from a specific machine: those vary by source, hardware and settings, and inventing them would be worse than leaving them out. The RF ranges given are the practical values that HandBrake's quality guidance, the x264/x265 encoders' own defaults and the wider encoding community have converged on, and the instruction to test a one-minute sample is the calibration method rather than a claim that any single number is correct for every file.
Frequently asked questions
Is HandBrake really free, and is it safe to install?
Yes. It is open source — most of the code under GNU GPL version 2, some portions under BSD 3-Clause — built by volunteers, with no account and no paid tier. The official download page names exactly three legitimate sources: handbrake.fr, the HandBrake GitHub project, and Flathub for Linux, and publishes checksums and OpenPGP signatures for verification. Any site selling a licence key or an activation code is selling something that does not exist.
What RF value should I use for a 1080p video?
Start at RF 22 and adjust in steps of two. RF 20–24 is the practical range for 1080p, with 22 or 23 the usual preset default. Lower means higher quality and bigger files; RF 0 is lossless and enormous. Blocking in dark scenes means lower the number; a clean file that is still too large means raise it. Screen recordings tolerate the high end of that range far better than grainy footage does.
Will compressing a video with HandBrake make it visibly worse?
Not if you stay in the sensible range for your content. Every re-encode loses something, because the encoder discards detail it judges hard to see; the question is whether you can see it at normal viewing distance on the screen you will use. A 1080p screen recording at RF 23 is usually indistinguishable. The damage becomes obvious when people lower the resolution first — that cut is irreversible, whereas RF is a setting you can change and re-test.
Why did my HandBrake output come out bigger than the input file?
Usually one of three things: the source was already efficiently encoded and you set a higher quality target than it was made at, you are in Average Bitrate mode with a value above the source's real bitrate, or you kept extra audio tracks, passed through a surround track, or enabled a filter that outweighed the video savings. Check the source bitrate first — if it is already low, there is little left to remove.
Is hardware encoding faster but worse quality?
Generally yes on both counts. NVENC, AMD VCN, Intel Quick Sync and Apple VideoToolbox encode several times faster than software x264 or x265, and at the same output size the software encoders typically look better, especially at low bitrates. HandBrake's own VideoToolbox page notes its scaler produces less sharp output than the software scaler. Use hardware when time matters more than the last few percent of quality, such as batch-converting a folder overnight.
Next: if you are assembling a free toolkit, see our roundup of free tools that replace paid apps, or the ShareX vs Snipping Tool comparison for capture. And if the video you are trying to shrink is a screen recording that came out enormous to begin with, our guide to the seven OBS recording settings beginners get wrong explains why it grew that large — CBR rate control is usually the culprit, and fixing it at capture time saves you the encode entirely.
Get more guides like this
Free tools, VPNs and realistic side income — tested hands-on, delivered weekly.
Subscribe free