I've spent the last three years testing AI character platforms — from polished consumer companion apps to the developer tools powering them behind the scenes. When Convai started showing up in NVIDIA spotlights, Second Life integrations, and GamesBeat coverage, I spent four weeks actually putting it through its paces: the playground, the Unreal Engine plugin, and the character editor under real conditions.
Here's what I found: Convai is technically ahead of almost everything in this space. The knowledge bank, scene perception, and voice pipeline are genuinely impressive. But there's a latency problem that Convai's own developer forum has documented since January 2025, and as recently as March 2026, users were reporting 8 to 10 second delays before a character replied. For a platform built around real-time conversation, that's a significant gap between promise and reality.
Convai (pronounced "convey") is a San Jose, California-based developer platform for building AI-powered characters in games, virtual worlds, and XR training simulations. Founded in 2022 by Purnendu Mukherjee — formerly a Senior Deep Learning Software Engineer at NVIDIA who built Nvidia's first 3D avatar chatbot — the platform integrates with Unity, Unreal Engine, Roblox, PlayCanvas, and Second Life. Characters can hear voice input, perceive 3D environments, draw on custom knowledge banks, and respond with synchronized lip animation. In late 2025, Convai launched a beta Web SDK bringing voice-first character interactions directly into browsers, and added the ability for characters to be interrupted mid-response — a meaningful update for live gameplay.
After four weeks of testing, I'd say Convai is worth it — specifically for developers building interactive virtual experiences who have the technical tolerance to deal with an occasionally unstable backend.
Knowledge bank accuracy: Upload a PDF, text file, or game lore document and the character draws on it accurately during conversation. I tested this with a 3MB worldbuilding document and the character referenced it correctly — no hallucinated plot details that weren't in the source file.
Scene perception: The vision pipeline works. Characters can identify objects in the 3D scene and respond contextually — a capability that consumer companion apps don't come close to offering.
Game engine depth: The Unreal Engine plugin (v4.0 beta as of early 2026) includes emotion signals, action callbacks, and lip-sync as standard. Setting up a fully voiced, reactive NPC from scratch took under two hours.
Latency: Developers in Convai's own forum reported 8–10 second response times across all characters in March 2026 — "the STT service is also cutting out," one user noted. This is a recurring issue, not a one-off.
Long-term memory bug: SDK version 3.6.9-hotfix-2 broke long-term memory in Unreal Engine. Affected users reported characters telling them they had no memory of prior conversations — directly contradicting what paid plans advertise.
Hidden flagship LLM interaction cap: Each plan has a secondary limit on how many interactions can use the best AI models. On Indie Dev, only 1,500 of your 3,000 monthly interactions can use flagship models. The remaining 1,500 auto-downgrade without warning.
If these limitations are dealbreakers, two alternatives worth considering: OurDream AI for consumer-grade character creation with no SDK required, or Candy AI if you want polished chat and image generation without touching a game engine.
What it can do:
Real-time hands-free voice-to-voice conversations with AI characters via the new Web SDK (no push-to-talk required)
Custom knowledge bank upload — documents, PDFs, and text files that shape what your character knows and says
Vision and scene perception — characters identify objects in the 3D environment and respond contextually to them
Long-term memory, so characters remember past interactions per user session (when functioning correctly)
High-quality lip-sync and full-body animation, including support for Ready Player Me, MetaHuman, Daz3D, and Reallusion character assets
Emotion signaling and action callbacks — characters express mood states and trigger in-game behaviors based on conversation context
Plugin integrations with Unity, Unreal Engine, Roblox, PlayCanvas, Discord, and Second Life
Current limitations:
No dedicated mobile app
Primarily English-only (limited multilingual support via third-party TTS on higher tiers)
Long-term memory broken in recent Unreal SDK versions as of early 2026 — check the developer forum before shipping
Free plan capped at approximately 4,000 interactions per month with no flagship LLM access
Zero NSFW content of any kind — Convai is a developer and enterprise platform, not a companion app
If these limitations are dealbreakers, two simpler alternatives worth considering: OurDream AI for browser-based companion creation with no technical setup, or Joi AI for voice-forward character interaction without any SDK overhead.
Almost every new developer gets caught by this. Convai's plans are advertised by total monthly interaction count — but every paid tier also carries a secondary cap that determines how many of those interactions can use the best available AI models.
On the Indie Dev plan at $29/month, you get 3,000 total interactions per month. That's the number in the headline. What isn't in the headline is that only 1,500 of those interactions can use a flagship model like GPT-4.1. Once that sub-limit is exhausted, your character automatically switches to a less capable model for the rest of the month — mid-session, without any alert or notification inside the application.
I found this buried in Convai's documentation under "Core AI Settings," well away from the pricing page. The developer forum has threads where users spent days troubleshooting why their character responses suddenly became less coherent halfway through their testing cycle. The answer, in most cases, was a depleted flagship LLM quota they didn't know existed. This isn't a scam — it's a documented feature — but the way it's surfaced (or not surfaced) in the purchase flow pushes developers toward the Professional tier ($99/month) faster than the pricing page alone would suggest. If you're building anything beyond a proof-of-concept prototype, price for Professional from the start.
