Skip to content
RespawnStack

Guides

AI NPC dialogue in Unity & Godot: tools, cost, Steam rules

In short

To add AI NPC dialogue to a Unity or Godot game, first decide whether lines are written during development (Steam: Pre-Generated) or generated while the game runs (Live-Generated, which means describing your guardrails to Valve).
Rest of this summary For live dialogue, pick a character platform with an engine SDK (Convai and Inworld for Unity, Player2 for Godot and Unity) or a raw LLM API (Claude, OpenAI) behind your own server. Keep API keys out of the build, add guardrails, and budget the per-call cost.

Live AI dialogue changes what you tell Steam, what you pay every time a player talks, and where your API key ends up. This guide compares three NPC platforms and two raw LLM APIs from their own docs, then walks through the integration in seven steps.

Not tested hands-on. Untested (research-based). Built from Convai, Inworld, Player2, Anthropic and OpenAI pricing pages, docs and terms, plus Unity, Godot and Steamworks documentation, quoted with fetch dates and re-checked. We have not integrated any of these services hands-on and have not measured latency or token use. This is not legal advice.

5 tools · 41 vendor, engine and Steamworks doc quotes plus ledger rows · updated 2026-11-18 · last checked 2026-11-19

NPC dialogue options side by side: Convai, Inworld, Player2, Claude API, OpenAI API

Each cell is copied from the vendor's own pricing page, docs or terms, or from Steamworks. Open "Quotes" under a row to see the exact text and the date we fetched it. Prices are in US dollars. Vendors count usage in different units (interactions, credits, tokens), so the price row doesn't compare one to one.

  • What you are buying

    Convai
    A character platform: you build the NPC on Convai and the SDK connects to it. Long Term Memory is listed on every plan except Free.
    Inworld
    A runtime for building characters yourself. Inworld says its legacy Character Studio has been retired; the runtime covers knowledge retrieval, safety checks, long-term memory and voice.
    Player2
    An AI NPC platform with engine integrations. Player2 pays the API costs and players run the Player2 App.
    Anthropic Claude
    A raw LLM API. You write the character prompt, memory and guardrails yourself.
    OpenAI ChatGPT and GPT API
    A raw LLM API, plus a separate Moderation endpoint you can use as a guardrail.
    Quotes behind this (9)
  • Cheapest price

    Convai
    Free: 100 interactions a month. Indie Dev: $29/month ($22 billed yearly) for 3,000.
    Inworld
    On-Demand is $0 pay as you go; LLM Inference starts at $0.07 per 1M tokens on open models. Paid plans start with Creator at $25/month in credits.
    Player2
    No fees for creators: Player2 pays the API costs, and creators get a donation of 10% of its gross revenues.
    Anthropic Claude
    Claude Haiku 5.5: $0.10 per 1M input tokens and $0.50 per 1M output (prompts up to 100,000 tokens). Sonnet 5.5: $2 / $10.
    OpenAI ChatGPT and GPT API
    gpt-6-luna on the Flex row: $0.05 per 1M input tokens and $0.25 per 1M output. gpt-6.1-terra on that same Flex row: $0.50 / $2.50.
    Quotes behind this (9)
    • Free To test out the platform and it’s functionalities $0 /mo 100/month Interaction Quota 60 mins Cloud Rendered Avatar Studio and Convai Sim 1 Character Session Concurrency Get started Compare all features Indie Dev For creators and developers starting to build an application $29/mo $22 /mo Billed yearly 3000/month Interaction Quota 600 mins Cloud Rendered Avatar Studio and Convai Sim 1 Character Session Concurrency Subscribe Compare all features Professional For creators and developers ramping up development $99/mo $69 /mo Billed yearly 10,000/month Interaction Quota 2,400 mins Cloud Rendered Avatar Studio and Convai Sim 3 Character Session Concurrency Subscribe Compare all features Scale For building with a team and small scale production $499/mo $299 /mo Billed yearly 50,000/month Interaction Quota 10,200 mins Cloud Rendered Avatar Studio and Convai Sim 15 Character Session Concurrency
      Convai · Verified · fetched · convai.com/pricing
    • On-Demand Creator Builder Developer Growth Enterprise Price Free $25/mo $100/mo $300/mo $1,500/mo Custom
      Inworld · Verified · fetched · docs.inworld.ai/portal/billing
    • No, you don't have to pay any fees to use Player2! It has a generous free daily recharge to use premium models, we have free gpt-oss-120b model without any charge, and we share revenue with creators
      Player2 · Verified · fetched · player2.game/faq
    • It is completely FREE for creators , and you receive a donation of 10% of our gross revenues .
      Player2 · Verified · fetched · player2.game/forcreators
    • LLM Inference from $0.07 /1M tokens on open models
      Verified · fetched · inworld.ai/pricing
    • It is completely FREE for creators , and you receive a donation of 10% of our gross revenues .
      Verified · fetched · player2.game/forcreators
    • Claude Haiku 5.5 For high-volume, latency-sensitive tasks such as classification, extraction, and routing $0.10 / MTok for prompts up to 100,000 tokens $0.50 / MTok
      Verified · fetched · platform.claude.com/docs/en/about-claude/pricing
    • Claude Sonnet 5.5 The best combination of speed and intelligence $2 / MTok $10 / MTok
      Verified · fetched · platform.claude.com/docs/en/about-claude/pricing
    • gpt-6.1-terra $0.50 $2.50 $0.025 $0.625 $1.25 gpt-6-luna $0.05 $0.25 $0.005 $0.0625 $0.125
      Verified · fetched · developers.openai.com/api/docs/pricing
  • Player caps

