Tools / AI Avatar & Talking Head
Tavus
Real-time conversational AI humans and replicas, built API-first for developers
Disclosure: the Visit Website button may be an affiliate link — it supports the directory at no extra cost to you. Saving is private: it is stored only in your browser.
Facts checked · Nov 2026 — the pricing, free-plan and watermark facts on this page were checked against Tavus's official pages. See how we verify and the listing changelog.
Price alerts for Tavus
AI video tools change prices and credit rules without warning — Kling has raised prices roughly 3× before. Get emailed when Tavus's price, credits or free plan changes.
Our verdict
Best for developers and product teams who need a real-time AI human inside their own product — Tavus sells conversations by the minute, not a creator studio.
Tavus at a glance
| Starting price | Developer Starter plan from $59/month; free Basic tier available |
| Pricing type | Freemium |
| Free plan | Yes |
| API available | ✓ |
| Category | AI Avatar & Talking Head |
| Entry price estimate (our calculator) | $59/month, minutes-based |
| How the billing works | Developer Starter plan; free Basic tier also exists. Published overage: $0.37/min conversational on free tier, $0.32/min on Starter; video generation $1/min free tier, $0.90/min Starter |
What Tavus really costs
Tavus publishes its overage rates, so the unit economics are visible instead of guessed. Conversational video beyond your included minutes costs $0.37 per minute on the free Basic tier and $0.32 per minute on Starter ($59/month, which also includes 100 minutes). Pre-recorded video generation overage is $1.00 per minute on Basic and $0.90 per minute on Starter. Growth ($397/month) switches to scaling discounts instead of a listed rate.
Read those numbers as the price of a live AI human's time. A 10-minute support-style conversation that runs past your included allowance costs about $3.70 in overage on the free tier's rate; a product holding hundreds of such conversations a month lives or dies on the included-minutes ladder, which is why the jump from Starter's 100 minutes to Growth's 1,250 minutes is the real pricing story.
$0.37
Per extra conversational minute, pay-as-you-go on the free Basic tier
$0.32
Per extra conversational minute on Starter ($59/month, 100 minutes included)
$1.00
Per extra generated-video minute on Basic ($0.90 on Starter)
25 min
Free conversational minutes every month on the Basic developer tier
Rates are from Tavus's official pricing comparison table, checked Nov 2026. Overage is rounded to the nearest 6 seconds, and each conversation has a 30-second minimum charge. PALs consumer accounts are priced separately (Free, $20 and $50/month).
What is Tavus?
Most avatar tools render a video file: you paste a script, wait, and download a presenter reading it. Tavus is built around a different job — a live, two-way video conversation with an AI human (a PAL, in Tavus's current language) who sees you, hears you, remembers earlier conversations and responds in real time. Under the hood that is the Conversational Video Interface: a pipeline of perception, turn-taking, speech and full-face generation models, delivered over WebRTC through Tavus's APIs.
Alongside live conversation, Tavus trains replicas — hyperrealistic copies of a real person, face and voice — from a short source video. A replica can hold live conversations or star in pre-recorded videos generated through the video-generation API. There are two account families, and that matters for pricing: Developer accounts (the API product reviewed here) and PALs accounts, a consumer companion product with its own Free, Plus ($20/month) and Max ($50/month) tiers. If you are evaluating Tavus as a video tool, the developer tiers are the relevant ones.
How Tavus works
- 1Create a free developer (Basic) account — you get 25 minutes of conversational video and 5 minutes of video generation a month, plus access to 25 stock replicas.
- 2Pick a stock replica for testing, or train a custom replica from a short video of your subject (custom trainings are a paid-plan quota: 3 a month on Starter, 7 on Growth).
- 3Give the conversation context: upload a knowledge base (documents or websites), set objectives and guardrails, and connect function calling if the replica should act, not just talk.
- 4Embed the conversation in your product through the API or the no-code portal, and test turn-taking and latency with real users before scaling.