Convai is infrastructure. It's the engine room. If you're not building a game or XR simulation, it's the wrong starting point.
OurDream AI is where non-developers should go instead. No SDK. No API keys. No Unreal plugin to configure. You design a character in a browser — choosing from 19 voice styles, 40+ personality presets, and hundreds of character traits — and you're having a conversation within minutes. The DreamCoin system handles image generation, video clips, and voice messages in the same interface you're already in. The premium plan runs $19.99/month, or $9.99/month on an annual plan. That's less than Convai's Indie Dev tier, with none of the setup overhead. Convai builds the engine. OurDream lets you drive it without needing a mechanic.
Convai has been featured in NVIDIA's developer blog, integrated by Linden Lab as the backbone of Second Life's AI character generation system, and covered by GamesBeat as a leading NPC platform. Thousands of studios and indie developers have been using it since its first public product launch in March 2023. The developer forum at forum.convai.com is actively maintained, with staff responding to most bug reports and technical threads within 24–48 hours. The platform is ISO 27001 certified. The technology is genuine and ahead of the consumer companion market. The reliability issues are real and recurring, which is the honest caveat. Good platform, unstable backend — that's the current state.
Convai is worth it for developers and small studios building interactive games, training simulations, or XR experiences where character depth matters — knowledge banks, scene perception, and game engine integration are capabilities that no consumer companion app on the market provides. The free tier (around 4,000 interactions/month depending on current plan limits) is sufficient to validate a prototype at no cost. For anything production-bound, budget for the Professional tier at $99/month, because the Indie Dev plan's hidden 1,500 flagship LLM cap will frustrate you faster than the $29 price tag. If you want to talk to an AI character and never touch a game engine, Convai is not the answer — start with OurDream AI instead.
Convai is a developer and enterprise platform with zero tolerance for adult, explicit, or sexual content. The character response pipeline includes built-in content moderation, and there is no mode, setting, or tier that enables NSFW outputs. All interactions are processed on Convai's cloud infrastructure and subject to usage policy enforcement. Convai does not display advertising and does not use character interaction data for third-party advertising purposes. Developers building consumer-facing applications with Convai's API are responsible for ensuring their own downstream content policies comply with local regulations.
Convai's usage policies prohibit:
Generating explicit, sexual, or adult content through any character or API endpoint
Creating characters designed to harass, threaten, or incite hatred toward any individual or group
Impersonating real, identifiable public figures or brands without documented authorization
Building applications that bypass age verification or content safety mechanisms in consumer-facing products
Attempting to jailbreak character responses or circumvent Convai's moderation pipeline
Unauthorized resale, redistribution, or sub-licensing of Convai API access or character assets
Violations result in account suspension and revocation of API credentials.
Convai processes voice inputs and text interactions on its cloud infrastructure to generate character responses. Chat logs and audio data are stored server-side. The platform holds ISO 27001 certification, indicating audited information security management practices. Convai's stated privacy policy does not include third-party data sales. Developers integrating the API into consumer applications should disclose to their end users that interaction data is processed by Convai's servers — this is not handled automatically by the SDK.
Convai enforces age verification through account registration and applies content moderation to all character outputs at the API level. The platform is explicitly designed for gaming, enterprise training, and educational contexts. Age-gating of end-user applications built on Convai's API is the developer's responsibility, not Convai's — a point that developers deploying consumer-facing tools should build into their own compliance workflow.
Convai Technologies Inc. is headquartered in San Jose, California, and was incorporated in 2022. The company secured a $5 million seed round from Dune Ventures, Monta Vista Capital, and Rainfall Ventures. Founder and CEO Purnendu Mukherjee previously served as a Senior Deep Learning Software Engineer at NVIDIA's Conversational AI organization — where he built NVIDIA's first 3D avatar chatbot, demonstrated by CEO Jensen Huang at GTC 2020. Convai has been covered by GamesBeat, featured in NVIDIA's developer blog, and selected by Linden Lab as the AI character technology underpinning Second Life's character generation system.
Support: [email protected] and developer forum at forum.convai.com
Discord: approximately 3,137 members — discord.com/invite/9fSFn7aPUa
Developer Forum: forum.convai.com — active, staff response typically within 24–48 hours
Documentation: docs.convai.com — includes changelogs, SDK guides, and API reference
Platform updates: pushed regularly; check docs.convai.com/changelog for the latest
Convai offers a freemium model with five tiers under the Creator module and a separate Enterprise module for large-scale custom deployments.