    Convai
    Monthly Active End Users: 1 on Free, 75 on Indie Dev, 250 on Professional. The pricing table lists 1,250 for Scale, but the FAQ's example says Scale is capped at 200, so confirm with Convai. Concurrency is 1 session on Free and Indie Dev.
    Inworld
    Concurrent generations: 2 on free On-Demand (5 paid), 10 on Creator, 50 on Builder.
    Player2
    No player cap is stated on the pages we read. Each player needs the Player2 App open.
    Anthropic Claude
    Billed per token, so cost scales with players and lines. Rate limits not quoted here.
    OpenAI ChatGPT and GPT API
    Billed per token, so cost scales with players and lines. Rate limits not quoted here.
    Quotes behind this (7)
    • Just keep the Player2 App open on your device. You don't need to run any commands or worry about extra API keys
      Player2 · Verified · fetched · player2.game/faq
    • Monthly Active End Users Specifies the number of unique users who can interact with your AI Character each month. 1 75 250 1,250 2,500
      Verified · fetched · convai.com/pricing
    • Example: In the Scale plan, you have a cap of 200 Monthly Active End Users.
      Verified · fetched · convai.com/pricing
    • Character Session Concurrency The maximum number of simultaneous interactions an API key can support, regardless of the number of existing AI characters 1 1 3 15 30
      Verified · fetched · convai.com/pricing
    • On-Demand Creator Builder Developer Growth Enterprise Concurrent generations Free: 2, Paid: 5 10 50 150 500 Custom
      Verified · fetched · docs.inworld.ai/portal/billing
    • Claude Haiku 5.5 For high-volume, latency-sensitive tasks such as classification, extraction, and routing $0.10 / MTok for prompts up to 100,000 tokens $0.50 / MTok
      Verified · fetched · platform.claude.com/docs/en/about-claude/pricing
    • gpt-6.1-terra $0.50 $2.50 $0.025 $0.625 $1.25 gpt-6-luna $0.05 $0.25 $0.005 $0.0625 $0.125
      Verified · fetched · developers.openai.com/api/docs/pricing
  • Unity hookup

    Convai
    Convai Unity SDK via Package Manager or Asset Store. Needs Unity 6000.0.80f1 or later and an internet connection at editor time and runtime.
    Inworld
    Unity AI Runtime SDK with a Character template.
    Player2
    Unity Asset from Player2.
    Anthropic Claude
    No Unity SDK on the pages we read. Call the HTTP API through your own server with UnityWebRequest.
    OpenAI ChatGPT and GPT API
    No Unity SDK on the pages we read. Call the HTTP API through your own server with UnityWebRequest.
    Quotes behind this (10)
  • Godot hookup

    Convai
    Not on the pages we read: Convai's docs we fetched cover Unity and an Unreal Engine plugin. Check Convai before planning a Godot build.
    Inworld
    Not on the page we read: it names Unity, Unreal Engine and Node.js.
    Player2
    Godot Plugin. The FAQ also lists Defold.
    Anthropic Claude
    Call the HTTP API through your own server with Godot's HTTPRequest node.
    OpenAI ChatGPT and GPT API
    Call the HTTP API through your own server with Godot's HTTPRequest node.
    Quotes behind this (5)
  • Who owns the generated lines

