STREAMING GUIDE / STREAMING HARDWARE

Use NVIDIA Broadcast microphone effects in OBS without double processing

Choose the physical microphone in Broadcast and its processed virtual microphone in OBS. Compare recordings before stacking more filters.

START HERE

Set your real microphone as NVIDIA Broadcast’s input, then select Microphone (NVIDIA Broadcast) as the one microphone source in OBS. Start with one effect and compare an untreated recording with a processed one. Capturing both devices, or applying suppression again in several apps, makes troubleshooting much harder.

Confirm support and record an untreated sample

This guide covers NVIDIA Broadcast on a supported Windows/RTX setup. Check NVIDIA’s current requirements before installing: the effects and driver requirements change, and instructions for the older RTX Voice app are not interchangeable with every Broadcast release.

Before changing OBS, record your physical microphone directly for a baseline. Include normal speech, a quiet ending to a sentence, keyboard sounds, and a louder phrase. Keep the microphone position fixed for the comparison. If the original already crackles, cuts out, or clips, first fix the device level or connection; an effect cannot reliably reconstruct audio that never arrived intact.

Connect the physical microphone to Broadcast, then OBS

  1. Open NVIDIA Broadcast and select the actual USB/interface/headset microphone in its microphone input selector. Avoid an ambiguous Default selection for this test.
  2. Enable the microphone effect you want to evaluate, such as Noise Removal. Leave unrelated camera and speaker effects off initially.
  3. In OBS, add one Audio Input Capture source selecting Microphone (NVIDIA Broadcast), or choose it as the existing global Mic/Aux device. Use one method.
  4. Disable any OBS capture of that same physical mic that would run alongside the processed version. Speak and identify which single meter should respond.

The direction matters. Choosing the processed virtual device as Broadcast’s own input can create a circular route. Keep a written pair: “Broadcast input = physical device; OBS input = Broadcast.” If you later replace the microphone, update the first choice before changing OBS.

Evaluate one effect at a time

Record the same short script with the effect disabled and enabled. Compare both at a similar listening level, including pauses between sentences. Listen for lost consonants, clipped beginnings, watery artifacts, or breaths that become distracting. A quieter room tone is not an improvement if the audience struggles to understand words.

Begin with OBS’s additional noise suppression and gate disabled for this comparison. If you need a gate later, add it only after the processed signal is stable. Discord or a meeting app can also apply its own processing when it consumes the Broadcast microphone; evaluate that route separately instead of assuming a good OBS recording guarantees the same call sound.

Improve microphone placement before pushing processing strength. Moving the microphone closer at a sensible input gain can improve the voice-to-background balance without asking an effect to remove as much competing sound.

Find where a silent or damaged signal begins

TestWhat to check next
Physical mic is silent in a direct recordingHardware mute, Windows input permission, device selection, or the microphone connection.
Physical mic works; Broadcast does notBroadcast’s selected input, enabled effect, supported driver, and its own preview/test controls.
Broadcast works; OBS is silentOBS’s named input, source mute, and recording-track assignment.
Voice becomes damaged only with processingDisable the effect, compare again, then reduce or remove processing before changing the whole route.

If you hear your voice twice, compare the raw and processed source selections before adjusting sync offsets. Two different processing paths can sound like an echo even when each source is healthy by itself.

Test with the game running

Effects use system resources, and NVIDIA recommends enabling only the effects you need. A test at the desktop does not represent a demanding game plus OBS, webcam processing, and a live call. Repeat your speech recording under the workload you actually intend to stream.

If the signal becomes unstable only during that workload, first remove optional effects and compare again. Keep the raw-microphone OBS source available but disabled as a fallback; switch routes deliberately instead of leaving both enabled. Avoid a driver or app update immediately before an important broadcast when you have no time to repeat the check.

Verify mute and restart behavior

Check a hardware mute, OBS output mute, and your call-app mute separately. Confirm the saved file goes silent when the relevant control is used. Restart Broadcast and OBS, then confirm the named virtual microphone returns. The local microphone test can compare available inputs, but its browser processing may differ from OBS, so your final acceptance check is an OBS recording.

Sources & version notes

Documentation checked November 18, 2026. Use the platform and version notes in this guide when comparing settings. The troubleshooting checks are intended for your own setup; this review date does not represent hardware testing.