Background noise ruins EVP sessions, investigation live streams, and team communication faster than anything else. A fan humming, wind outside, or your own PC whirring can bury the faint audio anomalies you’re trying to capture. NVIDIA Broadcast noise removal solves this by using AI to strip unwanted sound from your microphone input in real time, no expensive hardware upgrades, no post-production editing, no audio engineering degree required.
At Haunt Gears, we test and recommend gear that makes paranormal investigations more effective. NVIDIA Broadcast has become a go-to software tool for investigators who record audio, run live streams from haunted locations, or coordinate with teammates over voice chat. It pairs well with the EVP recorders and microphones already in your kit, and it’s free for anyone with a supported NVIDIA GPU.
This guide walks you through downloading, installing, and configuring NVIDIA Broadcast for noise and echo removal. You’ll learn which filters to use, how to dial in sensitivity settings for different environments, and how to route the cleaned audio into your recording or streaming software. Whether you’re setting up for a solo investigation or a full team session, you’ll have broadcast-quality audio by the end of it.
What NVIDIA Broadcast noise removal does
NVIDIA Broadcast noise removal is an AI-powered audio filter that runs in real time on your GPU. Instead of using traditional frequency-based noise gates, it uses a deep learning model trained on thousands of hours of audio to identify and separate speech from background sound. The result is a clean audio signal with ambient noise stripped out before it ever reaches your recording software, streaming platform, or voice call.
What makes this different from a simple noise gate is how it adapts to changing conditions. A noise gate cuts audio below a set volume threshold, which means any noise louder than your voice will pass right through. The AI model in NVIDIA Broadcast constantly analyzes your audio and removes noise even when it overlaps with speech. A vacuum cleaner running in the next room, rain hitting a window, or the hum of an air conditioner gets filtered out without cutting off your voice mid-sentence.
For paranormal investigators, this means you can record in noisy environments like basements, attics, or outdoor locations and still capture cleaner audio for EVP analysis.
The two core filters
NVIDIA Broadcast gives you two separate audio filters, and understanding what each one targets helps you pick the right combination for your setup.
Noise Removal focuses on steady and sudden ambient sounds: HVAC systems, fans, keyboard clicks, traffic outside, and other environmental noise. Room Echo Removal targets a different problem entirely. It handles the reverb and echo created when your voice bounces off hard surfaces. If you investigate in large empty buildings with concrete walls, room echo removal is the filter that addresses that specific issue.
| Filter | What it removes | Best for |
|---|---|---|
| Noise Removal | Fans, HVAC, keyboard, environmental noise | Any recording environment |
| Room Echo Removal | Reverb, room reflections, echo | Untreated rooms, large open spaces |
You can run both filters simultaneously. NVIDIA recommends this for most setups, and in practice, combining them produces noticeably cleaner audio than running either filter on its own.
What it does not do
The filters are powerful, but they have limits worth knowing before you start relying on them. Noise Removal does not improve your microphone’s base frequency response or add clarity that wasn’t captured to begin with. It removes unwanted sound; it does not enhance the underlying signal you’re recording.
Room Echo Removal performs best in moderately reverberant spaces. In an extremely large area with severe echo, you may notice the filter producing an unnatural, slightly hollow quality in your voice. Adjusting the filter intensity setting, which you’ll configure in a later step, is the fix for that. Knowing these boundaries before you start saves time when you run into unexpected results during a session.
Step 1. Check your GPU, OS, and drivers
Before you download anything, confirm your system meets the minimum requirements. NVIDIA Broadcast noise removal runs entirely on your GPU, so your graphics card is the most important factor. If your hardware does not qualify, the app will install but the AI filters will not activate, leaving you with no noise processing at all.
GPU requirements
NVIDIA Broadcast requires a GeForce RTX, Quadro RTX, or TITAN RTX GPU with at least 4 GB of VRAM. The AI processing happens on the Tensor Cores built into RTX-series cards, which is why older GTX cards do not qualify even if they have plenty of VRAM. Check your exact GPU model by pressing Windows key + X, selecting Device Manager, then expanding the Display adapters section.
If you are unsure which RTX card you have, the NVIDIA Control Panel lists your full GPU model name under System Information.