    Convai
    Not verified. The pricing page lists data ownership and control as an Enterprise feature, so read Convai's terms before shipping.
    Inworld
    Inworld assigns its rights in Outputs to you, but licenses the service solely for internal business purposes as the docs permit. Read the terms before a game calls it live.
    Player2
    You own your Models and their Generations, but grant Elefant AI a perpetual, sublicensable license to commercialize them.
    Anthropic Claude
    Under the Commercial Terms (API keys), the customer owns its Outputs, with an IP indemnity for paid use.
    OpenAI ChatGPT and GPT API
    The Services Agreement (API) says the customer owns all Output; the IP indemnity covers API customers. Output may not be unique.
    Quotes behind this (10) · a quote failed its last re-check
    • Data Ownership and Control - Enterprise clients own and control their data, including all input and output data during development
      Convai · Verified · fetched · convai.com/pricing
    • you retain any right, title, and interest that you have in the Inputs you submit. Subject to your compliance with our Terms, we assign to you all of our right, title, and interest in Outputs.
      Inworld · Verified · fetched · inworld.ai/terms
    • Inworld AI grants you a limited, non-exclusive, non-transferable, non-sublicensable license to access and use the Services and Models solely for your internal business purposes and as permitted by the documentation.
      Inworld · Verified · fetched · inworld.ai/terms
    • As between you and Elefant AI, you also own any text, images, audio, or video the Model generates ("Generations") that are elicited by you.
      Player2 · Verified · fetched · player2.game/tos
    • You grant Elefant AI, to the fullest extent permitted under the law, a nonexclusive, worldwide, royalty-free, fully paid-up, transferable, sublicensable, perpetual, irrevocable license to copy, display, upload, perform, distribute, transmit, make available, store, access, modify, exploit, commercialize, and otherwise use the Model and all Generations elicited by you for any purpose
      Player2 · Verified · fetched · player2.game/tos
    • As between the parties and to the extent permitted by applicable law, Anthropic agrees that Customer (a) retains all rights to its Inputs, and (b) owns its Outputs.
      Anthropic Claude · Verified · fetched · anthropic.com/legal/commercial-terms
    • Customer’s paid use of the Services (which includes data Anthropic has used to train a model that is part of the Services) in accordance with these Terms or Outputs generated through such authorized use violates any third-party intellectual property right.
      Anthropic Claude · Verified · fetched · anthropic.com/legal/commercial-terms
    • As between Customer and OpenAI, to the extent permitted by applicable law, Customer: (a) retains all ownership rights in Input; and (b) owns all Output. OpenAI hereby assigns to Customer all OpenAI’s right, title, and interest, if any, in and to Output.
      OpenAI ChatGPT and GPT API · Verified · fetched · openai.com/policies/services-agreement/
    • OpenAI’s indemnification obligations to API customers under the Agreement include any third party claim that Customer’s use or distribution of Output infringes a third party’s intellectual property right.
      OpenAI ChatGPT and GPT API · Verified · fetched · openai.com/policies/service-terms/
    • Due to the nature of our Services and artificial intelligence generally, output may not be unique and other users may receive similar output from our Services. Our assignment above does not extend to other users’ output or any Third Party Output.
      OpenAI ChatGPT and GPT API · Verified · fetched · openai.com/policies/terms-of-use

Runtime NPC dialogue options

Three character platforms and two raw LLM APIs. Platforms give you characters, memory and an engine SDK. Raw APIs are cheaper per token but leave prompts, memory, guardrails and key handling to you.

  • Character platform with a Unity SDK (Unity 6000.0.80f1+, internet at runtime) and an Unreal plugin. Ship builds in Auth Token mode, not API Key mode. Plans cap Monthly Active End Users.

    Cheapest paid
    Indie Dev, $29/month ($22 billed yearly): 3,000 interactions, 75 Monthly Active End Users
    Free tier
    Free: 100 interactions a month, 1 Monthly Active End User
    Quotes behind this (5)
    • Free To test out the platform and it’s functionalities $0 /mo 100/month Interaction Quota 60 mins Cloud Rendered Avatar Studio and Convai Sim 1 Character Session Concurrency Get started Compare all features Indie Dev For creators and developers starting to build an application $29/mo $22 /mo Billed yearly 3000/month Interaction Quota 600 mins Cloud Rendered Avatar Studio and Convai Sim 1 Character Session Concurrency Subscribe Compare all features Professional For creators and developers ramping up development $99/mo $69 /mo Billed yearly 10,000/month Interaction Quota 2,400 mins Cloud Rendered Avatar Studio and Convai Sim 3 Character Session Concurrency Subscribe Compare all features Scale For building with a team and small scale production $499/mo $299 /mo Billed yearly 50,000/month Interaction Quota 10,200 mins Cloud Rendered Avatar Studio and Convai Sim 15 Character Session Concurrency
      Convai · Verified · fetched · convai.com/pricing
    • Unity version 6000.0.80f1 or later Scripting backend Mono or IL2CPP Internet connection Required at editor time and runtime
      Convai · Verified · fetched · docs.convai.com/api-docs/plugins-and-integrations/convai-unity-sdk/get
    • Data Ownership and Control - Enterprise clients own and control their data, including all input and output data during development
      Convai · Verified · fetched · convai.com/pricing
    • Monthly Active End Users Specifies the number of unique users who can interact with your AI Character each month. 1 75 250 1,250 2,500
      Verified · fetched · convai.com/pricing
    • API Key mode is for local development, not for a build you distribute.
      Verified · fetched · docs.convai.com/api-docs/plugins-and-integrations/convai-unity-sdk/get
  • Agent Runtime with a Unity AI Runtime SDK and an Unreal Agent Runtime SDK; the legacy Character Studio is retired. Outputs are assigned to you, but the service license is for internal business purposes.

