Media between people
Media routed between participants: each producer × the people consuming it. Nobody consumes their own stream.
- Audio
- $0.10/1K min
- SD video
- $0.75/1K min
- HD video
- $1.50/1K min
Real-time media, telephony, and AI agent infrastructure with transparent usage rates. Bring your preferred providers and keep control of your margins.
Either way it is one balance: a plan credits your account with minutes, metered use deducts from the same minutes. Plan minutes are calling credits scaled by destination, not clock minutes — 1,000 buys roughly 100,000 talk minutes to the US, 10,000 to India.
Cloud phone, meetings, AI agents, widgets or your own app on the SDKs — it all runs on the same infrastructure and bills the same underlying minutes.
A plan just bundles the same minutes up front, sized to how often you run rooms and calls. Every path — plan, pay-as-you-go or SDK — pays the same rates underneath.
Start free, connect your own SIP, or add larger meetings, recording, phone, and translation.
Try small rooms and 1-on-1 calls
Use your own SIP at the lowest yearly cost
Regular team meetings without credits
Meetings plus phone and translation credits
Large rooms, events, and higher usage
How plan minutes work
Plan minutes are calling credits, not clock minutes. Every destination has its own per-minute calling rate, so the same credits stretch much further where calling is cheap. For example, 1,000 plan minutes buy roughly:
You're only billed for what you actually use — at each destination's calling rate. Nothing is charged for minutes you don't spend.
Which Plan is Right for You?
For Small Rooms or AI Agents
Choose Starter or Pro — great for small capacity rooms. Perfect for telephony and AI agents with low concurrent usage.
For Large Events or Webinars
Choose Pro or Scale — perfect for large capacity rooms and events. Ideal for webinars, town halls, and large meetings with many concurrent participants.
| Feature | Free | Connect | Starter | Pro | Scale |
|---|---|---|---|---|---|
| Max participants | 4 | 4 | 100 | 250 | 3,000 |
| Meeting duration | 15 min | 15 min | 24 hours | 24 hours | 24 hours |
| Video quality | SD (480p) | SD (480p) | HD (720p) | HD+ (1080p) | HD+ (1080p) |
| Cloud recording | - | - | 1 hour per session | 3 hours per session | 10 hours per session |
| Phone / SIP access | No | Own SIP only | Yes | Yes | Yes |
| Phone call credits | - | - | - | ~100 min/mo | ~1000 min/mo |
| AI / translation | Own config | Own config | Included | ~100 min/mo | ~1000 min/mo |
| Priority support | - | - | - | Yes | Yes |
Building an app with video/voice features?
Every call, meeting, agent and recording reduces to streams moving for minutes. Where a stream goes decides which meter it runs on.
Media routed between participants: each producer × the people consuming it. Nobody consumes their own stream.
Any stream that ends at, or is processed on, our servers: recording, an AI agent listening, noise cancellation, transcription feeds.
One caller, one agent, 10 minutes of audio
Your STT, LLM and TTS providers bill you directly.
Two people, HD camera both ways, 30 minutes
Nothing reaches the server as a sink, so only meeting minutes apply.
5 people on SD video, everyone on camera, 15 minutes
A camera that is off is not a producer, so it adds nothing.
Two hosts, audio, 60 minutes, each track recorded
The first composed recording is included.
Each billing interval records who was sending, how many people received each stream, at what quality, the rate that applied and the pricing model version. Open any room, see every line, and check the total yourself.
Start with live participant usage or recording output. The estimate updates as you change room size, media quality, duration, frequency, layout, and delivery choices.
Model the live media a room receives and sends
2 streams come in.
Each is delivered to the other 29 people; nobody receives their own.
2 × 29 = 58 deliveries × 60 min × 6 events at $0.0001 per audio stream-minute.
Quality is set per stream, not by room size: a 1080p presenter and 360p cameras bill as exactly that, and up to 3,000 people can send at once with no per-room seat tier.
Infra-first rates for calls, multimodal sessions, and real-time agents. Bring your own AI providers and keep the margin.
Carrier numbers and AI-model usage stay with your chosen providers. MediaSFU charges the real-time transport, agent runtime when enabled, and recording output only when you create it.
| Service | Media only | AI-ready total | Discounted rates |
|---|---|---|---|
| Phone Call (Telephony) Bring your own number. | $0.0001 | $0.0021 Media + agent infrastructure | Medium$0.00168Media only $0.00008 High$0.00126Media only $0.00006 |
| Web/Mobile (Audio Only) Non-telephony audio calls or agents. | $0.0001 | $0.0021 Media + agent infrastructure | Medium$0.00168Media only $0.00008 High$0.00126Media only $0.00006 |
| Web/Mobile (Multimodal) Non-telephony video or multimodal calls or agents, from QnHD (320×180). SD and up: see the quality matrix. | $0.000375 | $0.003375 Media + vision agent infrastructure | Medium$0.0027Media only $0.0003 High$0.002025Media only $0.000225 |
Eligible MediaSFU usage is measured monthly. The medium and high tiers apply automatically when the threshold is reached; carrier, AI-provider, and optional service charges remain separate.
Connect your preferred STT, LLM, and TTS providers. MediaSFU handles the real-time media and orchestration layer without bundling AI usage into a higher platform rate.
Reference rates link to each provider. Included services and provider costs differ, so use the calculator below with your own workload.
Every stream bills at the resolution it actually runs at. A 1080p speaker with three 360p viewers is one FHD minute plus three SD minutes — nothing is pooled, and nobody is downgraded to keep the bill in a tier.
