Convai
Convai is a conversational AI platform for building 3D characters and NPCs that see, speak and act inside games, XR and virtual worlds, using no-code studios, Unity and Unreal plugins, and open APIs.
What is Convai?
Convai, pronounced convey, is an infrastructure layer that gives three-dimensional characters a voice, a memory and a sense of their surroundings. The premise is stated plainly on the homepage: humans understand the world through their bodies, so an artificial intelligence needs to be embodied and spatially aware to reach a comparable depth of understanding. In practice that means virtual humans and disembodied AI characters that hear and see what is around them, then answer in natural language with speech, gestures and contextually appropriate actions. The workflow is organised in three movements: craft a character's mind, embody that mind in an avatar, then deploy the avatar inside a custom environment. Two doors lead in. Creators use no-code tools that run in the browser, with Avatar Studio handling appearance, clothing, environment, lighting, camera work and animation, and Convai Sim building larger simulation scenes. Developers use plugins for Unreal Engine, Unity and web runtimes such as Three.js, React, PlayCanvas and Babylon.js, alongside open APIs, Pixel Streaming embedding, a modding framework and MCP servers. Underneath sits a single real-time pipeline: speech recognition, a language model, speech synthesis, NeuroSync lip sync, memory, vision and what Convai calls Action-to-Dialogue, with a first-token latency the company advertises below 200 milliseconds. Characters can be grounded in a multimodal Knowledge Bank assembled from uploaded documents, follow a scripted narrative flow, see through a webcam or a first-person view, and chain animations into complex actions. Speech covers more than 65 languages, one to four per character, with voices sourced from Google Cloud, Microsoft Azure, OpenAI and ElevenLabs. Three uses are put forward: role-play characters for learning and training, a brand representative answering product and company questions around the clock, and companions that improve onboarding and retention in social virtual worlds. Deployment targets run from websites and mobile applications to video games, virtual and augmented reality and physical installations. The platform is avatar-agnostic, accepting anything from a cube to a highly detailed model, provided lip sync receives the blendshapes it expects. The publisher is Convai Technologies Inc, whose engineering team it describes as drawn from Google, Nvidia and Amazon.
What it does
- Craft a character's mind: role, objectives, backstory and personality
- Embody that mind in any 3D avatar and deploy it inside your own environment
- Ground the character's answers in a multimodal Knowledge Bank built from uploaded documents
- Give characters sight through a webcam or a first-person view
- Chain animations into complex, context-aware actions
- Script a narrative flow that keeps a conversation on story
- Speak one to four languages per character, drawn from more than 65 languages and 1,200 voices
When to use Convai / When not to
A quick filter to help you decide if Convai is the right fit.
When to use Convai
- Game developers building conversational NPCs in Unity or Unreal Engine
- XR studios producing immersive role-play training simulations
- Brand and marketing teams deploying an AI brand representative available around the clock
- No-code creators designing 3D characters in Avatar Studio or Convai Sim
- Indie developers prototyping on the permanent free tier of 100 monthly interactions
When not to use Convai
- Teams that want a plain text website chatbot rather than an embodied 3D character
- Audio producers looking for a dubbing or voice-over studio
- Buyers in regulated sectors who need a data processing agreement, a subprocessor list and a published postal address
- Organisations that require European data residency, since hosting is in the United States only
- Anyone who must budget before signing up, because no paid plan amount is published on the site
How to use Convai
A typical end-to-end flow, from setup to results.
