OBS-Studio.net Independent OBS Studio reference
Sign in Create account
Fix OBS Tutorials Settings Tools Overlays Plugins Games Hardware Platforms Latest version FAQ Learning paths Search Sign in Create account

Recording and streaming output settings that do not fight each other

The two outputs have different constraints

A live stream has to fit a fixed upload budget and be decodable in real time by every viewer, so it uses CBR at a bitrate the platform accepts. A local recording has no bandwidth limit and only has to be decodable by your editor, so it should use a quality-targeted mode and a higher bitrate. Using stream settings for recordings throws away quality you already paid for in GPU time.

Recording settings that hold up in an editor

  • Switch Settings → Output to Advanced and select the Recording tab.
  • Use CQP (NVENC/AMF) or CRF (x264) rather than CBR. Lower numbers mean higher quality and larger files.
  • Record to a container that survives a crash. MKV can be recovered if OBS stops unexpectedly; MP4 written directly can be left unplayable. OBS can remux MKV to MP4 afterwards from the File menu.
  • Record to a different physical drive from the one the game streams assets off, if you have one.

Reuse the stream encoder only when you must

Setting the recording encoder to "use stream encoder" means one encode is shared by both outputs. It is the cheapest option and the right one if your hardware is already at its limit, but the recording inherits the stream's CBR bitrate. If the encoder has headroom, run a second independent encode for the recording.

Canvas, output and scaling

Set the base (canvas) resolution to your monitor resolution so sources are not scaled twice, then set the output (scaled) resolution to what you actually send. Downscaling from 1440p to 1080p with the Lanczos filter looks better than dragging sources to fit a smaller canvas, and it keeps one layout usable for both outputs.

Frame rate

Pick 60 fps for fast action and 30 fps for talking-head or slide content. Fractional frame rates such as 59.94 exist for broadcast interoperability; unless you have a specific reason, an integer frame rate avoids a class of sync problems in editors.

Test the whole chain

Record and stream simultaneously for five minutes with your busiest scene, then check the Stats dock. Zero skipped and zero dropped frames with both outputs running is the only proof that the two are not competing for the same encoder.