- 5Watch the minute meter. Every conversation carries a 30-second minimum charge, and usage beyond your included minutes bills per minute at your plan's overage rate.
- 6Scale on Starter ($59/month) for early production or Growth ($397/month) when you need 1,250 included minutes, 100+ stock replicas and 10 concurrent streams.
Core features
Conversational Video Interface (CVI)
The full real-time pipeline — vision, turn-taking, speech and rendering — so a replica can hold a natural face-to-face conversation at 1080p over WebRTC, rather than playing back a rendered file.
Custom replicas
Train a hyperrealistic replica of a real person, including a custom voice model. Starter includes 3 trainings a month and Growth 7; extra trainings are a flat per-replica fee ($65 on Starter, $40 on Growth, per the pricing page).
Video generation
The same replicas can star in pre-recorded generated videos. Included allowances are small next to conversational minutes — 5 minutes a month on free, 10 on Starter, 100 on Growth — with per-minute overage after that.
Perception, memory and guardrails
Replicas can read visual cues, keep persistent memory across conversations, and work to objectives with guardrails. A knowledge base can be built from files or websites so answers stay grounded in your material.
Function calling and your own LLM
Conversations can call external tools and APIs mid-dialogue, and you can bring your own language model instead of Tavus's default — the pieces that turn a talking face into a working agent.
Stock replica library
25 stock replicas on the free tier and 100+ on Growth and above, so you can prototype a conversational experience before training anyone custom.
Languages and compliance at the top end
30+ languages across plans. SOC 2 and HIPAA compliance reports, white-labelling and guaranteed SLAs sit on Enterprise, which is also where concurrency becomes custom.
The free plan — the full truth
The free developer tier is a genuine prototype budget, which is rare in the avatar category. You get 25 minutes of conversational video and 5 minutes of video generation every month, access to 25 stock replicas, the full API suite and a no-code portal. For a developer deciding whether real-time AI humans fit their product, that is enough to build and demo a real integration, not just watch a sample.
What the free tier does not settle is polish and scale: one concurrent stream, best-effort support, and no included custom replica trainings — training your own person is a paid-plan quota. Tavus's pricing comparison lists watermark-free generation among its paid-plan features, and the free tier's own watermark terms are not spelled out on the page, so do not assume free generated videos are clean until you have exported one and checked. Conversational minutes beyond your allowance bill at $0.37 per minute, so an unmonitored public demo can quietly become a bill.
- • 25 conversational minutes and 5 video-generation minutes per month on the free developer tier
- • 1 concurrent stream on the free tier (3 on Starter, 10 on Growth)
- • No included custom replica trainings — those start on Starter (3/month)
- • Each conversation carries a 30-second minimum charge
- • Watermark terms for free generated video are not stated plainly on the pricing page — verify on your first export
Tavus pricing plans
| Plan | Price | Credits included | Best for |
|---|---|---|---|
| Basic (Developer) | $0/month | 25 conversational minutes + 5 video-generation minutes/month; 25 stock replicas; 1 concurrent stream | Prototyping a conversational AI human and proving the integration works |
| Starter | $59/month + pay-as-you-go | 100 conversational minutes + 10 video-generation minutes/month; 3 custom replica trainings/month; 3 concurrent streams | Startups putting a first live replica experience into production |
| Growth | $397/month + pay-as-you-go | 1,250 conversational minutes + 100 video-generation minutes/month; 7 replica trainings/month; 100+ stock replicas; 10 concurrent streams | Teams productionising conversations at real volume |
| Enterprise | Custom | Volume discounts, custom concurrency, 100% white-labelled experience, SLAs and compliance reports | High-volume or regulated deployments |
Developer plans are from Tavus's official pricing page, checked Nov 2026. Tavus also sells separate PALs consumer accounts (Free with 15 call minutes, Plus $20/month with 150 minutes, Max $50/month with 500 minutes) — those are a companion product, not the developer API tiers above.
