AI in Video Calls: Background Blur, Noise Cancellation & More
By Marcin Piekarski builtweb.com.au · Last Updated: 11 February 2026
TL;DR: Zoom, Teams, and Google Meet use AI to blur backgrounds, cancel noise, improve lighting, and more. Learn what's happening behind the scenes.
TL;DR
Every major video calling app uses AI to make you look and sound better. Background blur separates you from your messy room. Noise cancellation removes barking dogs and construction sounds. Auto-framing keeps you centred even when you move. Live captions convert speech to text in real time. All of this happens on your computer or phone during the call, processing dozens of video frames and audio samples every second.
Why it matters
Video calls went from a convenience to a necessity during the pandemic, and they are not going away. Whether you work remotely, attend virtual meetings, or call family overseas, you spend hours on video every week.
The difference between a good video call and a terrible one often comes down to AI. Without noise cancellation, a barking dog makes you inaudible. Without background blur, you are constantly self-conscious about your surroundings. Without auto-framing, you look off-centre and unprofessional on a laptop placed at an odd angle.
These AI features are not gimmicks. They solve real problems that make video communication harder, and understanding how they work helps you get the best results from your setup.
How background blur and replacement works
Background blur is one of the most visible AI features in video calls. Here is what happens behind the scenes every time it is active.
The AI analyses each video frame (typically 30 frames per second) and performs a task called "semantic segmentation." This means it classifies every pixel in the image as either "person" or "background." It identifies your head, shoulders, arms, hands, and torso, creating a detailed outline that separates you from everything behind you.
Once it knows where you are and where the background is, it applies a blur effect to the background pixels while leaving you sharp. For background replacement (using a virtual beach or office), it replaces the background pixels with the chosen image instead of blurring them.
The tricky part is getting the edges right. Your hair, fingers, and the space between your arm and body are all difficult boundaries. Early background blur implementations often made hair look like a fuzzy halo or caused arms to disappear when raised. Modern AI models have gotten much better at these edge cases, though you will still notice occasional glitches with fast movements or complex backgrounds.
The AI model that does this segmentation runs directly on your device. It needs to be fast enough to process 30 frames per second without noticeable delay, which is why it uses your computer's GPU or a dedicated AI chip when available.
How AI noise cancellation works
Noise cancellation is arguably the most useful AI feature in video calls, and it works differently than you might expect.
Traditional noise cancellation (like in headphones) uses inverse sound waves to cancel out noise. AI noise cancellation in video calls takes a different approach. It is trained on massive datasets of human speech mixed with various background noises — dogs, children, traffic, keyboards, construction, doorbells, washing machines. The AI learns to recognise the patterns of human speech and separate them from everything else.
During a call, the AI processes your audio in real time. It identifies which parts of the audio signal are your voice and which parts are noise. Then it strips out the noise and transmits only your voice to the other participants.
The results can be remarkable. Krisp, Nvidia RTX Voice, and the built-in noise cancellation in Zoom, Teams, and Meet can all eliminate loud background noise so effectively that the other person has no idea your dog was barking right next to you.
There are limitations. If you try to intentionally share audio (like playing a song), the AI might filter it out because it does not recognise it as speech. Some systems have trouble with multiple voices in the same room, occasionally cutting out parts of your speech that overlap with other speakers.
Auto-framing and eye contact correction
Auto-framing solves a common problem: you set up your laptop at an angle, sit off to one side, or lean back in your chair, and suddenly you are barely visible in the corner of the frame. Auto-framing AI detects your face and body position and digitally adjusts the camera view to keep you centred and properly framed.
On devices with multiple cameras or wide-angle lenses (like Apple's Centre Stage on iPads), auto-framing can also follow you as you move around a room. The camera physically stays still, but the AI crops and pans the digital image to track you smoothly.
Eye contact correction is a newer feature offered by tools like Nvidia Broadcast and Apple's FaceTime. The problem it solves is subtle but significant: when you look at your screen to see the other person, your camera sees you looking down rather than straight at it. This makes it feel like you are not making eye contact.
Eye contact AI subtly adjusts the position of your eyes in the video to make it look like you are looking directly at the camera. It is a small change, but it makes video conversations feel noticeably more natural and engaged.
Lighting adjustment and video enhancement
AI lighting correction brightens your face when your room is dark or when a window behind you creates a silhouette effect. The AI detects your face and selectively brightens and colour-corrects it without over-processing the rest of the image.
Some platforms go further with AI-powered video enhancement. Google Meet's "studio look" and similar features adjust contrast, smooth skin texture slightly, and optimise colours to make you look more like you are in a professional studio than a home office.
These features make a real difference when you do not have ideal lighting, which is most of the time. A single overhead light creates harsh shadows under your eyes. A window behind you turns you into a silhouette. AI corrects both of these common problems automatically.
Live captions and transcription
Real-time captions convert spoken words into text overlaid on the video call. This AI feature serves multiple purposes.
For hearing-impaired participants, live captions make video calls accessible. For people in noisy environments who cannot use speakers, captions let them follow the conversation silently. For non-native speakers, seeing words written out alongside hearing them improves comprehension. And for everyone, captions help when audio quality is poor.
The AI behind live captions uses speech recognition models similar to those powering voice assistants. They process audio in real time, converting speech to text with increasingly impressive accuracy. Modern systems handle different accents, speech patterns, and even multiple languages.
Many platforms also offer meeting transcription — a complete written record of the call that you can review later. AI-powered meeting summaries can even extract action items, key decisions, and talking points from the transcription.
Performance and hardware considerations
AI features in video calls require processing power. Background blur, noise cancellation, and auto-framing all run simultaneously during a call, and each one consumes CPU or GPU resources.
