Convai Alternative for Unity and Indie Game Developers

The trouble with interaction quotas shows up after the demo, not during it. Convai is a cloud platform for conversational 3D characters, with a Unity SDK, an Unreal Engine plugin, web SDKs, lip sync and a large voice library, and it serves training, XR and brand-avatar projects as well as games. Indie teams usually start looking for a Convai alternative when they model what a real launch would cost, how many players can talk at once, and where the AI runs.

This post works through those three questions with a concrete example, then shows how Neocortex handles the same game.

How does Convai count usage?

Convai's FAQ defines the unit: every character response is one interaction, and each plan has a monthly quota.

• Indie Dev includes 3,000 interactions a month.
• Only 1,500 of those can use flagship LLMs. After that, the character has to run on a non-flagship model for the rest of the month.
• 500 interactions can use HQ third-party voices. Past that, those voices are unavailable unless you bring your own ElevenLabs key and pay ElevenLabs separately.
• Quotas reset at the start of each billing cycle. Unused interactions do not carry over.
• Higher plans add a cap on monthly active end users: the Scale plan allows 200, and bigger audiences need a plan arranged with sales.

On top of the quota sits concurrency, which Convai sets per deployed API key, not per NPC. It limits how many sessions can be connected at once. With a limit of 1, you can have ten NPCs in a scene, but only one player can talk to any of them at a time.

What does a launch weekend look like on Indie Dev?

Take a small narrative RPG with four talkative NPCs: an innkeeper, a quest giver, a merchant and a companion.

• 300 players try the game on launch weekend.
• Each player has about 15 exchanges across the four NPCs in a first session.
• That is 4,500 character responses in two days.

That one weekend is already past the Indie Dev monthly quota of 3,000. The flagship model allowance of 1,500 runs out first, so the characters switch to a non-flagship model partway through the weekend, and the same companion may start answering differently. If 20 of those players are mid-conversation at the same moment, your key's concurrency limit has to be at least 20, or some of them wait.

None of this is a problem for a scripted demo, a trade-show booth or a training simulation with a known number of users. It is a problem for a game where you do not control how much people talk.

Can Convai run offline?

Not on the self-serve plans. Convai runs in the cloud, and every conversation goes through its servers. Its documentation describes on-premises deployment with Docker, but the images are not publicly available: you contact Convai to get access. Offline support has been requested on Convai's community forum.

For a game that has to work on a plane, in a classroom with a locked-down network, or for a studio with strict data rules, that leaves a sales conversation as the only route.

Is Neocortex a good Convai alternative for Unity?

Every hosted AI service meters usage somehow, and Neocortex is no exception: plans come with monthly credits. What changes is what you can predict, what you can test for free, and who controls the character.

A plan price you can read. The Free plan has 2 characters and 200 monthly credits for personal use. Starter is EUR 9 a month for personal and academic projects. Pro is EUR 29 a month and covers commercial use, group chat and 5 custom voices. Team is EUR 99 a month with 5 seats and 20 custom voices.

Testing never touches credits. Talking to a character in the browser editor is free on every plan, so writers can rewrite a personality fifty times without spending your launch budget.

Behaviour you can see. A Neocortex character is a graph of nodes: Agent, Personality, Emotion, Knowledge, World, Objects, Intents, Actions, Language, Accent and Voice. A designer can see why a character behaves the way it does and change it without touching code. Conversation Flows keep a character on a story beat and return checkpoint keywords (flowState) your game reacts to.

Built for gameplay. Your game sends nearby objects and recent events. Replies come back as lines with an emotion tag, plus actions such as OPEN_DOOR with a targetId, which your code executes. On Pro and Team, up to six characters can share one conversation over the API.

Local while you prototype. You can run local models through Ollama in Unity for text chat while you build, then switch to the cloud. Full self-hosting is a Custom enterprise arrangement.

Convai vs Neocortex side by side

Neocortex Unity builds also get microphone support in WebGL, so browser players can speak to your characters.

For a wider look at how quota, credit and per-minute models compare, see our AI NPC pricing models post.

Our recommendation

If you are building a game where players talk freely to several NPCs, start by modelling a launch weekend before you pick a platform. With Neocortex, you can build the character in the browser today, test it for free until it sounds right, drop it into Unity with a sample scene, and move to Pro when the game is ready to sell.

Competitor details were last checked in October 2026. Plans change often, so check Convai's pricing page for current limits.

Frequently Asked Questions