    Cheapest paid
    Creator, $25/month in credits
    Free tier
    On-Demand, $0 pay as you go; LLM Inference from $0.07 per 1M tokens on open models
    Quotes behind this (6)
    • On-Demand Creator Builder Developer Growth Enterprise Price Free $25/mo $100/mo $300/mo $1,500/mo Custom
      Inworld · Verified · fetched · docs.inworld.ai/portal/billing
    • Unity Download the Unity AI Runtime SDK and use the Character template to build AI characters in Unity. Unreal Engine Download the Unreal Agent Runtime SDK and use the Character template to build characters in your engine.
      Inworld · Verified · fetched · docs.inworld.ai/character-studio-deprecation
    • you retain any right, title, and interest that you have in the Inputs you submit. Subject to your compliance with our Terms, we assign to you all of our right, title, and interest in Outputs.
      Inworld · Verified · fetched · inworld.ai/terms
    • Inworld AI grants you a limited, non-exclusive, non-transferable, non-sublicensable license to access and use the Services and Models solely for your internal business purposes and as permitted by the documentation.
      Inworld · Verified · fetched · inworld.ai/terms
    • The legacy Inworld Character Studio has been retired.
      Verified · fetched · docs.inworld.ai/character-studio-deprecation
    • LLM Inference from $0.07 /1M tokens on open models
      Verified · fetched · inworld.ai/pricing
  • AI NPC platform with a Godot Plugin and a Unity Asset. Player2 pays the API costs; each player keeps the Player2 App open. You grant Elefant AI a broad license to your Models and Generations.

    Cheapest paid
    No fees; creators receive a 10% revenue donation
    Free tier
    Free for creators
    Quotes behind this (5)
    • No, you don't have to pay any fees to use Player2! It has a generous free daily recharge to use premium models, we have free gpt-oss-120b model without any charge, and we share revenue with creators
      Player2 · Verified · fetched · player2.game/faq
    • We provide multiple creation tools including Godot/Unity/Defold AI NPC SDKs, a web-based story creator, and comprehensive APIs.
      Player2 · Verified · fetched · player2.game/faq
    • Just keep the Player2 App open on your device. You don't need to run any commands or worry about extra API keys
      Player2 · Verified · fetched · player2.game/faq
    • You grant Elefant AI, to the fullest extent permitted under the law, a nonexclusive, worldwide, royalty-free, fully paid-up, transferable, sublicensable, perpetual, irrevocable license to copy, display, upload, perform, distribute, transmit, make available, store, access, modify, exploit, commercialize, and otherwise use the Model and all Generations elicited by you for any purpose
      Player2 · Verified · fetched · player2.game/tos
    • Use our engine integrations for Godot and Unity to build full AI NPC games with complete control over gameplay. • Godot Plugin • Unity Asset
      Verified · fetched · player2.game/forcreators
  • Claude API, a RESTful API you call from your own server. Under the Commercial Terms the customer owns Outputs and Anthropic may not train on Customer Content.

    Cheapest paid
    Claude Haiku 5.5: $0.10 / $0.50 per 1M input / output tokens (prompts up to 100,000 tokens)
    Free tier
    Not quoted for the API
    Quotes behind this (4)
  • OpenAI API, called from your own server. The Moderation endpoint is free to use as one guardrail layer. The Services Agreement assigns Output to the customer.

    Cheapest paid
    gpt-6-luna Flex row: $0.05 / $0.25 per 1M input / output tokens
    Free tier
    Moderation endpoint is free; model use is paid
    Quotes behind this (3)
    • As between Customer and OpenAI, to the extent permitted by applicable law, Customer: (a) retains all ownership rights in Input; and (b) owns all Output. OpenAI hereby assigns to Customer all OpenAI’s right, title, and interest, if any, in and to Output.
      OpenAI ChatGPT and GPT API · Verified · fetched · openai.com/policies/services-agreement/
    • gpt-6.1-terra $0.50 $2.50 $0.025 $0.625 $1.25 gpt-6-luna $0.05 $0.25 $0.005 $0.0625 $0.125
      Verified · fetched · developers.openai.com/api/docs/pricing
    • The moderation endpoint is free to use
      Verified · fetched · developers.openai.com/api/docs/guides/moderation

Prices, caps and engine support are quoted from each vendor's pricing page, docs or terms. Interactions, credits and tokens are the vendors' own units and don't compare one to one. Latency is not listed because none of the quoted pages give a measured figure.

