What It Really Costs to Add Video Chat to Your Website in 2026

Adding video chat to a website costs almost nothing to prototype and a surprising amount to run. WebRTC, the technology behind browser video calls, is free and already built into every modern browser. The bills come from what sits behind it: TURN relay bandwidth, a signaling server that can never go down, media servers for group calls, moderation for live sessions and, in more markets every year, age checks.

The short answer: a small one-to-one call feature can run on infrastructure costing anywhere from nearly nothing to a few hundred dollars a month. Managed video APIs list at roughly $3.50 to $4 per 1,000 participant-minutes for HD video in 2026. Once strangers can talk to each other on your site, automated moderation and age assurance usually become the largest lines in the budget, often bigger than everything else combined. Every vendor price in this guide was checked against the vendor’s own pricing page in October 2026.

This guide breaks the cost down line by line, with current list prices and a worked monthly example, so you can quote a client a number that survives launch. The same math applies whether you want to add video chat to a website built on WordPress or are estimating the cost to build a video chat app from scratch.

Live Video Chat or Recorded Video? Check Before You Budget

A surprising number of “we want video on the site” requests turn out to mean recorded video: a product demo, an onboarding tour, a founder’s welcome message. Recorded video streams from a host or CDN, caches well and costs a fraction of anything live. If that is the real brief, our guide to integrating custom video into WordPress covers placement, loading speed and accessibility without any real-time infrastructure at all.

Live video chat is a different animal. Two or more people see and hear each other in real time, and nothing can be cached. It comes in three shapes, and each one is priced differently:

• One-to-one calls between known users, such as a coach and a client or a tutor and a student. Cheapest to run and easiest to moderate.
• Group calls, such as live classes or member meetups. These need a media server, which changes the bandwidth math completely.
• Calls between strangers, such as random video chat or social discovery. The most expensive shape by far, because moderation and age checks stop being optional.

Pin down which shape the client actually needs before anyone writes a line of code. Every number below depends on it.

Why a Video Chat Demo Is Cheap and Production Is Not

WebRTC ships inside Chrome, Safari, Firefox and Edge. Nobody pays a license fee to use it, and visitors need no plugin or install. A competent developer can build a working two-person call in a weekend.

That weekend demo is where budgets go wrong. It usually connects two browsers on the same office network, directly, with no relay, no moderation and no age gate. The client watches it work and assumes the feature is nearly finished. Real users arrive from mobile carriers, hotel Wi-Fi and locked-down corporate networks, and a meaningful share of them cannot connect directly at all. The gap between the demo and production is where the money goes.

The Cost Centers at a Glance

Cost center	What drives it	2026 ballpark
TURN relay bandwidth	The 10% to 25% of sessions that cannot connect directly	$0.05 to $0.80 per GB at list price; about 1.4 to 2.3 GB per relayed 720p call-hour
Group call routing (SFU)	Participants per call	Grows with every extra person on the call
Signaling server	Uptime, not capacity	Cheap to run, expensive when it goes down
Moderation	Minutes of live video between users	Usually the largest line once strangers meet
Age assurance	Signups, plus lost conversion	Small per check, large in drop-off
Maintenance	Browser, OS and SDK updates	Developer time every month, forever

Each line is explained below, with the numbers you need to estimate it.

Cost 1: TURN Relay Servers (The Bandwidth Bill)

WebRTC always tries a direct peer-to-peer path first, using STUN to discover each device’s public address. That works for most sessions. Somewhere between 10% and 25% of them fail, depending on the audience, because strict mobile carrier NAT or a corporate firewall blocks the direct route. One in five is a common planning figure, and audiences heavy on business users or mobile data tend to sit at the high end.

Those sessions fall back to a TURN server, which relays the full audio and video stream in both directions. Every gigabyte that passes through it is billed.

How Much Data Does a Video Call Use Per Hour?

The arithmetic is simple once you know the bitrate. A stream running at 1 Mbps moves about 0.45 GB per hour. A 720p WebRTC call typically runs at 1.5 to 2.5 Mbps in each direction, so one relayed one-to-one call pushes roughly 1.4 to 2.3 GB through the relay every hour. Some providers meter traffic on both sides of the relay, which can double that figure.

