TL;DR — OBS Studio is free, open-source, scriptable via obs-websocket, and runs on Windows/macOS/Linux. It beats QuickTime and Game Bar because it treats recording like a pipeline: sources → scenes → encoder → file. This post is a practical config walkthrough with the trade-offs (MKV vs MP4, NVENC vs x264, 30 vs 60 fps), the audio filter chain you should always apply, and the mistakes that bite you in post. Skip to the config tables if you just want the numbers.
OBS Studio recording interface with scenes, sources, and audio mixer panels on a creator's desk setup
Why OBS is the right tool for devs
If you've ever tried to record a terminal session + webcam + system audio with QuickTime, you know the pain: one track, no mixer, no overlays, no way to swap layouts mid-recording. OBS solves this by exposing a compositor model that should feel familiar if you've ever touched a shader pipeline or a video editor's node graph.
The pieces:
Sources — input nodes (display capture, window capture, webcam, mic, browser, image, media file).
Scenes — named compositions of sources. Think of them as saved layouts.
Audio Mixer — per-source volume + filters, routable to up to 6 separate tracks in the output file.
Encoder — NVENC / AMD HW / Apple VT / x264, configurable bitrate and container.
Output — MKV or MP4 on disk, RTMP/SRT to Twitch/YouTube/Facebook.
OBS is the de facto standard for recording and streaming as of 2026, used by tens of millions of creators, educators, businesses, and gamers. It's free, open source, no watermark, no time limit, no subscription.
Install (30 seconds)
macOS
brew install --cask obs
Windows (winget)
winget install OBSProject.OBSStudio
Linux (Ubuntu/Debian)
sudo add-apt-repository ppa:obsproject/obs-studio
sudo apt update && sudo apt install obs-studio
Or just grab the installer from:
https://obsproject.com
First launch → Auto-Configuration Wizard → pick "Optimize for recording, I will not be streaming" if you're capturing local video. It benchmarks your CPU/GPU and picks an encoder + bitrate. Trust it for the first pass; tune later.
System baseline for 2026:
Platform Minimum Recommended
Windows Win10 x64, 4 GB RAM, DX11 GPU Win11, 16 GB RAM, dedicated GPU
macOS macOS 11 Big Sur macOS 13+, Apple Silicon
Linux Ubuntu 20.04+ Latest LTS, 8 GB RAM
The minimum-viable scene graph
Most devs overthink this. Here's the setup I use for screencasts:
Scene: "Demo"
├── Display Capture (monitor 1)
├── Video Capture Device (webcam, bottom-right, 320×240)
├── Image (logo.png, top-right, 15% opacity)
├── Audio Input Capture → Track 1 (mic)
└── Audio Output Capture → Track 2 (desktop)
Scene: "Intro"
├── Media Source (intro.mp4, loops: false)
└── Audio Output Capture → Track 2
Scene: "BRB"
├── Image (brb-card.png)
└── Media Source (lofi-loop.mp3)
Key property: sources are shared by reference across scenes. Move the webcam once, it updates everywhere. This is why "one giant scene with toggled visibility" is an anti-pattern.
OBS Studio Scenes panel showing multiple named scenes like Intro, Screen + Webcam, and Starting Soon
Source types, picked correctly
Source Captures Use when
Display Capture entire monitor presentations, multi-window demos
Window Capture single app window IDE-only tutorials, hiding Slack
Game Capture full-screen game gaming; lower overhead than Display
Video Capture Device webcam / HDMI capture card face cam, DSLR
Audio Input Capture mic narration
Audio Output Capture system audio app sounds, tutorial music
Browser Source URL rendered live chat overlays, alerts, HTML widgets
Text (GDI+) rendered text titles, live stats
Order in the Sources list = z-index. Top of list = frontmost. Drag to reorder.
Audio: the part 90% of people get wrong
Audio quality is what separates "YouTube tutorial" from "looks like someone's cousin recorded it." Route each source to its own track so you can fix it in post without re-recording.
Settings → Audio → Advanced Audio Properties
Mic/Aux → Track 1 ✓
Desktop → Track 2 ✓
(both mixed → Track 1 ✓ for a convenience master)
Settings → Output → Recording → Audio Track: 1,2,3
Now your output file has separate stems. Your future self editing in Resolve/Audition will thank you.
The mic filter chain (apply in this order)
Right-click mic in the mixer → Filters → Add:
- Noise Suppression (RNNoise method; always on)
- Noise Gate (threshold: ~-40 dB, for noisy rooms)
- Compressor (ratio 4:1, threshold -18 dB, attack 6ms, release 60ms)
- Limiter (ceiling -1.5 dB; prevents clipping)
- Gain (if the mic is quiet, +6 to +12 dB here) Target levels on the mixer meter:
Mic peaks: -12 dB to -6 dB
Desktop: -6 dB to -3 dB (duck below mic)
Output config: MKV + hardware encoder
Settings → Output → Output Mode: Advanced
├── Recording Path: /Users/you/Recordings
├── Recording Format: mkv # crash-safe; remux to mp4 later
├── Encoder: NVIDIA NVENC H.264 # or AMD HW, Apple VT, x264
├── Rate Control: CBR
└── Bitrate: 15000 Kbps # 1080p30 sweet spot
Why MKV? If OBS crashes or your power dies mid-recording, MKV recovers gracefully. MP4 finalizes its index only on clean close — crash and the whole file can be unreadable. Remux later, zero quality loss:
File → Remux Recordings → select .mkv → outputs .mp4
No re-encode, instant, same bitrate.