How to add AI NPC dialogue to a Unity or Godot game, step by step

Seven steps, from the decision that changes your Steam answers to the one that keeps your bill bounded. Each step cites the vendor, engine or Steamworks text it relies on. We have not built this integration hands-on.

  1. Step 1. Decide: pre-written with AI, or generated while the game runs

    If you use AI to write lines during development and ship them as text or audio, Steam calls that Pre-Generated. If the NPC generates lines while the game runs, it is Live-Generated, which adds a requirement: you describe your guardrails in the Content Survey. Pre-generating is the simpler path; everything below is for live dialogue. Full walkthrough: Steam AI disclosure.

    Quotes behind this (2)
    • Pre-Generated: Any kind of content that ships with your game and is consumed by players that is created with the help of AI tools during development.
      Verified · fetched · partner.steamgames.com/doc/gettingstarted/contentsurvey
    • Live-Generated: Any kind of content created with the help of AI tools while the game is running. In addition to following the same rules as Pre-Generated AI content, this comes with an additional requirement - in the Content Survey, you'll need to tell us what kind of guardrails you're putting on your AI to ensure it's not generating illegal content.
      Verified · fetched · partner.steamgames.com/doc/gettingstarted/contentsurvey
  2. Step 2. Pick a platform or a raw API

    A platform gives you a character, memory and an engine SDK: Convai (Unity SDK, Unreal plugin), Inworld (Unity and Unreal runtime SDKs) or Player2 (Godot Plugin and Unity Asset). A raw API (Claude, OpenAI) is billed per token and leaves the character logic to you. For Godot, Player2 is the only one of the five whose pages we read that names a Godot integration; for the others you call an HTTP API.

    Quotes behind this (5)
  3. Step 3. Check the license and the player caps before you build

    Read the ownership terms for the exact service. Inworld assigns Outputs to you but licenses the service for internal business purposes. Player2 lets you own Generations but takes a broad license to commercialize them. Convai's ownership terms are not verified in our ledger. Then check caps: Convai's plans limit Monthly Active End Users (1 on Free, 75 on Indie Dev), and Inworld's plans limit concurrent generations. See who owns AI game assets.

    Quotes behind this (5) · a quote failed its last re-check
    • you retain any right, title, and interest that you have in the Inputs you submit. Subject to your compliance with our Terms, we assign to you all of our right, title, and interest in Outputs.
      Inworld · Verified · fetched · inworld.ai/terms
    • Inworld AI grants you a limited, non-exclusive, non-transferable, non-sublicensable license to access and use the Services and Models solely for your internal business purposes and as permitted by the documentation.
      Inworld · Verified · fetched · inworld.ai/terms
    • You grant Elefant AI, to the fullest extent permitted under the law, a nonexclusive, worldwide, royalty-free, fully paid-up, transferable, sublicensable, perpetual, irrevocable license to copy, display, upload, perform, distribute, transmit, make available, store, access, modify, exploit, commercialize, and otherwise use the Model and all Generations elicited by you for any purpose
      Player2 · Verified · fetched · player2.game/tos
    • Monthly Active End Users Specifies the number of unique users who can interact with your AI Character each month. 1 75 250 1,250 2,500
      Verified · fetched · convai.com/pricing
    • On-Demand Creator Builder Developer Growth Enterprise Concurrent generations Free: 2, Paid: 5 10 50 150 500 Custom
      Verified · fetched · docs.inworld.ai/portal/billing
  4. Step 4. Keep your API key out of the shipped build

    Convai's Unity docs are explicit: API Key mode is for local development, and a build made in that mode contains your account key, obfuscated but not encrypted. For anything you ship, use Auth Token mode, where a server you control issues short-lived tokens. Editorial: the same logic applies to a Claude or OpenAI key. Put it on a small server of your own and have the game call that server, never the vendor directly.

    Quotes behind this (4)
  5. Step 5. Call it from the engine

    With an SDK, install it in the engine: Convai through Unity's Package Manager or the Asset Store, Inworld's Unity AI Runtime SDK, Player2's Godot Plugin or Unity Asset. With a raw API, send HTTP from your game to your server. In Unity, UnityWebRequest handles HTTP and you send with SendWebRequest. In Godot, use the HTTPRequest node; don't make simultaneous requests on one node, and for Android exports enable the INTERNET permission. Convai's SDK also needs an internet connection at runtime, so plan for what an NPC says when the player is offline.