Here is how that works out at common quality levels, using typical WebRTC bitrates:

Call quality	Typical bitrate per stream	Data per stream per hour	Relayed one-to-one call per hour
Audio only	32 to 64 kbps	0.015 to 0.03 GB	0.03 to 0.06 GB
360p video	About 0.5 Mbps	About 0.23 GB	About 0.45 GB
540p video	About 1 Mbps	About 0.45 GB	About 0.9 GB
720p video	1.5 to 2.5 Mbps	0.68 to 1.13 GB	1.4 to 2.3 GB
1080p video	3 to 4 Mbps	1.35 to 1.8 GB	2.7 to 3.6 GB

Real bitrates move with the codec, the amount of motion on screen and network conditions, so treat these as planning figures rather than guarantees.

You will often see “about 1 GB per hour” quoted as a rule of thumb. That only holds for lower-resolution video, around the 540p row above. If your product allows HD, budget for the higher number, and consider capping resolution: limiting calls to 720p is one of the cheapest ways to cut a relay bill.

TURN Server Pricing in 2026

List prices for managed TURN vary enormously. Cloudflare Realtime charges $0.05 per GB of egress, with the first 1,000 GB each month free (shared with its SFU service). Twilio’s Network Traversal Service lists $0.40 per GB in the US and Europe, rising to $0.60 in parts of Asia and $0.80 in São Paulo and Sydney. The same traffic can therefore cost eight to sixteen times more depending on the provider and region you pick, so price your users’ actual geography rather than a single headline rate.

Self-Hosted coturn or a Managed TURN Service?

You can run TURN yourself with coturn, the open-source server most self-hosted setups use. Your per-gigabyte cost drops to whatever your cloud provider charges for egress. In exchange you take on servers that must be monitored, patched, scaled and placed close to your users, plus TURN over TLS on port 443 for the strictest corporate firewalls. For a small team, a managed provider is usually cheaper once you price your own time honestly.

One testing habit saves a lot of grief here: force every call through the relay during development by setting iceTransportPolicy: 'relay', then confirm the selected candidate in chrome://webrtc-internals. If relayed calls do not work in staging, one in five of your real users will not connect in production.

Cost 2: Group Calls Need an SFU

Peer-to-peer works well for two people. With three or more, a “mesh” setup makes every browser upload a separate copy of its video to every other participant, and phones start struggling at around four people. Group calls are normally routed through a Selective Forwarding Unit (SFU), a media server that receives each stream once and forwards it to everyone else.

The catch is that SFU bandwidth grows much faster than headcount. In a four-person call at 1 Mbps per video stream, each participant receives three streams, so the server sends out about 12 Mbps, or roughly 5.4 GB per hour. Double the call to eight people and each one receives seven streams: about 56 Mbps and 25 GB per hour. Techniques such as simulcast and showing only the active speakers keep this in check, but they need to be designed in from the start.

Open-source SFUs such as LiveKit, mediasoup and Janus are free to run on your own servers. Managed options bill per gigabyte or per participant-minute. Either way, a feature request that says “group calls” belongs in a different budget bracket from one that says “calls”.

Cost 3: The Signaling Server

Before two browsers can connect, they swap connection details through your signaling server. Signaling uses very little bandwidth, so the direct cost stays low, usually a small pair of servers for redundancy. Availability matters far more than capacity. If signaling goes down, nobody can start a call, even though every other part of your stack is healthy.

Most teams run signaling as a small WebSocket service in Node.js or Go, kept separate from the main website host so a traffic spike on the site cannot take calls offline. That separation matters on WordPress in particular, because PHP’s short request cycle is a poor fit for thousands of long-lived connections. A sensible pattern is to let WordPress handle accounts and issue short-lived tokens, and let the signaling service do the real-time work. If you build that bridge as a custom plugin, keep it lean; our performance optimization tips for custom WordPress plugins apply directly.

Use the same token logic for TURN. Generate time-limited credentials for each session instead of embedding a permanent username and password in your JavaScript, or strangers will find your relay and run their own traffic through it on your bill.

Cost 4: Moderation, the Line Most Quotes Leave Out

This is where video chat projects fail, and it is almost always missing from the original quote.

Text and images can be scanned after they are posted. Live video happens once and is gone. If users who do not know each other can talk on your site, you need moderation during the session itself: automated screening, human reviewers, or both. Neither works without a report button, a blocking system, a way to stop banned users from coming back under a new account, and somebody who actually reads the reports.

What Automated Video Moderation Costs

A common approach is to sample frames from each live stream and send them to an image moderation API. AWS puts the default price of Amazon Rekognition’s image moderation API at about $0.001 per image. Sampling one frame every five seconds from each person in a one-to-one call means 24 images per call-minute, or roughly $0.024 per minute of conversation. At 5,000 call-hours a month, that is 7.2 million images and up to about $7,200, somewhat less once volume tiers apply. That is before a single human reviewer is paid to handle flags, appeals and user reports.

