LiveKit Ship and Scale both work out to $0.01 per agent minute

LiveKit's paid self-serve plans work out to the overage rate: Ship at $50 includes 5,000 agent session minutes and Scale at $500 includes 50,000, which is $0.01 a minute each, the same as the $0.01 per-minute overage the page lists. The plan fee is therefore prepaid minutes plus concurrency and features, not a discount. Sume does not host live agent sessions; its text-to-speech is about $0.036 to $0.038 per minute of speech, against LiveKit's listed TTS range of $0.009 to $0.030 a minute.

Figures are from the LiveKit pricing page, read on 2026-10-03; the per-minute division is my arithmetic.

What the page lists

Each plan includes inference credits and a global edge network. Beyond the plan, the page prices telephony and model lines per minute.

The rate hidden in the plans

Divide each fee by its included minutes: $50 / 5,000 = $0.01 and $500 / 50,000 = $0.01. Build is free for its first 1,000 minutes, after which the overage rate applies. What the higher plans buy is concurrency (5, 20, then 600 sessions), more deployments and, on Scale, features such as role-based access and HIPAA compliance listed on the page.

So the choice between plans is about how many calls overlap, not about unit price. A product with a handful of simultaneous calls but 3,000 minutes a month would pay $50 on Ship, or $0 plus 2,000 overage minutes at $0.01 ($20) on Build, but Build allows only 5 concurrent sessions.

A live minute, added up

At the cheapest ends of the page's ranges, a minute is $0.01 (session) + $0.0025 (STT) + $0.0090 (TTS) + $0.0002 (LLM) = $0.0217. At the dearest ends of the text-model lines it is $0.01 + $0.0095 + $0.0300 + $0.0379 = $0.0874; the page also lists a realtime model at $0.0676 a minute, which would be higher. Telephony adds $0.01 a minute for US local inbound. Those sums are mine; the page does not publish a combined figure.

What Sume does and does not do

Sume text to speech is $0.0475 per 1,000 characters, up to 20,000 characters per request. Speech to text is $0.01 per minute of audio, up to 600 seconds per job.

Per-minute speech figures assume 750 to 800 characters a minute, the range Cartesia publishes for its voices; a script's real length varies with pace.

Sume has no agent runtime, no WebRTC room and no phone line. Both audio steps are jobs; see Jobs and results. Its place beside a LiveKit agent is the recorded material.

• Pre-render greetings and fixed notices as audio files. See pre-rendering fixed replies.
• Transcribe saved sessions at $0.01 a minute in jobs up to 10 minutes.
• Produce rendered avatar video; see LiveKit avatar providers versus rendered clips.