Best OBS Recording Settings: Quality, File Size, and the CBR Mistake
Best OBS Recording Settings
Most OBS recording problems come from a single root cause: people configure recording the way they configure streaming.
That is understandable — it is the same application, the settings look similar, and streaming guides vastly outnumber recording guides. But streaming and recording have opposite constraints, and the setting that is mandatory for one is actively harmful for the other.
Get that one thing right and the rest follows.
The Core Difference: Bandwidth vs Disk
Streaming sends video down a network connection with a fixed ceiling. Exceed it and viewers buffer. So streaming requires CBR — constant bitrate — because the platform needs a predictable, unvarying data rate. The encoder must spend the same bits every second regardless of what is on screen.
Recording writes to a local disk with no such ceiling. There is no reason to spend a fixed number of bits per second, and every reason not to.
Consider what happens with CBR at 12 Mbps while you record a coding tutorial. For ten minutes the screen shows a mostly-static editor with a blinking cursor. CBR spends 12 Mbps on it anyway — 90 MB per minute to encode a nearly-unchanging image. Then a fast scroll happens, and CBR still spends 12 Mbps, which now is not enough, so that moment looks blocky.
CBR gives you the worst of both: enormous files and artifacts exactly where quality matters.
The fix is quality-based rate control — CQP on NVENC and AMD, CRF on x264. You specify a quality level and the encoder spends whatever bitrate is needed to hit it. Static screen, almost no data. Fast motion, plenty of data. File sizes commonly drop by more than half while quality improves.
This is the single highest-impact change available in OBS recording settings, and it is the one most guides bury.
Recommended Settings
Set Settings > Output > Output Mode to Advanced, then open the Recording tab.
Encoder
| Hardware | Choose |
|---|---|
| NVIDIA GTX 1050+ | NVENC H.264 (or HEVC on RTX) |
| AMD RX 400+ | AMF H.264 / HEVC |
| Intel with Quick Sync | QuickSync H.264 / HEVC |
| Apple Silicon / recent Intel Mac | Apple VT H264 Hardware Encoder |
| No hardware encoder | x264 |
Hardware encoders run on dedicated silicon, leaving your CPU free for the game or application you are recording. On any modern system this is the right default.
x264 is still slightly more efficient per bit — it produces marginally smaller files at the same quality — but it consumes CPU that your recorded application probably needs more. Use it when you have no hardware encoder, or when you have a high core count and are recording something CPU-light.
Rate control and quality
- Rate Control: CQP (NVENC/AMF/QuickSync) or CRF (x264)
- CQ Level / CRF: 20
The scale runs roughly:
| Value | Result |
|---|---|
| 15–18 | Visually lossless, large files |
| 19–23 | Excellent quality, sensible size — start here |
| 24–26 | Noticeable softening on detail |
| 27+ | Visible blocking and artifacts |
Start at 20. If files are larger than you want, move to 22 before touching anything else.
Keyframe interval
0 (auto) for recording, or 2 if you plan to edit the footage. Shorter keyframe intervals make seeking and cutting more responsive in an editor, at a modest cost in file size. This matters more than it sounds — scrubbing footage with a long keyframe interval is genuinely unpleasant.
Preset
NVENC exposes P1–P7; use P5 (Quality) or P6 (Slower). x264 uses named presets; veryfast is the standard recommendation, faster or fast if you have CPU headroom.
Slower presets do more analysis and produce smaller files at the same quality. They also cost more resources, so the correct choice depends on what else your machine is doing.
Profile
High. There is no reason to use Main or Baseline for local recording — every modern decoder handles High Profile.
Audio
192 kbps AAC for most content. 160 kbps is fine for voice-only. Go to 256 or 320 only if music fidelity genuinely matters, since audio is a small fraction of total size either way.
MKV vs MP4: Record to MKV
This is the second setting people get wrong, and the consequences are worse than a large file.
MP4 writes its index at the end of the file. If OBS crashes, your machine loses power, or you force-quit mid-recording, that index is never written and the file is typically unrecoverable. Hours of footage, gone.
MKV is written incrementally. Interrupt it and you get a valid file containing everything up to the failure.
Record to MKV, then use File > Remux Recordings in OBS to produce an MP4 when you need one. Remuxing is a container change, not a re-encode — it copies the existing streams into a new wrapper. It is lossless and takes a few seconds regardless of recording length.
Recent OBS versions also offer hybrid MP4, which writes recovery information as it goes and is far more crash-resistant than standard MP4. If your editor refuses MKV, that is a reasonable middle ground. If it accepts MKV, plain MKV remains the safest option.
Our MKV vs MP4 comparison covers the format differences, and converting MKV to MP4 covers the conversion in more depth.
Video Settings That Affect Size
Under Settings > Video:
Base (Canvas) Resolution — your monitor's native resolution.
Output (Scaled) Resolution — what actually gets recorded. Downscaling from 1440p to 1080p is one of the largest file-size reductions available, and for most content the difference is invisible on playback.
Downscale Filter — Lanczos if you have GPU headroom, Bicubic otherwise. Bilinear is noticeably softer; avoid it.
FPS — 60 for gameplay, 30 for screen recording and tutorials. This is worth dwelling on: recording a coding tutorial or software walkthrough at 60 fps roughly doubles file size for content where nothing moves fast enough to benefit. Most screen recordings should be 30 fps.
Codec: H.264 or HEVC?
HEVC produces roughly 30–50% smaller files at equivalent quality. If your GPU supports hardware HEVC encoding — NVIDIA GTX 1650 and newer, recent AMD cards, Apple Silicon — the performance cost over H.264 is minimal.
The catch is compatibility downstream. Some editors handle HEVC poorly, some upload pipelines choke on it, and older machines struggle to decode it smoothly during playback.
Recommendation: H.264 if the footage goes straight into editing. HEVC if you are archiving raw recordings or your pipeline is known to handle it. See HEVC vs H.264 for the full comparison.
A Complete Starting Configuration
Output > Recording:
- Type: Standard
- Recording Format: MKV
- Encoder: hardware encoder for your GPU
- Rate Control: CQP / CRF
- CQ Level: 20
- Keyframe Interval: 2
- Preset: P5 Quality (NVENC) or veryfast (x264)
- Profile: high
Output > Audio:
- Track 1 Bitrate: 192
Video:
- Output Resolution: 1920×1080
- Downscale Filter: Lanczos
- FPS: 30 for screen recording, 60 for gameplay
Record two minutes, check quality and size, adjust CQ from there. That loop takes five minutes and beats copying anyone's numbers verbatim.
The Part After Recording
Even well-configured recordings are large, and that is correct — a recording master should be high quality, because you cannot recover detail you never captured.
The mistake is treating that master as your delivery file. A CQ 20 recording is excellent for editing and considerably larger than anything you should upload or share. Which means the workflow has two distinct compression stages:
- Recording — quality-based rate control, generous settings, MKV. Optimised for fidelity and crash safety.
- Delivery — compressed to a target, in a format the destination accepts. Optimised for size.
Trying to make one setting serve both is why people end up with recordings that are simultaneously too big to upload and too compressed to edit well.
For the second stage, Compresto handles it on macOS — drop in an OBS recording, set a target size or quality, and get back something you can actually upload. Hardware-accelerated H.264 and HEVC, batch processing for a folder of recordings, MKV and MP4 in and out, and nothing leaves your machine.