There are legal duties behind this as well. In the US, online service providers who become aware of child sexual abuse material must report it to NCMEC, and you need a documented process for preserving evidence and escalating it. Build that process before launch, not after the first incident.

Why Retrofitting Moderation Is Expensive

Adding moderation to a live product means rebuilding session handling, storage and admin tools at the same time, while users are on the platform. Planning it before launch costs a fraction of that.

Reports, blocks and appeals are workflows like any other, and each one needs an owner and an end-to-end test before go-live. It is the same discipline our website redesign checklist for the workflows that break after launch applies to forms, bookings and payments: if a workflow has no owner, it does not really exist yet. Moderation cost also scales with usage rather than headcount, which makes it behave unlike almost every other line in a typical web project budget.

Cost 5: Age Assurance and Privacy Law

Compliance has changed the math for any site that connects people over video. In the UK, the Online Safety Act covers user-to-user services, which includes most live video chat between users. Services must assess whether children are likely to use them and, if so, the risks they face, and those that allow pornography or other content most harmful to children must use what Ofcom calls highly effective age assurance. Self-declaration, a simple “I am over 18” checkbox, does not qualify. Ofcom’s guidance on highly effective age checks lists methods it considers capable of meeting the bar, including facial age estimation, open banking, photo-ID matching, digital identity services, credit card checks, mobile network operator checks and email-based age estimation.

In the US, age verification rules are a state-by-state patchwork aimed mostly at adult content and social media, and the Supreme Court upheld Texas’s age-verification law for adult sites in 2025. In the EU, the Digital Services Act adds its own duties to protect minors on online platforms. If your client serves several markets, check each one and get legal advice for the strictest.

Each verification check costs a small amount per user. The bigger cost is conversion. Every verification step removes a share of signups, and a product that depends on enough people being online at the same time suffers twice from that drop. Facial age estimation also brings privacy obligations: face data can count as biometric data under the GDPR and laws such as Illinois’s Biometric Information Privacy Act, so consent, retention limits and vendor contracts belong in the budget too.

The Hardest Case: Video Chat Between Strangers

Random video chat and social discovery carry every cost above at its maximum. Audiences are mobile-heavy, which pushes more sessions onto relays. Users are anonymous, which makes moderation essential. And because anyone can sign up, age assurance moves from a consideration to a requirement in a growing list of markets.

Omegle is the cautionary tale. The random chat site reached tens of millions of monthly users before its founder shut it down in November 2023, saying that running it was “no longer sustainable, financially nor psychologically” after years of fighting misuse and the lawsuits that followed. Scale did not make the costs easier. It made them larger.

That is why a site whose audience wants to meet new people on camera, such as a language-exchange blog or a hobby community, is usually better off pointing visitors to an established video chat platform built and staffed for that job than trying to build one. Before you link to any platform, check its minimum age, its reporting and blocking tools, and its published safety and child protection policies. Never send an audience that includes under-18s to stranger chat.

Editor’s note: we mention this platform as one example of a dedicated social video chat service. It is not an endorsement, and we have no commercial relationship with it. Review any platform’s safety policies yourself before recommending it to your audience.

Build, Buy or Link Out? How to Decide

Start with one question. Is live video your product, or a feature of something else?

If it is the product, build it properly. Budget for relay bandwidth, moderation and compliance from day one, and treat all three as operating costs that grow with success. If it is a feature, one of the cheaper routes below is usually the better call. Keep in mind that development is a separate bill: published agency estimates for a basic video chat MVP mostly fall between about $20,000 and $50,000, before any of the running costs above.

Route A: A Managed Video API

Providers such as Daily, Agora and Zoom’s Video SDK handle relays, media servers, SDKs and scaling for you. Their 2026 list prices cluster around $3.50 to $4 per 1,000 participant-minutes for HD video, usually after a free monthly allowance. Daily, for example, lists $0.004 per participant-minute after 10,000 free minutes a month, with automatic volume discounts from 100,000 minutes. Recording, transcription and compliance features often cost extra.

This suits teams who need calling inside their own product and do not want to run servers. The trade-off is margin. Per-minute pricing is gentle at low volume and becomes the dominant cost at high volume, which is when teams start looking at Route B.

Route B: Open-Source WebRTC on Your Own Servers

An open-source SFU such as LiveKit or mediasoup, paired with coturn or a cheap managed TURN service, can cut per-minute costs dramatically at scale. What you save on the invoice you spend on engineers: deployment, monitoring, upgrades and on-call cover. It pays off once volume is high and predictable, rarely before.