    Quotes behind this (9)
  6. Step 6. Write the guardrails Steam will ask about

    For Live-Generated content, the Content Survey asks what guardrails stop your AI generating illegal content, and Valve says it doesn't want to ship Live-Generated AI Adult Only Sexual Content at this time. Options from the docs: OpenAI's Moderation endpoint is free and its results can filter content or route it for review (it doesn't classify audio), and Inworld's runtime lists safety checks. Editorial: also constrain the system prompt to your game's world, cap reply length, and log flagged lines so you can describe the setup honestly in the survey.

    Quotes behind this (6)
  7. Step 7. Budget the ongoing cost, then disclose

    Steam's FAQ covers games that connect to a Live-Generated AI service with ongoing per-call costs: you manage access to that service for your players and collect payment through a Steam-supported method, such as the game's price, microtransactions, a subscription or DLC. Claude Haiku 5.5 is $0.10 per 1M input tokens, and OpenAI's Flex row lists gpt-6-luna at $0.05 per 1M input tokens; Player2 pays its API costs itself. Finish by answering the Content Survey; Valve compares answers with your build and store page.

    Quotes behind this (8)

Caveats we could only partly check

What the sources say, and what they leave open. Latency is not here because none of the vendors' pages we quote give a measured number for game dialogue, and we have not tested one.

  • Offline play

    Convai's Unity SDK lists an internet connection as required at runtime. Player2 needs its App open on the player's device. A raw API call needs a network round trip to your server and the vendor. None of the five pages we read describe an offline mode.

    Quotes behind this (3)
  • Convai's player cap is stated two ways

    The pricing table lists 1, 75, 250, 1,250 and 2,500 Monthly Active End Users across Free to Business, but the FAQ's example says the Scale plan has a cap of 200. Ask Convai before relying on the Scale number.

    Quotes behind this (2)
    • Monthly Active End Users Specifies the number of unique users who can interact with your AI Character each month. 1 75 250 1,250 2,500
      Verified · fetched · convai.com/pricing
    • Example: In the Scale plan, you have a cap of 200 Monthly Active End Users.
      Verified · fetched · convai.com/pricing
  • Inworld's terms vs shipping

    Inworld's pricing page shows 'Commercial license' on On-Demand and assigns Outputs to you, but the terms license the service solely for internal business purposes as the docs permit. We have not seen a game-specific clause. Read the terms, or ask Inworld, before players call it live.

    Quotes behind this (3)
    • Commercial license Community support Choose On-Demand
      Inworld · Verified · fetched · inworld.ai/pricing
    • you retain any right, title, and interest that you have in the Inputs you submit. Subject to your compliance with our Terms, we assign to you all of our right, title, and interest in Outputs.
      Inworld · Verified · fetched · inworld.ai/terms
    • Inworld AI grants you a limited, non-exclusive, non-transferable, non-sublicensable license to access and use the Services and Models solely for your internal business purposes and as permitted by the documentation.
      Inworld · Verified · fetched · inworld.ai/terms
  • Units don't compare

    Convai sells interactions, Inworld sells credits and per-token inference, and Claude and OpenAI bill per million tokens. We have not measured how many tokens a typical NPC exchange uses, so we do not convert between them.

    Quotes behind this (4)
    • Free To test out the platform and it’s functionalities $0 /mo 100/month Interaction Quota 60 mins Cloud Rendered Avatar Studio and Convai Sim 1 Character Session Concurrency Get started Compare all features Indie Dev For creators and developers starting to build an application $29/mo $22 /mo Billed yearly 3000/month Interaction Quota 600 mins Cloud Rendered Avatar Studio and Convai Sim 1 Character Session Concurrency Subscribe Compare all features Professional For creators and developers ramping up development $99/mo $69 /mo Billed yearly 10,000/month Interaction Quota 2,400 mins Cloud Rendered Avatar Studio and Convai Sim 3 Character Session Concurrency Subscribe Compare all features Scale For building with a team and small scale production $499/mo $299 /mo Billed yearly 50,000/month Interaction Quota 10,200 mins Cloud Rendered Avatar Studio and Convai Sim 15 Character Session Concurrency
      Convai · Verified · fetched · convai.com/pricing
    • On-Demand Creator Builder Developer Growth Enterprise Price Free $25/mo $100/mo $300/mo $1,500/mo Custom
      Inworld · Verified · fetched · docs.inworld.ai/portal/billing
    • Claude Haiku 5.5 For high-volume, latency-sensitive tasks such as classification, extraction, and routing $0.10 / MTok for prompts up to 100,000 tokens $0.50 / MTok
      Verified · fetched · platform.claude.com/docs/en/about-claude/pricing
    • gpt-6.1-terra $0.50 $2.50 $0.025 $0.625 $1.25 gpt-6-luna $0.05 $0.25 $0.005 $0.0625 $0.125
      Verified · fetched · developers.openai.com/api/docs/pricing