Bitrate cheat sheet
Use case Resolution FPS Bitrate Encoder
Tutorial / software demo 1920×1080 30 12,000–20,000 Kbps NVENC or x264
Gaming 1920×1080 60 20,000–40,000 Kbps NVENC
Talking head / podcast 1920×1080 30 8,000–15,000 Kbps any
Presentation 1920×1080 30 8,000–12,000 Kbps any
4K 3840×2160 30 40,000–80,000 Kbps NVENC / VT
Storage math: at 20,000 Kbps you burn ~150 MB/minute, so a 60-minute session is ~9 GB. Check disk before hitting record. Running out of space mid-session is the single most preventable failure mode.
Video settings
Settings → Video
Base (Canvas) Resolution: 1920×1080 # match your display
Output (Scaled) Resolution: 1920×1080
Downscale Filter: Lanczos # best quality
Common FPS: 30 # 60 only if you actually need it
60 fps roughly doubles file size at the same bitrate. Use it for gameplay or motion-heavy content, not for somebody typing in a terminal.
OBS Studio Output Settings panel showing encoder, bitrate, recording path, and MKV format configuration
Hotkeys > clicking
Settings → Hotkeys. Bind at minimum:
Start Recording → F9
Stop Recording → F10
Mute/Unmute Mic → F8
Push-to-talk Mic → (hold) Right Alt
Switch to Scene "Demo" → Ctrl+1
Switch to Scene "BRB" → Ctrl+2
You never want the audience watching your mouse hunt for a button in the OBS window during a live demo.
Virtual Camera: OBS as your Zoom webcam
Tools → Start Virtual Camera. OBS now shows up as a camera device in Zoom, Meet, Teams, Discord. Full composited scene — webcam + slides + overlays + filtered audio — piped into any conferencing app. Zero extra hardware.
obs-websocket: scripting OBS
For anything repeatable, use the built-in obs-websocket (Tools → WebSocket Server Settings). Example in Node:
import OBSWebSocket from 'obs-websocket-js';
const obs = new OBSWebSocket();
await obs.connect('ws://127.0.0.1:4455', 'your-password');
// Switch scene
await obs.call('SetCurrentProgramScene', { sceneName: 'Demo' });
// Start recording
await obs.call('StartRecord');
// ... do stuff ...
await obs.call('StopRecord');
await obs.disconnect();
This is how you wire OBS into CI-style recording workflows, Stream Deck macros, or "record this test run automatically" tooling.
Plugins worth installing in 2026
Plugin What it does
StreamFX 3D transforms, masks, shader filters
Advanced Scene Switcher trigger-based scene changes (window focus, idle, schedule)
OWN3D Pro animated overlay library
DroidCam Virtual Output phone as wireless webcam
Downstream Keyer persistent overlay above all scenes
obs-websocket remote API (bundled in modern OBS)
Install path:
Windows: C:\Program Files\obs-studio\obs-plugins\
macOS: ~/Library/Application Support/obs-studio/plugins/
Linux: ~/.config/obs-studio/plugins/
Restart OBS after dropping files in.
Common mistakes (a.k.a. things I've done)
Single audio track for mic + desktop — can't fix noise without ducking your game. Route to separate tracks.
60 fps on a static terminal recording — doubles file size, zero perceptible benefit.
No filter chain on the mic — sounds amateur even with a Shure SM7B.
Forgot to change the recording path — output lands in ~/Videos/ and you don't find it for a week.
One monster scene with everything visible-toggled — switching scenes is faster and cleaner.
A/V drift on webcam — open the Video Capture Device properties and nudge the audio offset in ms. Calibrate with a clap test.
MP4 instead of MKV — one crash and the whole file is toast.
Segment, don't marathon
Record in chunks (per topic/chapter) and stitch in post. Trade-off: slightly more files to manage. Payoff: a crash loses one segment not the whole thing, edits are faster, and shorter clips retain viewers better on YouTube. Teams that switch to segment-based recording for webinars and training typically cut post-production time by 40–60%.
Shipping to a global audience
Once you've got a clean MP4, the highest-leverage next step is localization — no re-recording required. Tools like VideoDubber run the full pipeline (transcription → translation → voice cloning → lip-sync) across 150+ languages.
Workflow:
- File → Remux Recordings (mkv → mp4)
- Upload to https://videodubber.ai (or paste a YouTube URL)
- Pick target languages, review auto-transcript
- Download dubs, ship to language-specific channels Per Gartner benchmarks, localizing support/product videos into the top 3–5 customer languages deflects 30–50% of support tickets from non-English markets (self-service ≈ $1.84/issue vs $13.50 for agent contacts). Full methodology in the multilingual customer support video guide, and if you're eyeing Bilibili's 340M Chinese users, see the Bilibili repurposing guide.
Key takeaways
OBS is a free, open-source compositor for recording + streaming on Win/macOS/Linux.
Scenes and sources are a graph, not a timeline — design accordingly.
Default to MKV + hardware encoder (NVENC/AMD HW/Apple VT) + multi-track audio.
Mic filter chain: Noise Suppression → Noise Gate → Compressor → Limiter. Non-negotiable.
Bind hotkeys. Use obs-websocket for automation. Record in segments.
Translate with VideoDubber to multiply reach without re-recording.
Download OBS Studio → · Dub your recordings with VideoDubber →
Reference: https://videodubber.ai/blogs/how-to-record-with-obs/.
Top comments (0)