Here is a quick reference for supported GPU families:
| GPU Family | Supported |
|---|---|
| GeForce RTX 2000 series | Yes |
| GeForce RTX 3000 series | Yes |
| GeForce RTX 4000 series | Yes |
| GeForce GTX series | No |
| AMD or Intel GPUs | No |
OS and driver requirements
Your system needs Windows 10 or Windows 11 (64-bit) to run NVIDIA Broadcast. The app does not support macOS or Linux at this time. On the driver side, you need Game Ready Driver version 456.38 or newer, or Studio Driver 456.38 or newer. Older drivers will cause the installation to fail or the noise filters to not load when you open the app.
To update your drivers, open GeForce Experience, click the Drivers tab, and install the latest available version. If you prefer a clean install, download the driver directly from the NVIDIA website, select Custom Install, and check the box for "Perform a clean installation." Restart your PC after the driver update finishes before you move to Step 2. Skipping that restart causes conflicts that show up as a blank effects panel or missing filter options inside the app, which wastes time troubleshooting something that a simple reboot would have prevented.
Step 2. Install NVIDIA Broadcast and pick devices
With your drivers updated and your GPU confirmed, you’re ready to get NVIDIA Broadcast installed. The download file is roughly 200 MB and the full installation takes about five minutes, so this step moves quickly as long as your system meets the requirements from Step 1.
Download and install the app
Head to the official NVIDIA Broadcast page at nvidia.com/en-us/geforce/broadcasting/broadcast-app and click Download Now. Run the installer with administrator privileges. Follow the on-screen prompts, accept the license agreement, and let the installer finish without interrupting it. When it completes, restart your PC before opening the app. Skipping the restart causes device detection to fail, which shows up as empty dropdown menus when you launch NVIDIA Broadcast for the first time.
If GeForce Experience prompts you to update it alongside NVIDIA Broadcast, complete that update first. Running an outdated version of GeForce Experience alongside a fresh Broadcast install can cause conflicts in the effects panel.
Select your input and output devices
When NVIDIA Broadcast opens, you land directly on the Microphone tab. This is where you configure nvidia broadcast noise removal filters, and device selection must happen before the AI effects become available. At the top of the tab, you’ll find two dropdown menus you need to configure:
- Microphone (source): Select your physical microphone or audio interface. This must be your actual hardware device, not a virtual cable or loopback device.
- Speakers (output): Select the headphones or speakers you monitor through during a session. This controls what you hear, not what gets recorded.
Choose your real hardware microphone from the source dropdown. If it appears more than once in the list, pick the entry that matches the exact product name shown in your Windows Sound settings. After you select the source, NVIDIA Broadcast loads its AI processing model, which takes a few seconds on first run. You will see a brief loading indicator before the effect controls appear below the device selectors.
Once the model finishes loading, the filter toggle buttons become active, and the app is ready for configuration. Keep your headphones selected as the monitoring output so you can hear the difference as soon as you enable filters in Step 4.
Step 3. Route audio through the Broadcast mic
Installing NVIDIA Broadcast and enabling filters is only half the job. The processed audio does not automatically flow into your recording or communication apps. You need to route it manually by selecting the NVIDIA Broadcast virtual microphone as the audio input in Windows and in each app you use for investigations, streams, or calls.
NVIDIA Broadcast creates a virtual microphone device called "Microphone (NVIDIA Broadcast)" that acts as a clean audio pass-through for any app on your system.
Set the virtual microphone in Windows Sound settings
Open your Windows Sound settings by right-clicking the speaker icon in the taskbar and selecting Sound settings. Scroll to the Input section and set the default input device to "Microphone (NVIDIA Broadcast)". Setting this as your Windows default ensures any app that inherits the system default will automatically receive the processed audio without additional configuration.
To confirm the routing is working, speak into your microphone and watch the input level bar in Windows Sound settings. The bar should respond to your voice, confirming that the processed signal is passing through the virtual device. If the bar shows no activity, return to the NVIDIA Broadcast app and verify that a physical microphone is still selected as the source in the Microphone tab.
Route the virtual mic in individual apps
Some apps override the Windows default and use their own saved audio device. You need to manually update the microphone input inside each app to point at "Microphone (NVIDIA Broadcast)". Here is where to find that setting in the most common tools:
| App | Where to change microphone input |
|---|---|
| OBS Studio | Settings > Audio > Mic/Auxiliary Audio |
| Discord | User Settings > Voice & Video > Input Device |
| Microsoft Teams | Settings > Devices > Microphone |
| Zoom | Settings > Audio > Microphone |
| Audacity | Edit > Preferences > Devices > Recording > Device |
After updating each app, do a quick test recording or join a voice channel to confirm the selected device is active. With nvidia broadcast noise removal filters enabled in the Broadcast app, every app you update here will receive the cleaned audio signal automatically, even when you switch between them mid-session.
