Asset 20 8 2
Does AI recommend your business? Run the free check →

Join 15,000 business owners, marketers and entrepreneurs. The Sunday newsletter you'll be annoyed only arrives once a week.

Article

LiveKit Alternatives: 9 Real-Time Voice and Video Picks

The best LiveKit alternatives split into hosted WebRTC platforms and video SDKs billed by the minute, and open-source frameworks or servers you run yourself. Decide first whether you want to own the media servers; that one choice narrows nine options to three or four.

Are you replacing LiveKit because of the bill, because of the ops work, or because you're building a voice agent and want a different framework around it? Each answer leads to a different shortlist, because the tools below aren't one market: some of them carry audio and video, one is a framework for the AI conversation on top, and some are raw building blocks for your own servers.

Hosted or self-run: make this call first

Six of the nine are hosted services: you add an SDK, they run the media servers around the world, and you pay for the minutes your users spend connected. Three can run on your own infrastructure: two open-source projects and one streaming server you install yourself or launch from a cloud provider's catalogue.

Choose hosted if your team is small, your users are spread across regions, or you'd rather not be paged at 3am because a TURN server ran out of bandwidth. Hosted platforms handle the routing, recording and global edge for you. The price is a per-minute bill that grows in a straight line with use, and less say over where media travels.

Choose self-run if you have steady, high volume where per-minute pricing starts to look silly, strict rules about where audio and video may be processed, or a product that needs control over the media pipeline the hosted APIs won't give you. The price is people: someone who understands WebRTC, NAT traversal, codecs and autoscaling, on call.

A useful middle path: start hosted, keep your own code behind a thin wrapper around the SDK, and revisit the decision when the monthly bill reaches the cost of the engineer you'd hire to run it yourself. The open-source side is covered in more depth in the roundup of open-source video conferencing software.

Hosted platforms and SDKs

Daily

From the team behind Pipecat, a hosted platform for real-time voice, video and AI features aimed squarely at developers.

Agora

One of the bigger names here, offering voice, video, chat and voice AI APIs, plus low-code tools for teams wanting less custom work.

Stream Video

In-app calling, both video and audio, delivered through an API and a set of SDKs for teams who want calling inside their own product.

100ms

Video calls and livestreaming inside your app, with a customisable interface and APIs for managing rooms.

VideoSDK

Building a platform that needs embedded calls or AI voice agents? This smaller vendor sells audio, video and live streaming APIs for exactly that.

ZEGOCLOUD

Voice, video, chat and live streaming APIs bundled together for real-time interaction inside your own app.

Frameworks and servers you run yourself

Pipecat

Open-source framework for voice and multimodal AI agents, handling the conversation logic rather than the transport underneath.

mediasoup

A server-side WebRTC library for engineers who want to build their own conferencing back end from parts.

Ant Media Server

Streaming server speaking WebRTC, HLS, RTMP and more, which you can install on your own machines or start from a cloud provider's catalogue.

Video calls and voice agents are different jobs

A video call product cares about many people seeing each other: layouts, screen share, recording, moderation. A voice agent cares about one person talking to software with as little delay as possible: getting audio to a speech model fast, handling interruptions, and passing the reply back before the silence feels awkward.

That's why one tool on this list is a framework rather than a transport. It sits on top of a transport and organises the chain of speech recognition, language model and speech synthesis. You still need something to carry the audio. Three of the hosted vendors here mention AI or voice agents on their own sites, so you can buy both layers from one place or mix them.

If you're only adding calls to an app, ignore the agent side entirely and pick on SDK quality for the platforms your users are on. The walkthrough on how to add video calls to your app covers the build steps.

Work with me

Want AI doing the heavy lifting in your marketing?

I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.

How the minutes add up

The hosted services mostly meter by the minute, and the detail that changes your bill is who gets counted. Participant minutes count every person in the room, so a one-hour call with four people is four hours of billable time. Some vendors price audio well below video, which matters for voice agents, since an agent call is usually audio only between one person and a bot.

For scale, four of the hosted vendors publish a per-minute starting rate on their own pricing pages this month, and those rates run from about a tenth of a cent to about ten cents a minute, with audio at the cheap end. Three of those four also show free minutes to start, from a few hundred up to ten thousand. Run your expected call length, room size and monthly sessions through each page before you trust any headline rate.

Self-run tools move the cost to servers. One server on this list is metered per instance-hour, which rewards keeping rooms dense on fewer machines; the open-source projects cost nothing to license and everything to operate.

You can get a working prototype running on most of the list before a bill arrives. Four of the nine have a free plan or free minutes, two have a free trial or starter credit, and the two open-source projects are free to download; only one shows no pricing on the pages read for this list.

Do you make real-time voice, video or voice AI platforms? Tool makers can ask to be considered for a featured place on this list.

A trial period that tells you more than the docs

Pick two candidates and put the same small feature on each: a two-person call with a mute button and a recording, or a voice agent that answers one question. Ship it to five colleagues on different networks, including one on mobile data and one behind a strict office firewall.

Over the week, watch four things. Join time on a bad connection. How the SDK behaves when someone's network drops for ten seconds and comes back. Whether the dashboard shows per-session quality data you can act on. And the bill for that week multiplied out to your real expected usage, counting participants the way the vendor does. The firewall test alone knocks out more options than any feature grid.

Which way to go

An app adding one-to-one or small group calls: a hosted video SDK with good mobile support. You'll be live in days and won't need anyone who knows what a TURN server is.

A voice AI product: an agent framework on top of a hosted transport, priced on audio minutes rather than video. Check interruption handling in your trial, because that's what users notice first.

Large audiences watching a few speakers: a streaming server that can hand WebRTC off to HLS for the crowd. If you're weighing simpler hosted rooms instead, see the video calling API comparison.

High, steady volume or strict data rules: self-run, with a hosted platform kept as a fallback for regions you don't cover.

Want your tool on this list? Featured spots and reviews here are paid partnerships. Here's how to partner with me.

Frequently asked questions

Can I switch from LiveKit without rewriting my front end?

Not entirely. Each platform ships its own client SDK, so the code that joins rooms, publishes tracks and handles events changes. Keeping that code behind your own small interface means the rest of your interface survives the move.

Do I need a separate provider for speech recognition and voices?

Usually yes. Transport platforms carry the audio and agent frameworks organise the conversation, but the speech-to-text, language model and voice output typically come from model providers you choose and pay separately.

Is WebRTC enough for streaming to thousands of viewers?

It can be, but it gets expensive and complex at that scale. Many teams keep WebRTC for the people speaking and send the audience an HLS stream, accepting a few seconds of delay for much cheaper delivery.

Published and maintained by the Lilach Bullock team, covering marketing, AI and business growth.
Your buyers are asking AI who to use. Does it say you?

See for free whether ChatGPT, Claude, Perplexity, Gemini and Google name you, and get the plan to become the answer.

Check my AI visibility →
Sundays only

Get the Sunday newsletter.

One email a week. AI experiments, marketing tactics, and the workflows Lilach is building right now in her own business.

Subscribe free

Let’s get your marketing running on AI.

Book a free 30-minute call

We figure out what you need, where AI fits in, and what working together would look like.

Book the call →

Or take the 30-second calculator

You’ll see the hours and the money quietly leaking out of your week, and the three workflows worth building first.

Take the calculator →

Or grab the free AI resource library

Prompt packs, templates, checklists, and swipe files. The exact tools I build for paying clients. Yours, free.

Get the library →
Keep reading

More from the blog.