- Create a Convai account with an email address and password, or with Google or GitHub where that option is available in your region
- Open the Convai Playground, the dashboard where characters are built and tested
- Start from one of the sample characters supplied in the dashboard, or from a blank one
- Write the character's backstory, personality and objectives, then adjust the Core AI Settings and the moderation filters
- Upload documents into the Knowledge Bank so the character answers from your own material
- Design the narrative flow and define the actions the character is allowed to perform
- Choose one to four languages, English being the default, then pick a voice from the list filtered by that language
- Give the character a body in Avatar Studio, or build a scene in Convai Sim, without writing any code
- To ship inside an application, install the Unity plugin, the Unreal Engine plugin or a web plugin, or call the Core APIs directly
- Publish, bearing in mind the contractual duty to display the Convai logo and a Powered by Convai line at start-up unless waived in writing
Pros & Cons
Pros
- A complete real-time pipeline in a single service: speech recognition, language model, speech synthesis, lip sync, memory, vision and actions
- Deep and documented engine coverage, from Unity and Unreal Engine to web plugins and Pixel Streaming
- Avatar-agnostic and platform-agnostic, from MetaHuman and Reallusion to Meta Quest and Apple Vision Pro
- Two genuine entry points: no-code studios for creators, APIs and SDKs for developers
- A permanent free tier of 100 interactions per month, enough to build a proof of concept
- Contractual recognition that the customer owns the content its characters produce
- An Enterprise tier that collects nothing by default and offers on-premise deployment
Cons
- No paid price is readable anywhere: the pricing page renders in JavaScript and publishes no amount in its HTML
- No postal address appears on any page of the site
- A single mailbox handles support, legal matters, privacy requests and service of process
- No data processing agreement, no subprocessor list and no EU representative under Article 27
- The privacy policy dates from February 2023 while the terms were revised in September 2025
- On the Creator tier, refusing the licence granted over your content means leaving the service
- Parts of the FAQ are out of date, describing shipped features as still in progress
Pricing & Plans
Convai operates a freemium model. A permanent free tier is available and is documented at 100 interactions per month, which the company presents as sufficient for most developers to test their characters and build a proof of concept. Paid subscription tiers exist and are billed in United States dollars, payable on issue of the invoice and non-refundable unless otherwise agreed. However, no amount for any paid plan is published in a verifiable first-party source: the pricing page is rendered on the client side and carries no figure in its HTML, and neither the FAQ nor the company's own articles state one. No entry price or currency can therefore be reported here. Billing is metered in interactions, an interaction being a user input followed by a character response in text or voice, with monthly quotas that reset at each billing cycle and do not carry over.
- 100 interactions per month
- 3
- 000 interactions per month
- of which at most 1
- 500 on flagship models such as GPT-4o
- Claude 3.5 Sonnet or Gemini 1.5 Pro
- plus 500 interactions using high-quality third-party voices
- price not published
- higher quotas
- with a cap on monthly active end users illustrated at 200 in the documentation
- price not published
- voice cloning
- no data collection by default
- on-premise deployment and a dedicated agreement
- price on request
- A business plan is also referred to in Convai's own article on bringing your own language model
- with no published terms
Data, GDPR & hosting
A consolidated view of how Convai handles your data.
GDPR overview
The term GDPR appears nowhere on the pages reviewed: Convai neither claims nor denies compliance. The privacy policy does carry a section for individuals in the EEA, the United Kingdom and Switzerland setting out lawful bases, namely contract, legitimate interest, legal obligation and withdrawable consent, along with rights of access, portability, rectification, erasure, objection and restriction, with complaint links to the EDPB, the ICO and the Swiss authority. Beyond that the framework is thin. No representative is designated under Article 27, no data protection officer is named, no data processing agreement is offered and no subprocessor list is published; every request goes to one support address. Personal data is transferred to servers in the United States with no mention of standard contractual clauses. Registration is barred below 13, and below 16 for EU residents. The policy is dated 21 February 2023.
Who owns the data?
Convai's terms state that the customer owns the Customer Content and the AI Generated Content at all times. Customer Content covers the inputs used to build a character, such as backstory, skills and personality, together with the scene data, metadata, voice and text a deployed character receives from end users. AI Generated Content covers the text, speech, animations and actions the character produces in return. Anything Convai itself creates in order to deliver the service remains Convai's property, and the customer holds only a non-exclusive, non-transferable licence to use it. Customers must have obtained from their own end users the rights covering any data those users supply.
Reuse rights
Ownership does not mean the material is left untouched. On the Creator tier the terms grant Convai a worldwide, royalty-free licence to store and use Customer Content and AI-generated content in order to develop and improve the service, unless something else has been agreed in writing; the only remedy offered to a customer who objects is to stop using the service or delete the account. On the Enterprise tier the position is reversed: Convai states it does not collect, store or use any data generated by those users unless they have given express and explicit consent, with processing governed by the applicable agreement and by the customer's own configured preferences. Separately, the privacy policy allows personal data to be used to run, secure, personalise and improve the service, and to be shared with unnamed hosting, cloud, analytics, email and marketing providers. Convai states that it does not and will not sell personal information. No mechanism allowing a customer to exclude their data from model training is documented anywhere on the site.