Some providers price by the aggregate or composite size of everything a participant receives, so adding people pushes every minute into a higher tier — or you cap everyone's quality to stay under it. Others charge one flat video rate, 180p or 1080p.
| Resolution | MediaSFU | 100ms | Daily | Agora |
|---|---|---|---|---|
| QHD (2560×1440) | $3.00 | — | — | — |
| FHD (1920×1080) | $2.25 | $4.00 | $4.00 | $8.99 |
| HD (1280×720) | $1.50 | $4.00 | $4.00 | $3.99 |
| SD (640×360) | $0.75 | $4.00 | $4.00 | $3.99 |
| QnHD (320×180) | $0.375 | $4.00 | $4.00 | $3.99 |
| Audio only | $0.10 | Not published | $0.99 | $0.99 |
| Resolution | MediaSFU | 100ms | Daily | Agora |
|---|---|---|---|---|
| QHD (2560×1440) | $18 | — | — | — |
| FHD (1920×1080) | $15 | $27 | $13.50 | $28 |
| HD (1280×720) | $12 | $27 | $13.50 | $14 |
| SD (640×360) | $6 | $13.50 | $13.50 | $14 |
| QnHD (320×180) | $3 | $13.50 | $13.50 | $14 |
| Audio only | $2 | $4 | — | — |
MediaSFU rates are per participant stream, before volume discounts. Other providers' rates are from their public price pages; — means not offered.
Pick a workload, set your minutes, and watch the gap.
Agent infrastructure - bring your own STT, LLM, and TTS
At 50,000 monthly minutes, the estimated annual difference versus the closest listed alternative is $4,800. Competitor figures use their public list rates; services they bundle differ.
MediaSFU captures participant media before composition, preserving useful inputs for transcripts, review, compliance, language workflows, and new layouts after the session.
Audio, video, screen share, and whiteboard are retained as separate tracks.
Original source quality remains available for analysis and post-session processing.
Generate a composed recording, additional layouts, or HLS/VOD delivery.
Build for web, mobile, and game surfaces with framework-specific packages and guides.
Deploy real-time experiences near users across six continents.
Use E2EE, token authorization, domain controls, IP filtering, and webhooks.
Just need video meetings for your team?
Audio transport is $0.0001 per minute ($0.10 per 1,000 minutes), video transport starts at $0.000375 per minute at QnHD (320×180) and is $0.00075 at SD (640×360), and AI agent infrastructure is $0.002 per minute ($2 per 1,000 minutes). Volume discounts reach up to 40%.
Meeting minutes are media routed between people: one participant's stream delivered to another, billed at the streaming rates (audio $0.10 per 1,000 minutes, HD video $1.50). Recording minutes are any stream that ends at, or is processed on, MediaSFU servers: a recording, an AI agent listening to a caller, noise cancellation, a transcription feed. Those bill at the recording rates (audio $2 per 1,000 minutes, HD video $12).
Routing is counted as each media producer × the people consuming it, per minute. Nobody consumes their own stream, so a 1:1 video call is 2 × 1, four people with one camera on is 1 × 3, and a five-person meeting with every camera on is 5 × 4. A camera that is off is not a producer, so it is not billed. An open room where nobody is sending media keeps a small presence rate of a quarter of a stream-minute per person, which covers its live connection, chat and signalling. Larger rooms also get automatic room-size tiers.
At list rates, before room-size and volume discounts: a 10-minute AI voice agent call is about $0.021, a 30-minute 1:1 HD video consult about $0.09, a 15-minute five-person SD standup about $0.23, and a recorded one-hour two-host audio podcast about $0.25. The worked examples on this page show each calculation.
No — and that is the point. MediaSFU charges only for the real-time infrastructure and orchestration layer. You bring your own STT, LLM, and TTS providers (BYOK) and pay them directly with no markup, so you keep control of quality and margin.
Developers get 50,000 free build minutes: 20,000 production and 30,000 sandbox. On top of that you get 100,000 demo minutes, which apply to rooms you create from the dashboard and are separate from your SDK and API usage. Teams can use the Free plan for small rooms and 1-on-1 calls with no time-boxed trial.
If you are running meetings and calls for a team, pick a Teams plan (from $5.99/month, Connect from $10/year with your own SIP). If you are building an app, agent, or embedding communication into a product, use the usage-based developer rates.
On published rates, MediaSFU agent infrastructure at $0.002/min is roughly 25x below Vapi's ~$0.05/min platform fee, and well below typical bundled rates from Retell (~$0.055/min) or Bland (~$0.11/min). Bundled platforms include different services, so model your own volume and provider mix.
No. MediaSFU does not add bandwidth markup, and supported AI and SIP provider costs stay direct. Recording is track-based with the first composed version included free.
Launch calls, meetings, agents, translation, widgets, and recording without stitching together separate operating surfaces.
Real-time agents for calls, video sessions, intake flows, lead qualification, and 24/7 customer support.
AI + mediaAI-powered outbound calling with automated scheduling, smart retries, and real-time analytics.
Campaign operationsUnified contact management with call history, notes, tags, and CRM integration.
Contact workspaceJoin and place calls via phone with local numbers worldwide. Available on all plans.
TelephonyReal-time translation during meetings and calls. Break language barriers effortlessly.
LanguageProfessional virtual backgrounds and blur effects for polished video calls.
VideoDrop-in video, voice, and chat widgets you can embed in any website or app.
EmbedRecord meetings and calls to the cloud with automatic storage and easy playback.
Capture