I wanted to host completely livekit on premise wit...
# ask-ai
b
I wanted to host completely livekit on premise with egress also. What would be the cost for 250 concurrencies. include livekit oss + egress + worker hosting
t
Thanks for reaching out, Yethu. I'm an AI assistant and I'm here to help. Give me a moment while I research this topic.
Thanks — I’m glad you asked. I’ll walk through what’s included and what we need to estimate so you can get a realistic cost for running LiveKit fully on‑premise for 250 concurrent users (LiveKit OSS + Egress + Worker/Agent hosting). Short answer • LiveKit Open Source, Egress, and Agents/workers have no licensing fee — software cost is $0. • The total cost will be your infrastructure, bandwidth, storage, and operations (people) costs. To give a concrete estimate I need a few details about how you plan to use it. A few important cost drivers (please confirm or pick options) • Session topology ◦ Do you expect mostly 1:1 calls, many small group rooms (e.g., 5–10 participants), or large group rooms (e.g., 50+ participants / a single meeting with many concurrent streams)? • Media profile ◦ Typical video resolution and frame rate (e.g., 720p@30fps, 480p, audio-only). ◦ Will you use simulcast/VP8 SVC or single high‑res streams? • Egress usage ◦ Will you record all sessions, only some sessions, or stream out (livestream) to external platforms? ◦ Typical recording format/length and whether you’ll use a Chrome compositor/FFmpeg for recording (these are CPU intensive). • Worker/agent workload ◦ What will your agents do? (Simple orchestration, heavy media processing, transcription/AI inference). ◦ Do you expect many short agent jobs or long-running agents per room? • Deployment environment ◦ Will you run on physical servers in a private data center, or in a private cloud (VMs) / co-lo? Which cloud provider and region if applicable? • Reliability & headroom ◦ Desired redundancy (single server vs. active-standby vs. multi-node cluster), and target headroom for bursts (e.g., 20–30% spare capacity). What I can estimate/provide next • If you give the answers above I’ll produce: a. A recommended architecture (number and size of LiveKit servers, egress workers, and agent hosts). b. Approximate CPU, RAM, and network requirements. c. Rough monthly cost ranges for common hosting options (example: private cloud VM-based cost or colo rack cost), plus bandwidth and storage estimates. d. Notes on scaling and redundancy options so you can plan for growth. If you want a quick example now (so you have something to start with), tell me which of the following best matches your expected usage and I’ll return a sample cost estimate: • A: Mostly 1:1 calls, low-res video (480p), few recordings. • B: Many small group calls (5–10 participants), 720p video, occasional recordings. • C: Large rooms and/or heavy recording/streaming (>=50 participants / many recordings / livestreaming). Which option matches you, or send the details for the bullet list above and I’ll prepare a tailored estimate. Sources: The end of Participant Minute pricing | The end of Participant Minute pricing | The end of Participant Minute pricing | Pricing | LiveKit | Build voice, video, and physical AI | LiveKit
b
what would be the pricing
t
Thanks — I’m responding about pricing for running LiveKit OSS plus Egress and Agents on‑premise for 250 concurrent users. I can’t give a single accurate dollar amount without a few usage details because the cost is driven entirely by your infrastructure, bandwidth, storage, and operational choices. To produce a concrete estimate, please answer the following (or pick the closest option): 1. Session topology (pick one) ◦ A: Mostly 1:1 calls ◦ B: Many small group rooms (5–10 participants) ◦ C: Large rooms (20–50+ participants) or many participants publishing simultaneously 1. Video/audio profile ◦ Typical video resolution and frame rate (e.g., 480p@15fps, 720p@30fps, 1080p@30fps) ◦ Audio only or video+audio? 1. Egress behavior ◦ Do you record all sessions, some sessions, or only on demand? ◦ Do you stream live to external platforms (RTMP) and/or generate per‑participant recordings? ◦ Typical recording length per session (minutes) 1. Worker/agent workload ◦ What do agents do? (light orchestration, FFmpeg/Chrome compositing for recordings, AI inference/transcription) ◦ Are agents long‑running per room or short tasks? 1. Deployment environment ◦ Physical servers in your DC, private cloud VMs, or public cloud (which provider/region)? ◦ Any budget or cost constraints (e.g., prefer lower-cost VMs vs. high-performance instances)? 1. Reliability & headroom ◦ Desired redundancy (single node, multi-node cluster, active-active) and % spare capacity for bursts. If you want a quick example estimate instead of answering all questions, pick A/B/C from #1 and choose a video profile from #2 and I’ll produce a sample monthly cost range (including suggested server counts/sizes, bandwidth, and storage) for on‑prem or private‑cloud hosting.