Route C: Link Out to a Tool People Already Use

A community site, course platform or membership site whose members need to call each other can often skip the build entirely. Scheduled sessions on Google Meet, Zoom or Jitsi Meet, or a Discord server for the community, keep the social benefit for members and avoid relay bills, moderation staffing and most of the compliance burden. Some of them, Jitsi Meet among them, also let you embed video chat in a page if you want calls to feel native.

It sounds unambitious. It is also the right answer far more often than developers like to admit, particularly when calling is a nice-to-have rather than the reason people visit.

A Worked Example: 5,000 Call-Hours a Month

Picture a coaching membership site running one-to-one 720p calls at about 2 Mbps in each direction. Members log 5,000 call-hours a month, and 20% of calls need a relay. Here is how the main lines compare at 2026 list prices.

Line item	How it is calculated	Monthly cost
Relayed traffic	1,000 relayed hours × 1.8 GB per hour	1,800 GB
TURN on Cloudflare Realtime	(1,800 GB minus 1,000 free GB) × $0.05	About $40
TURN on Twilio (US or Europe)	1,800 GB × $0.40	About $720
Managed video API instead of DIY	600,000 participant-minutes × $0.0035 to $0.004, less volume discounts	About $2,000 to $2,400
Automated moderation (if users are strangers)	7.2 million sampled frames × up to $0.001	Up to about $7,200
Signaling	A small redundant server pair	Tens of dollars
Maintenance	Browser, SDK and security updates	Developer time, every month

Three lessons stand out. The relay bill alone swings eighteen-fold on provider choice. A managed API costs more than DIY bandwidth but replaces servers and engineering time, which for most small teams is the expensive part. And once strangers meet, automated moderation can outweigh every other line combined. For known users, such as a coach and a paying client, automated screening can be lighter, but reporting and blocking tools are still part of the build.

How These Numbers Were Calculated

Prices come from the vendors’ published list rates as checked in October 2026, before taxes, volume discounts or negotiated contracts, and each one is linked to its source where it first appears. Bandwidth uses the standard conversion of 0.45 GB per hour for each 1 Mbps stream. Participant-minutes count each person on a call separately, so one hour-long one-to-one call equals 120 participant-minutes. Real bills will differ with your audience’s networks, resolution caps and regions, so rerun the math with your own traffic before you quote.

A Checklist Before You Quote Video Chat to a Client

Run through these points before giving anyone a figure:

• Confirm the call shape: one-to-one, group, or strangers.
• Estimate monthly relayed minutes, not total minutes, and decide the maximum resolution.
• Confirm in writing who pays for bandwidth overage.
• Choose automated moderation, human review or both, and name who reads the reports.
• Check whether age assurance applies in each market the client serves.
• Budget for reporting, blocking and evidence-retention tools.
• Decide whether calls are recorded, and plan storage, consent and retention if they are.
• Include ongoing maintenance, not just the initial build.
• Agree what happens if usage grows ten times faster than forecast.

Video chat rarely arrives alone. It usually comes as part of a wider build, so pair this list with our guide on what a business should prepare before website development, which covers content, ownership and handover alongside features like this one.

Common Mistakes That Push Video Chat Over Budget

A few patterns show up again and again on projects that blow their budget.

Quoting from the prototype. The weekend demo used direct connections and no moderation. It is not a sample of production.

Testing only on office Wi-Fi. Carrier NAT pushes more sessions onto relays than office testing suggests. Test on mobile data, and force relay mode at least once.

Shipping with a free public TURN server. Free relays are fine for experiments. In production they bring rate limits, no uptime guarantee and no control over where your users’ media travels.

Allowing 1080p by default. Higher resolution raises every bandwidth line. Most one-to-one calls look fine at 720p.

Treating moderation as a phase two item. For any service where strangers meet, it belongs in the launch scope, and in the UK the Online Safety Act expects reporting tools from day one.

Ignoring the conversion cost of verification. Teams model the per-check fee and never model the signups they lose at that step.

The Bottom Line on Website Video Chat Costs

Video chat is not expensive to prototype. It is expensive to run safely at scale. The browser gives you the hard engineering for free, and in exchange you inherit bandwidth, moderation and compliance costs that grow in direct proportion to how successful the feature becomes.

Treat live video as an ongoing service rather than a one-off build. Price it that way from the start, choose the cheapest route that genuinely fits the use case, and the project will survive its first busy month.

What Site Owners Ask About WebRTC, TURN and Video Chat Pricing