Data retention & training
Hosting summary
Convai's privacy policy states that personal information is transferred from the user's location to the company's facilities and servers in the United States. That is the only jurisdiction named. No European region, no data residency option and no named hosting provider appear anywhere on the site; the policy refers only to categories of vendors, including hosting, cloud and other information technology services. On-premise deployment is listed among the Enterprise-ready options, without further detail. Technically, the domain resolves to an anycast address operated by Google and the site's static assets are served from Google Cloud Storage, which describes how the website itself is delivered rather than where customer data lives. Security is described as commercially reasonable technical, administrative and organisational measures, with an explicit caveat that the service is used at the user's own risk and that no transmission is ever fully secure. Buyers who need European hosting, a data processing agreement or a named list of subprocessors will find none of them documented.
Things to keep in mind
Risks and trade-offs to weigh before adopting Convai.
- On the Creator tier, everything written into a character can be used to develop and improve the service, and the only way out is to leave the platform
- Because these characters sound human, users may forget they are talking to a statistical model and take what it says at face value
- Vision through a webcam turns a play session into a camera feed: end users need to be told and their consent obtained
- You remain responsible for your own end users, including the notices they are owed and the consents they must give
- Immersive companions can become genuinely absorbing, and neither the site nor the terms address time spent or emotional attachment
- The ISO 27001 claim on the homepage is spelt with a mistake and no certificate is published to support it
- Liability is capped at USD 10 or twelve months of fees, disputes go to arbitration in California, and class actions are waived
Setup & Integrations
Technical difficulty
Setup difficulty depends on the route taken. The no-code route is genuinely light: a Convai account is the only stated prerequisite, and Avatar Studio and Convai Sim run in the browser, with sample characters available to start from. The developer route asks for ordinary integration work, installing the Unity plugin, the Unreal Engine plugin or a web plugin, or calling the Core APIs directly. One concrete constraint applies to lip sync, which requires an avatar carrying OVR, ARKit or Reallusion blendshapes. Public documentation, video tutorials, a community forum and a status page support both routes.
Deployment
Integrations
Supported languages
Behind Convai
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement Convai.
Frequently asked questions
What is Convai?
Which engines and platforms does Convai support?
Is there a free plan?
How much do the paid plans cost?
How is an interaction counted?
Can I use my own avatar?
How many languages and voices are available?
Who owns the content my characters produce?
Where is my data hosted and how long is it kept?
What is the minimum age to use Convai?
Should you pick Convai?
Convai occupies a narrow and well-defended position: making three-dimensional characters that can hold a conversation, perceive a scene and act inside it. The technical case is convincing and, unusually, documented. Engine plugins for Unity and Unreal Engine, web runtimes, an avatar-agnostic approach that accommodates MetaHuman, Reallusion and Ready Player Me, deployment targets stretching from desktop to Meta Quest and Apple Vision Pro, and a full real-time pipeline rather than a set of parts to assemble. The dual entry point is real rather than cosmetic: a designer can build a character in the browser while an engineer wires that same character into a game through the API. A permanent free tier of one hundred interactions per month makes evaluation genuinely free. The reservations are commercial and legal rather than technical. The pricing page publishes nothing readable, so the cost of the first paid tier cannot be established from the site: plan names and quotas are known, amounts are not. The corporate footprint is thin for a business selling to studios and brands, with no postal address, one mailbox covering support, legal and privacy matters, a privacy policy two and a half years older than the terms, and no data processing agreement, subprocessor list or European representative. Data is hosted in the United States only. On the Creator tier, content put into a character is licensed to Convai to develop and improve the service, and declining that licence means leaving; the Enterprise tier reverses the default and adds on-premise options. For a studio or a creator building conversational characters for games, training or brand experiences, Convai deserves a hands-on trial on the free tier. For a buyer carrying procurement or data protection duties, the price, the contract and the hosting are the questions to settle before signing.
- Choosing a selection results in a full page refresh.
- Opens in a new window.