On modern computers (made in the last three to four years), this is rarely a problem. Newer chips include dedicated AI processing units that handle these tasks efficiently. Apple's M-series chips, Intel's NPU-equipped processors, and Nvidia GPUs all accelerate AI video call features.
On older hardware, you might notice your computer getting warm, the fan running louder, or the video becoming choppy when multiple AI features are active. If this happens, try disabling features you do not need. Background blur typically uses the most processing power, so turning that off first usually gives the biggest improvement.
Tips for getting the best results
Use a front-facing light source. AI lighting correction helps, but starting with good light makes everything better. Face a window or place a lamp behind your monitor, pointing at your face. This gives the AI less work to do and produces a more natural result.
Test background blur before important calls. Not all implementations are equal. Test with your usual setup and check for glitches — disappearing hands, flickering edges, or slow response when you move.
Keep your software updated. Video call AI improves significantly with each update. Companies continuously retrain their models and optimise performance. Running an old version means missing out on better quality and lower resource usage.
Reduce unnecessary AI features to save resources. If your computer is struggling, disable features in order of processing cost: background replacement first (more expensive than blur), then background blur, then auto-framing. Keep noise cancellation on — it uses less processing power and makes the biggest difference to call quality.
Common mistakes
Relying on background blur for privacy. Background blur occasionally glitches, revealing parts of your background. Do not count on it to hide sensitive information (whiteboards with confidential data, documents on your desk). Position your camera so the real background is acceptable even if blur fails.
Disabling noise cancellation to share audio. If you want to share music or a video during a call, use the screen share audio feature instead of disabling noise cancellation. Disabling it exposes all your background noise for the rest of the call.
Ignoring bandwidth. AI features process video locally, but the video still needs to travel over the internet. A slow connection causes more quality issues than any AI feature can fix. Use a wired connection or sit close to your router for important calls.
Over-processing your video. Turning on every AI enhancement simultaneously (blur, lighting correction, beauty mode, eye contact) can make you look artificial. Use only the features that solve a real problem with your setup.
What's next?
- AI in Everyday Life — More places AI works behind the scenes in your daily routine
- Voice Assistants Explained — How speech recognition technology powers more than video calls
- AI in Smartphones — AI features in the device you use for mobile video calls
Frequently Asked Questions
Does background blur work well on older computers?
It depends on how old. Computers from the last three to four years typically handle background blur without issues. Older machines may experience lag, overheating, or choppy video. If performance is a problem, try reducing video resolution to 720p, closing other applications, or using a platform with lighter AI processing (Google Meet tends to be less resource-intensive than Zoom's virtual backgrounds).
Can AI noise cancellation remove all background noise?
It removes most background noise remarkably well — dogs, traffic, keyboards, construction. However, it can struggle with noise that overlaps with your voice frequency (like another person talking nearby) or very loud continuous sounds. It also cannot enhance poor microphone quality. Using a decent headset with a microphone gives the best results when combined with AI noise cancellation.
Is my video data sent to the cloud for AI processing?
For most platforms (Zoom, Teams, Google Meet), the AI features like background blur and noise cancellation run locally on your device. Your raw video with your real background is not sent to the company's servers. However, the processed video and audio are transmitted to other call participants through the platform's servers, as with any video call.
Why does background blur sometimes make my hair or hands disappear?
The AI needs to draw a precise boundary between you and the background, and thin or fast-moving features like hair strands, fingers, and arms in motion are difficult to classify accurately at 30 frames per second. Better lighting (especially a light behind your camera pointing at you) helps the AI distinguish edges more accurately. Sitting still and avoiding rapid hand gestures also reduces glitches.
Was this guide helpful?
Your feedback helps us improve our guides
About the Authors
Marcin Piekarski· Frontend Lead & AI Educator
Marcin is a Frontend Lead with 20+ years in tech. Currently building headless ecommerce at Harvey Norman (Next.js, Node.js, GraphQL). He created Field Guide to AI to help others understand AI tools practically—without the jargon.
Credentials & Experience:
- 20+ years web development experience
- Frontend Lead at Harvey Norman (10 years)
- Worked with: Gumtree, CommBank, Woolworths, Optus, M&C Saatchi
- Runs AI workshops for teams
- Founder of builtweb.com.au
- Daily AI tools user: ChatGPT, Claude, Gemini, AI coding assistants
- Specializes in React ecosystem: React, Next.js, Node.js
Areas of Expertise:
Prism AI· AI Research & Writing Assistant
Prism AI is the AI ghostwriter behind Field Guide to AI—a collaborative ensemble of frontier models (Claude, ChatGPT, Gemini, and others) that assist with research, drafting, and content synthesis. Like light through a prism, human expertise is refracted through multiple AI perspectives to create clear, comprehensive guides. All AI-generated content is reviewed, fact-checked, and refined by Marcin before publication.
Transparency Note: All AI-assisted content is thoroughly reviewed, fact-checked, and refined by Marcin Piekarski before publication.
Key Terms Used in This Guide
Related Guides
AI Spam Filters: Keeping Your Inbox Clean
BeginnerEver wonder how Gmail blocks 99%+ of spam? AI spam filters analyze billions of emails to catch junk before it reaches you.
4 min readAI Accessibility Features: Technology for Everyone
BeginnerAI makes technology accessible to people with disabilities—from screen readers to voice control to live captions. Discover how AI levels the playing field.
5 min readAI in Social Media: How Algorithms Shape What You See
BeginnerYour social media feed isn't random—AI decides what you see, when, and why. Learn how these algorithms work and how to take back control.
5 min read