Approximately 4,000 interactions per month (verify current limit at convai.com/pricing — figures vary across sources)
Access to the character creation playground and basic API
Knowledge bank: 1MB
Character session concurrency: 1
No flagship LLM access
No long-term memory
Access to the Developer Forum for support
Plan | Monthly Billing | Annual Billing |
Indie Dev | $29/month | ~$22/month |
Professional | $99/month | ~$69/month |
Scale | $499/month | Contact Convai |
Business | $1,199/month | Contact Convai |
Enterprise | Custom | Custom |
Indie Dev ($29/month) includes:
3,000 interactions/month total (1,500 capped for flagship LLMs)
Knowledge bank: 5MB
Long-term memory
Character session concurrency: 1
Cloud Rendered Avatar Studio access: 10 hours/month
Developer Forum support
Professional ($99/month) includes:
Up to 10,000–40,000 interactions/month depending on model (verify at convai.com/pricing)
Knowledge bank: 20MB
Higher character concurrency limits
Cloud Rendered Avatar Studio: 35 hours/month
Character Base API access — required for dynamic character creation via API
Priority support
Scale ($499/month) includes:
50,000+ interactions/month
Knowledge bank: 100MB
Designed for production-scale and multi-user deployments
Annual billing saves approximately 25% across the Indie Dev and Professional tiers — the Indie Dev plan drops from $29 to around $22/month, and Professional from $99 to around $69/month. There is no publicly listed promo code as of April 2026. The free tier is genuinely functional for prototyping and does not require a credit card. All paid plans auto-renew — review cancellation terms before upgrading, particularly if you are testing the Professional tier before a production decision.
Convai primarily supports English across its character voice, speech-to-text, and language understanding pipelines. Limited multilingual text-to-speech is available on higher-tier plans through third-party TTS integrations. If you are building for a non-English audience, check convai.com directly for current language availability before committing to a plan.
Main Platform
Website: https://convai.com/
Character Playground: https://convai.com/playground
Pricing: https://convai.com/pricing
Documentation & Support
Developer Docs: https://docs.convai.com/
Developer Forum: https://forum.convai.com/
API Reference: https://docs.convai.com/api-docs/
Community
Discord (~3,137 members): https://discord.com/invite/9fSFn7aPUa
Twitter/X: https://twitter.com/convaitech
Convai's biggest limitation isn't the latency or the flagship LLM cap — it's the requirement to understand APIs, game engine plugins, and SDK versioning just to have a conversation with your character. OurDream AI removes that barrier entirely. Where Convai puts character creation behind a developer workflow, OurDream gives you a browser-based studio where you select a personality, choose from 19 voice styles and 40+ presets, and start chatting within minutes of signing up. There's no Unity project, no Unreal Engine plugin, and no interaction quota to mentally track mid-month. The DreamCoin system handles image generation, video clips, and voice messages directly inside the same interface where you're already talking — something Convai's playground doesn't support at all. For users who want immersive character interaction rather than character infrastructure, OurDream covers vastly more of what most people actually want. The premium plan costs $19.99/month, or $9.99/month billed annually — less than Convai's Indie Dev tier, with none of the technical overhead.
Candy AI won't help you build a game NPC. But if Convai caught your attention because you wanted a deeply customized AI persona you could actually talk to, Candy AI delivers that in a finished consumer product. The chat quality is warm, consistent, and never drifts into generic AI-speak the way characters can when Convai's knowledge bank context runs thin. The V2 image engine produces crisp, consistent visual outputs from your companion, with each image costing around 2–4 tokens depending on settings. Candy AI runs on a token system that adds up fast — expect $50–$100/month for heavy use — but the product is polished in ways that a developer platform optimized for game engine integration simply isn't. The core distinction from Convai is intentional: Candy AI is designed to be used; Convai is designed to be built on. If you want to skip the infrastructure entirely and go straight to an actual conversation, Candy AI is the fastest path to a working companion experience.
Convai's voice pipeline is technically capable — but implementing it means embedding a plugin into a game engine or integrating the Web SDK into a web application. Joi AI skips all of that. It's a consumer-facing voice and chat companion app where voice quality is the primary feature, with conversations that feel natural and response times that compare favorably to what Convai's forum has been documenting. There's no developer account, no API key to generate, and no character session concurrency to worry about. The tradeoff is customization depth — you're working within Joi AI's predefined character options rather than uploading lore documents and training a knowledge bank around them. For users who want voice-first AI interaction without any technical setup, Joi AI is the clean answer. It won't give you a character that knows your game's worldbuilding, but it will give you a character that talks back in under a second.
For developers and studios: Convai is still the standard for AI NPC infrastructure. The knowledge bank accuracy, game engine depth, and scene perception haven't been replicated by any competitor at this price point. The latency problems are real but appear to be infrastructure-side — the underlying technology is sound. For everyone else: if the phrase "AI characters" made you think companion app rather than developer SDK, start with OurDream AI. The customization depth is comparable, the price is lower, and there's no learning curve between you and your first conversation.
Convai is the right choice for developers building interactive games, XR training, or virtual world experiences where character depth actually matters — knowledge banks, scene perception, and real-time voice are things no consumer companion app offers at any price. Go in knowing that the Indie Dev plan's hidden flagship LLM cap will push you toward Professional faster than expected, and that latency issues have been persistent enough to warrant checking the forum before any production launch. If you're not a developer — if you came here because you want to talk to a custom AI character and nothing more — Convai is not your product. Go to OurDream AI or Candy AI. You will be in a conversation in five minutes, not two hours of plugin configuration.