Editorial opinion

Our pick: which NPC dialogue option to start with

  • Pick Player2 for a Godot or Unity prototype or jam game with no AI bill: it pays the API costs and has a Godot Plugin. Accept that players need its App and that you grant Elefant AI a broad license.
  • Pick Convai for a Unity NPC with memory and an SDK, if the player caps fit (75 Monthly Active End Users on Indie Dev). Ship in Auth Token mode.
  • Pick Claude Haiku 5.5 or gpt-6-luna behind your own server when you want control and per-token pricing, and can write the prompts, memory and guardrails yourself. Use OpenAI's free Moderation endpoint as one guardrail layer.
  • Pick Pre-generate instead when you only need variety, not open conversation: write lines with AI during development, review them, and ship them. Steam treats that as Pre-Generated, with no guardrail section.

Picks and the steps marked 'Editorial' are our opinion from the quoted docs, terms and prices. We have not tested these services hands-on.

FAQ

What is the best AI NPC dialogue generator for Unity?
We have not tested any hands-on. From the docs: Convai and Inworld both ship Unity SDKs (Convai via Package Manager or the Asset Store; Inworld's Unity AI Runtime SDK), and Player2 has a Unity Asset with no fees to creators. With Claude or OpenAI you call the HTTP API from your own server with UnityWebRequest.
Sources (5)
  • Unity version 6000.0.80f1 or later Scripting backend Mono or IL2CPP Internet connection Required at editor time and runtime
    Convai · Verified · fetched · docs.convai.com/api-docs/plugins-and-integrations/convai-unity-sdk/get
  • Unity Download the Unity AI Runtime SDK and use the Character template to build AI characters in Unity. Unreal Engine Download the Unreal Agent Runtime SDK and use the Character template to build characters in your engine.
    Inworld · Verified · fetched · docs.inworld.ai/character-studio-deprecation
  • No, you don't have to pay any fees to use Player2! It has a generous free daily recharge to use premium models, we have free gpt-oss-120b model without any charge, and we share revenue with creators
    Player2 · Verified · fetched · player2.game/faq
  • Use our engine integrations for Godot and Unity to build full AI NPC games with complete control over gameplay. • Godot Plugin • Unity Asset
    Verified · fetched · player2.game/forcreators
  • UnityWebRequest handles the flow of HTTP communication with web servers.
    Verified · fetched · docs.unity3d.com/ScriptReference/Networking.UnityWebRequest.html
Can I add AI NPC dialogue in Godot?
Yes. Player2 has a Godot Plugin. Of the other four, none of the pages we read name a Godot integration, so you would call Claude, OpenAI or Inworld's API through your own server with Godot's HTTPRequest node.
Sources (3)
  • Use our engine integrations for Godot and Unity to build full AI NPC games with complete control over gameplay. • Godot Plugin • Unity Asset
    Verified · fetched · player2.game/forcreators
  • A node with the ability to send HTTP requests. Uses HTTPClient internally.
    Verified · fetched · docs.godotengine.org/en/stable/classes/class_httprequest.html
  • Unity Download the Unity AI Runtime SDK and use the Character template to build AI characters in Unity. Unreal Engine Download the Unreal Agent Runtime SDK and use the Character template to build characters in your engine.
    Verified · fetched · docs.inworld.ai/character-studio-deprecation
How much does AI NPC dialogue cost?
Convai's Free plan has 100 interactions a month and Indie Dev is $29/month for 3,000. Inworld's On-Demand is pay as you go with LLM Inference from $0.07 per 1M tokens on open models. Claude Haiku 5.5 is $0.10 per 1M input tokens and $0.50 per 1M output. OpenAI's Flex row lists gpt-6-luna at $0.05 per 1M input tokens and $0.25 per 1M output. Player2 charges creators nothing and pays the API costs.
Sources (5)
  • Free To test out the platform and it’s functionalities $0 /mo 100/month Interaction Quota 60 mins Cloud Rendered Avatar Studio and Convai Sim 1 Character Session Concurrency Get started Compare all features Indie Dev For creators and developers starting to build an application $29/mo $22 /mo Billed yearly 3000/month Interaction Quota 600 mins Cloud Rendered Avatar Studio and Convai Sim 1 Character Session Concurrency Subscribe Compare all features Professional For creators and developers ramping up development $99/mo $69 /mo Billed yearly 10,000/month Interaction Quota 2,400 mins Cloud Rendered Avatar Studio and Convai Sim 3 Character Session Concurrency Subscribe Compare all features Scale For building with a team and small scale production $499/mo $299 /mo Billed yearly 50,000/month Interaction Quota 10,200 mins Cloud Rendered Avatar Studio and Convai Sim 15 Character Session Concurrency
    Convai · Verified · fetched · convai.com/pricing
  • No, you don't have to pay any fees to use Player2! It has a generous free daily recharge to use premium models, we have free gpt-oss-120b model without any charge, and we share revenue with creators
