# How much does Tavus CVI cost per minute? Tavus pricing and the per\-hour math \(October 2026\) Tavus conversational video costs $0\.26 to $0\.35 per minute in overage, $15\.60 to $21 an hour\. Plan\-by\-plan math, the 30\-second minimum, and comparisons\. Published: 2026-10-05T10:09:17.069Z Updated: 2026-10-05T10:09:17.069Z Canonical: https://realtimeavatar.ai/blog/tavus-pricing-per-minute Markdown: [en](https://realtimeavatar.ai/blog/tavus-pricing-per-minute.md) As of October 2026, Tavus's Conversational Video Interface (CVI) costs $0.35 a minute on Builder, $0.31 on Growth and $0.26 on Business once a plan's included minutes are used up, which is $21.00, $18.60 and $15.60 per hour of live conversation. Included minutes work out to between $14.63 and $22.00 an hour depending on the plan, and every conversation is billed a 30-second minimum, rounded to the nearest 6 seconds. All Tavus figures here come from tavus.io/pricing as read on 5 October 2026. That page currently serves two different developer price tables in the same HTML (more on that at the end), so check it again before you budget. The Realtime Avatar figures come from our own pricing page; we make Realtime Avatar, so weigh our comparison accordingly. ## Tavus CVI plans and what an hour costs ``` Plan Monthly Included CVI minutes Cost per included hour Overage per min Overage per hour Max call Concurrent Free $0 20 free none none 5 min 1 Starter $22 60 $22.00 none none 5 min 1 Builder $59 175 $20.23 $0.35 $21.00 15 min 3 Growth $397 1,300 $18.32 $0.31 $18.60 60 min 10 Business $975 4,000 $14.63 $0.26 $15.60 60 min 15 Enterprise custom custom custom custom custom custom custom ``` Cost per included hour is the monthly fee divided by included hours, so it only holds if you use every included minute. Starter has no pay-as-you-go coverage: when the 60 minutes run out, calls stop until the next cycle. Builder, Growth and Business bill overage per additional minute. Every plan includes the full pipeline Tavus describes on the page (speech recognition, LLM, text-to-speech, WebRTC and its own rendering, perception and turn-taking models), so the per-minute rate is the whole conversation, not just the face. ### Things that add to the bill - The 30-second minimum per conversation. A user who opens a call and leaves after 8 seconds is billed 30. If your product has many short or abandoned calls, model this explicitly: 1,000 calls averaging 12 seconds are billed as 500 minutes, not 200. - Rounding to the nearest 6 seconds on each conversation. - Conversation recordings, listed at $0.03 a minute. - Custom face slots: 1 on Starter, 3 on Builder, 7 on Growth, 15 on Business. The older table on the same page lists a one-time fee per extra custom replica ($65 on its Starter, $40 on its Growth). - Concurrency. Builder allows 3 simultaneous sessions, Growth 10, Business 15. Peak concurrency, not monthly minutes, can force a plan upgrade. ## Worked examples Here is the cheapest Tavus plan for three usage levels, assuming the minutes are spread so concurrency limits are not the constraint: ``` Monthly usage Cheapest Tavus plan Tavus total Tavus per hour 10 hours (600 min) Builder: $59 + 425 x $0.35 $207.75 $20.78 50 hours (3,000 min) Growth: $397 + 1,700 x $0.31 $924.00 $18.48 100 hours (6,000 min) Business: $975 + 2,000 x $0.26 $1,495.00 $14.95 ``` Two observations. Growth only beats Builder once you pass roughly 1,140 minutes a month, and Business only beats Growth above roughly 3,165 minutes. And even at 100 hours a month the blended price is close to $15 an hour, because the included minutes are themselves priced near the overage rate. Three habits keep a Tavus bill close to the arithmetic above. Start a conversation only after an explicit user action, so page loads and accidental taps do not each cost 30 seconds. Set a maximum duration that matches the task rather than the plan ceiling, so a forgotten tab does not run to 60 minutes. And track abandoned calls separately from completed ones in your analytics, since short calls are where the minimum charge hurts most. The same habits apply to every vendor in this category, including us: most of them meter the whole session, whether anyone is talking or not. ## How that compares with other realtime avatar APIs Comparing at the same usage is more honest than comparing headline per-minute rates, because included minutes differ so much. The table below prices 100 hours a month on each vendor's cheapest self-serve plan for that volume, using each vendor's published rates as of October 2026. All four are full-stack options here (LiveAvatar in Full Mode, Anam in its default Turnkey