Pros
- + Real-time, two-way AI human conversation — a capability most avatar tools do not offer at all
- + Free tier is a usable prototype budget with full API access, not a crippled demo
- + Overage rates are published per minute, so costs can be modelled before launch
- + Replicas combine face, voice, memory, knowledge base and function calling in one system
- + Clear scaling ladder from free prototype to enterprise SLAs and compliance
Cons
- − The $59 to $397 step between Starter and Growth is a big commitment for a young product
- − Included video-generation minutes are small — this is a conversation platform first
- − Two account families (Developer vs PALs) make the pricing page harder to read than rivals
- − Per-minute billing punishes unmonitored public demos; the 30-second minimum charge adds up on short calls
Best for
- ✓ Developers embedding live AI-human conversations in a product
- ✓ Teams building AI receptionists, coaches, tutors or interviewers that must respond, not recite
- ✓ Products that need a specific real person replicated, with consent, at conversational quality
- ✓ Buyers who want published per-minute rates instead of opaque credits
Not ideal for
- ✕ Creators who want to paste a script and download a presenter video — HeyGen or Synthesia are simpler and cheaper for that
- ✕ Anyone needing high volumes of pre-recorded avatar video on a small budget
- ✕ Non-technical users expecting a polished no-code studio as the main experience
Tavus alternatives
See all alternatives →If your real job is script-to-presenter video rather than live conversation, HeyGen is the natural alternative: a creator-friendly studio from $24/month billed annually, with Tavus-style live avatars as a side feature rather than the core. Synthesia suits corporate training with stronger admin and compliance tooling. For the cheapest talking-head work and a mature API, D-ID starts around $5–6/month. Tavus earns its premium only when the human on screen has to listen and respond in real time — if nobody is talking back, you are paying for the wrong thing.
HeyGen
HeyGen turns a script into a video of a realistic digital presenter, with a large avatar library, voice cloning and translation features. It is widely used for marketing videos, product explainers and social content where a human-like presenter is needed without filming.
Synthesia
Synthesia is an enterprise-focused avatar video platform used heavily for training and internal communication. It offers a large stock avatar library, personal avatars, multi-language narration and compliance and admin features that larger organisations require.
D-ID animates a still photo or avatar into a speaking presenter and is well known for its API, which businesses use to generate personalised talking videos at scale. It also offers a studio for individual creators making presenter content cheaply.
Frequently Asked Questions
Is Tavus free?
The developer Basic tier is free and includes 25 minutes of conversational video and 5 minutes of video generation per month, 25 stock replicas and full API access. It is a prototype budget, not a production one: one concurrent stream and no included custom replica trainings.
How much does Tavus cost per minute?
Beyond included minutes, conversational video costs $0.37/minute on the free tier and $0.32/minute on Starter, per Tavus's pricing table (checked Nov 2026). Video generation overage is $1.00/minute on Basic and $0.90/minute on Starter. Growth and Enterprise use scaling discounts.
What is the difference between Tavus developer and PALs accounts?
Developer accounts are the API product: replicas, the Conversational Video Interface and per-minute billing, from free to $397/month Growth. PALs accounts are a consumer companion product — your own AI human to talk to — at Free, $20/month (Plus) and $50/month (Max).
Can Tavus clone a real person?
Yes. A custom replica — face plus a voice model — can be trained from a short video. Trainings are a paid quota (3/month on Starter, 7/month on Growth), with extra trainings billed per replica. Only replicate people with their consent; Enterprise adds a white-labelled consent workflow.
Does Tavus have an API?
Yes — the API is the product. Every developer plan, including the free tier, includes full access to the developer APIs plus a no-code portal, per the pricing page.
Tavus or HeyGen for live avatars?
Tavus is purpose-built for live, real-time conversation with published per-minute rates and conversation features such as memory, perception and function calling. HeyGen is primarily a video-generation studio that also offers interactive avatars. Choose Tavus when conversation is the product; choose HeyGen when rendered videos are.