    Player2 · Verified · fetched · player2.game/faq
  • LLM Inference from $0.07 /1M tokens on open models
    Verified · fetched · inworld.ai/pricing
  • Claude Haiku 5.5 For high-volume, latency-sensitive tasks such as classification, extraction, and routing $0.10 / MTok for prompts up to 100,000 tokens $0.50 / MTok
    Verified · fetched · platform.claude.com/docs/en/about-claude/pricing
  • gpt-6.1-terra $0.50 $2.50 $0.025 $0.625 $1.25 gpt-6-luna $0.05 $0.25 $0.005 $0.0625 $0.125
    Verified · fetched · developers.openai.com/api/docs/pricing
Does Steam allow AI-generated NPC dialogue?
Yes, with disclosure. Dialogue generated while the game runs is Live-Generated: you follow the Pre-Generated rules and also describe your guardrails against illegal content in the Content Survey. Valve says it doesn't want to ship Live-Generated AI Adult Only Sexual Content at this time.
Sources (2)
  • Live-Generated: Any kind of content created with the help of AI tools while the game is running. In addition to following the same rules as Pre-Generated AI content, this comes with an additional requirement - in the Content Survey, you'll need to tell us what kind of guardrails you're putting on your AI to ensure it's not generating illegal content.
    Verified · fetched · partner.steamgames.com/doc/gettingstarted/contentsurvey
  • we don't want to ship Live-Generated AI Adult Only Sexual Content at this time.
    Verified · fetched · partner.steamgames.com/doc/gettingstarted/contentsurvey
Who pays for the AI when players talk to NPCs?
You do, unless the platform does. Steam's FAQ says you manage access to the external service for your players and collect payment through a Steam-supported method. Player2 says it pays the API costs, but players need the Player2 App open.
Sources (3)
  • Just keep the Player2 App open on your device. You don't need to run any commands or worry about extra API keys
    Player2 · Verified · fetched · player2.game/faq
  • you’ll need to manage both access to that external service on behalf of your players and collecting payment from your player using a Steam-supported payment method.
    Verified · fetched · partner.steamgames.com/doc/gettingstarted/contentsurvey
  • Player2 pays for all the API costs for you.
    Verified · fetched · player2.game/faq
Can I put my OpenAI or Claude API key in the game?
We would not. Convai's own Unity docs say a build made in API Key mode contains your key obfuscated but not encrypted, and that anyone with the SDK source can reverse it; they recommend short-lived tokens from a server you control. Editorial: treat any LLM key the same way and proxy calls through your server.
Sources (3)
Do AI NPCs work offline?
Not with the options here, as far as their pages say. Convai's Unity SDK requires an internet connection at runtime, and Player2 needs its App running on the player's device. Raw APIs need a network call.
Sources (2)
Who owns AI-generated NPC lines?
It depends on the service. Claude's Commercial Terms and OpenAI's Services Agreement say the customer owns Outputs. Inworld assigns Outputs to you but limits the service license to internal business purposes. Player2 lets you own Generations but takes a broad license to them. Convai's terms are not verified in our ledger.
Sources (5) · a quote failed its last re-check
  • As between the parties and to the extent permitted by applicable law, Anthropic agrees that Customer (a) retains all rights to its Inputs, and (b) owns its Outputs.
    Anthropic Claude · Verified · fetched · anthropic.com/legal/commercial-terms
  • As between Customer and OpenAI, to the extent permitted by applicable law, Customer: (a) retains all ownership rights in Input; and (b) owns all Output. OpenAI hereby assigns to Customer all OpenAI’s right, title, and interest, if any, in and to Output.
    OpenAI ChatGPT and GPT API · Verified · fetched · openai.com/policies/services-agreement/
  • you retain any right, title, and interest that you have in the Inputs you submit. Subject to your compliance with our Terms, we assign to you all of our right, title, and interest in Outputs.
    Inworld · Verified · fetched · inworld.ai/terms
  • Inworld AI grants you a limited, non-exclusive, non-transferable, non-sublicensable license to access and use the Services and Models solely for your internal business purposes and as permitted by the documentation.
    Inworld · Verified · fetched · inworld.ai/terms
  • You grant Elefant AI, to the fullest extent permitted under the law, a nonexclusive, worldwide, royalty-free, fully paid-up, transferable, sublicensable, perpetual, irrevocable license to copy, display, upload, perform, distribute, transmit, make available, store, access, modify, exploit, commercialize, and otherwise use the Model and all Generations elicited by you for any purpose
    Player2 · Verified · fetched · player2.game/tos

More guides

Get an email when a tool's terms change

We email when our daily check finds a pricing or terms change on a tool we track, the same entries as the change log.