setup). ``` 100 hours a month (6,000 live minutes), cheapest self-serve plan for each, overage rates as of October 2026 Vendor Plan and math Total Per hour Tavus CVI Business: $975 + 2,000 min x $0.26 $1,495.00 $14.95 HeyGen LiveAvatar Business, Full Mode: $475 + 5,990 credits x $0.09 $1,014.10 $10.14 Anam Growth: $299 + 4,000 min x $0.12 $779.00 $7.79 Realtime Avatar Studio: $119 + 3,000 min x $0.08 $359.00 $3.59 ``` Caveats that matter. LiveAvatar bills overage in $100 increments, so its real charge rounds up. Anam meters the entire session, idle time included, and caps a single call at 2 hours on Growth. Realtime Avatar caps a single call at 30 minutes and its Studio plan allows 20 concurrent calls. At 10 hours a month the gap is wider: Realtime Avatar's Developer plan is $24 for 600 included minutes, against $207.75 on Tavus Builder. ## What the price difference buys you A lower per-minute rate is not the same product, and Tavus includes things the cheaper options here do not. Tavus trains a replica of a real person from a two-minute video or a photo, runs a perception model that reacts to what the camera sees, joins Google Meet and Zoom calls, lets you plug in your own OpenAI-compatible LLM, ships a hosted knowledge base and memories, and advertises SOC 2 and HIPAA compliance. If those are requirements, compare Tavus with Tavus-class products, not with the cheapest row. Realtime Avatar is narrower. It creates a character from one portrait, runs speech recognition, a language model you choose per session from the OpenAI and Gemini models it hosts, a Fish Audio voice and the rendered video under one meter of seconds on air, and lets your tools run in your own page or server. It has no renderer mode, no bring-your-own voice, no meeting bots and no hosted knowledge base. Plans: Starter $9 a month for 120 minutes, Developer $24 for 600, Studio $119 for 3,000, then $0.095, $0.085 or $0.08 a minute. The free sandbox is 1,020 credits, granted once per account, which is 17 live minutes. Tavus's free tier is 20 minutes a month, with 5-minute calls. ## A quick calculator you can paste If you want to rerun the numbers for your own volume, this is the arithmetic used above. Update the constants when the pricing page changes. ``` // Monthly CVI cost on each Tavus plan for a month of calls. // Rates from tavus.io/pricing as read on 2026-10-05. const plans = [ { name: "Builder", fee: 59, included: 175, overage: 0.35 }, { name: "Growth", fee: 397, included: 1300, overage: 0.31 }, { name: "Business", fee: 975, included: 4000, overage: 0.26 }, ]; // Every conversation bills at least 30 seconds, rounded to the nearest 6. export function billedMinutes(callSeconds: number[]) { return callSeconds.reduce((sum, s) => sum + Math.max(30, Math.round(s / 6) * 6), 0) / 60; } // Cheapest plan first, priced on billed minutes, not on time actually talked. export function tavusMonthlyCost(callSeconds: number[]) { const minutes = billedMinutes(callSeconds); return plans .map((p) => ({ plan: p.name, usd: p.fee + Math.max(0, minutes - p.included) * p.overage })) .sort((a, b) => a.usd - b.usd); } ``` ## A note on the two Tavus price tables On 5 October 2026 the HTML served at tavus.io/pricing contained two developer pricing tables. The one used in this post lists Free, Starter ($22), Builder ($59), Growth ($397), Business ($975) and Enterprise, and references the Phoenix-4.5 and Sparrow-2 models. The other lists Basic (free, 25 minutes), Starter at $59 with 100 minutes and $0.37 a minute overage, and Growth at $397 with 1,250 minutes and $0.32 a minute, alongside the older Phoenix-4 and Sparrow-1 names. Which one you see may depend on an experiment or a migration in progress. Either way, the hourly range is roughly $15 to $22 at overage, and the 30-second minimum appears in both. - [Realtime Avatar pricing](https://realtimeavatar.ai/pricing) - [Realtime Avatar quickstart](https://realtimeavatar.ai/docs/quickstart) - [Renderer or full stack: comparing realtime avatar APIs](https://realtimeavatar.ai/blog/renderer-or-full-stack-realtime-avatar-apis-compared) - [Realtime Avatar as a Tavus alternative](https://realtimeavatar.ai/blog/tavus-alternative-realtime-avatar-griffin) - [Realtime avatar API pricing: worked examples](https://realtimeavatar.ai/blog/realtime-ai-avatar-api-pricing) - [Tavus pricing](https://www.tavus.io/pricing) - [Tavus CVI product page](https://www.tavus.io/cvi) - [Tavus LLM layer docs](https://docs.tavus.io/sections/conversational-video-interface/pal/llm) - [LiveAvatar credits and subscriptions](https://docs.liveavatar.com/docs/faq/credits) - [Anam pricing](https://anam.ai/pricing)