Step 4. Enable noise removal and room echo removal
With your virtual microphone routed into your apps, you’re ready to activate the nvidia broadcast noise removal filters. Both filters live in the Microphone tab inside NVIDIA Broadcast, directly below the device selectors you configured in Step 2. You will see two labeled rows, one for Noise Removal and one for Room Echo Removal, each with a toggle switch on the left side.
Turn on noise removal
Click the toggle switch next to Noise Removal to enable it. The switch turns blue and a small AI indicator appears, confirming that the GPU is actively processing your microphone signal. Once activated, you should notice an immediate reduction in steady background sounds like fan noise, air conditioning hum, or ambient room tone.
Enabling Noise Removal before Room Echo Removal gives you a cleaner baseline signal, which makes the echo filter more effective when you stack them together.
To confirm the filter is working, speak at your normal volume while making a deliberate noise in the background, like snapping your fingers near the microphone or tapping the desk. The background sound should drop significantly in your monitoring output while your voice stays clear. If you hear no difference, check that your physical microphone is still set as the source device in the Broadcast app.
Turn on room echo removal
Room Echo Removal activates the same way, using its own toggle switch in the Microphone tab directly below the Noise Removal row. Enable it after Noise Removal is already running so you can hear the difference each filter adds independently. This filter targets the reverb your voice picks up in untreated rooms, which is common in the kinds of locations paranormal investigators typically work in, such as large empty buildings, stone basements, and abandoned structures.
Running both filters together produces the cleanest output. Here is a summary of what each filter state delivers:
| Filter State | Audio Result |
|---|---|
| Neither filter active | Raw microphone signal, full ambient noise |
| Noise Removal only | Ambient noise reduced, reverb still present |
| Room Echo Removal only | Echo reduced, ambient noise still present |
| Both filters active | Cleanest signal, ambient noise and echo both removed |
Step 5. Dial in strength, gain, and monitoring
Both filters are running now, but the default settings are a starting point, not a final configuration. NVIDIA Broadcast noise removal gives you a dedicated intensity slider for each filter, and adjusting these values makes the difference between audio that sounds natural and audio that sounds hollow or overprocessed. Spend two minutes in this step and your recordings will reflect it.
Adjust filter intensity
Each filter row in the Microphone tab has an intensity slider that controls how aggressively the AI model processes your signal. The slider runs from low to high, and the right setting depends on how noisy your environment is. Start both sliders at around 60 percent and listen to your monitoring output while making noise in the background.
Setting intensity too high causes your voice to sound thin and slightly robotic, which is a common artifact when the model works harder than the environment requires.
Use the following settings as a baseline and move from there based on what you hear:
| Environment | Noise Removal Intensity | Room Echo Removal Intensity |
|---|---|---|
| Quiet home office | 40-50% | 30-40% |
| Typical indoor investigation | 60-70% | 50-60% |
| Noisy environment (fans, AC, wind) | 80-90% | 60-70% |
| Large reverberant building | 70-80% | 80-90% |
Drop the intensity if your voice sounds unnatural. Raise it only if audible background noise is still bleeding through at lower settings. The goal is the lowest intensity that still removes the noise you’re targeting.
Set input gain and monitor the result
Your physical microphone’s input gain level in Windows Sound settings directly affects how hard NVIDIA Broadcast works. Open Sound settings, select your physical device under Input, and click Device properties. Set the input level between 70 and 85 percent to give the AI model a clean, strong signal without clipping.
Once your gain is set, put on headphones and enable the "Listen to this device" option under the NVIDIA Broadcast virtual microphone in Windows Sound settings. Speak normally, move around the space you plan to investigate in, and make noise near the microphone to simulate real session conditions. Adjust the intensity sliders in real time until the result sounds clean and natural at the same time.
Step 6. Use NVIDIA Broadcast in OBS and calls
Routing nvidia broadcast noise removal through the virtual microphone handles most apps automatically, but OBS Studio requires one extra configuration step to work correctly. Voice call apps like Discord and Zoom also have a common pitfall that silently bypasses the virtual device if you do not catch it. This step walks you through both scenarios so your cleaned audio reaches the right destination every time.
Configure NVIDIA Broadcast audio in OBS Studio
Open OBS Studio and navigate to Settings, then select the Audio tab. Under the Desk-1 or Mic/Auxiliary Audio dropdown, select "Microphone (NVIDIA Broadcast)" and click Apply. If you already had a microphone source added as a scene item rather than a global audio device, right-click that source in your sources list, choose Properties, and switch the device to the NVIDIA Broadcast virtual mic there instead.
Once you apply the change, check the OBS audio mixer at the bottom of the main window to confirm the meter is responding when you speak.
After the device is set, add a Gain filter inside OBS if your levels look low. Right-click the audio track in the mixer, select Filters, and add a Gain filter set between 3 and 6 dB. This compensates for any level reduction the NVIDIA processing introduces without you needing to raise your physical microphone input above the 85 percent ceiling from Step 5.
Set the virtual mic in Discord, Zoom, and Teams
Each voice call app stores its own microphone preference independent of Windows defaults, so you need to update each one manually. The table below shows exactly where to find the setting in the three most common apps investigators use for team coordination.
| App | Setting Path |
|---|---|
| Discord | User Settings > Voice & Video > Input Device |
| Zoom | Settings > Audio > Microphone |
| Microsoft Teams | Settings > Devices > Microphone |
After updating the input device, run a quick test call or use the app’s built-in mic test feature to confirm the virtual device is active and passing audio. Discord includes a "Let’s Check" microphone test under Voice & Video that plays back your input in real time, which makes it easy to verify the cleaned signal is coming through before you start a live session with your investigation team.
Step 7. Fix common issues and avoid artifacts
NVIDIA broadcast noise removal works reliably once it’s configured correctly, but a few specific problems come up regularly in the field. Most issues fall into one of three categories: the processed voice sounds unnatural, the virtual microphone disappears from your app, or you notice a delay between speaking and hearing the output. Each has a direct fix.
Voice sounds robotic or hollow
This is the most common artifact, and the cause is almost always intensity set too high for the actual noise level in your environment. When the AI model processes a signal that is cleaner than expected, it starts removing subtle voice harmonics along with the noise, which produces that hollow or over-processed quality.
Drop the Noise Removal intensity slider by 10 to 15 percent at a time and re-test until your voice sounds natural again.
Lower the intensity slider first before making any other changes. If you are also running Room Echo Removal at high intensity in a moderately treated space, reduce that slider as well. The fix is almost always in the intensity settings, not the device configuration.
Virtual microphone not detected
Some apps stop seeing the "Microphone (NVIDIA Broadcast)" virtual device after a Windows update or after NVIDIA Broadcast restarts during a session. The fastest fix is to close the app, restart NVIDIA Broadcast, and then relaunch the app so it scans for available devices again.
If the virtual microphone still does not appear, open Windows Device Manager, expand the Audio inputs and outputs section, and check whether the NVIDIA Broadcast device is listed. A yellow warning icon means the virtual driver needs reinstalling. Uninstall NVIDIA Broadcast fully through Windows Apps settings, restart your PC, then reinstall the app from the NVIDIA website.
Audio delay or latency spikes
Latency in NVIDIA Broadcast comes from GPU load, not the app itself. If your GPU is rendering a game or processing video at the same time as running the AI filters, processing time increases and you hear a delay.
Close background GPU tasks before starting a session. Running only NVIDIA Broadcast alongside your recording or streaming app keeps GPU headroom high and latency low throughout the investigation.
Wrap up and next steps
You now have nvidia broadcast noise removal fully configured, from GPU verification and installation through filter tuning, app routing, and troubleshooting. Your microphone signal goes through the AI model, strips out ambient noise and room echo, and delivers clean audio to every app you connect it to. That cleaner signal means less noise competing with faint audio anomalies during EVP sessions and sharper communication with your team on location.
Fine-tuning takes a session or two to dial in completely. Run a short test recording before each investigation and adjust the intensity sliders based on the specific environment you’re working in that night. Noisy basements and reverberant stone buildings need different settings than a quiet indoor location, so treat each space as its own calibration opportunity.
Better audio software works best alongside reliable field equipment. Browse the paranormal investigation gear at Haunt Gears to find tools built for serious investigators.
