gaforum

Games · 11 tools ranked

Best AI NPC and Dialogue Platforms in 2026

Convai comes first in our ranking of AI NPC and dialogue platforms, ahead of NVIDIA ACE and Inworld. AI NPC and dialogue platforms give game characters speech, memory and branching conversation through engine plugins, voice services and authoring tools.

Discuss this ranking in the forum

The top three

Ordering: editorial, based on what each vendor publishes about pricing, licensing and features as of the date above. Vendors can pay for more visibility; how that works is explained in our disclosure, linked above every ranking. No commission is earned on clicks or purchases, and we do not work as an affiliate.

What are AI NPC and dialogue platforms?

An AI NPC and dialogue platform lets a non-player character speak, answer players and remember context, using language models, speech synthesis and animation, or classic branching scripts. The category covers conversational engines such as Convai, authoring languages such as ink and Yarn Spinner, voice services, and tools that generate surrounding assets.

How we ranked these

Developers asking for an AI NPC tool usually mean one of three different things: a character that improvises, a writer's tool that structures planned scenes, or the art and voice that surround both. This ranking keeps those jobs in one list and orders them by how directly each gets a talking character into a shipped game.

We read vendor homepages, documentation, pricing pages and license terms, then compared published prices, supported engines and stated limits. Nothing here comes from running the products ourselves, and vendor claims are marked as such.

Conversational platforms sit at the top because they deliver the improvising character most readers want. Authoring tools follow, since many games are better off with written lines.

Asset generators close the list because they supply content around the character instead of the conversation itself.

Does the character actually converse?

Separate live conversation from authored branches. Convai and NVIDIA ACE generate responses with models, while ink, Yarn Spinner and articy:draft X play lines a writer composed. Pixel Crushers sits between them, adding a model addon to a scripted base.

Decide whether players will ask open questions or choose from options, because that single answer removes half the list. Asset generators such as Scenario, Layer and Ludo.ai do not converse at all.

Engine plugins and platforms

Check whether your engine has first-party support. Convai lists Unity and Unreal plugins plus web plugins; Yarn Spinner covers Unity, Godot and Unreal; ink offers Unity natively and Unreal through a community container; Dialogue System for Unity is Unity only; articy:draft X itself runs on Windows and macOS. A tool without a plugin for your engine costs integration hours that never show on a price sheet.

How the bill grows with players

Per-turn and per-character billing scales with player counts, while a license does not. Convai draws credits on every turn and session. Inworld quotes USD 25 per 1M characters for TTS-2.

ElevenLabs sells monthly credit tiers. ink costs nothing under MIT. NVIDIA ACE shows no price at all.

Estimate lines spoken per session before you pick a model-backed option.

Voice, latency and where inference runs

Spoken NPCs add speech-to-text, a language model and text-to-speech to each exchange, and every hop adds delay. NVIDIA ACE can run models on the player's GPU, which removes network round trips but limits hardware. Cloud services such as Inworld and ElevenLabs keep the player's machine free of that load.

Read the supported language list for voices and check whether a commercial license is included.

Compared at a glance

#ProductBest forPricingFree optionBased in
1ConvaiVoiced NPCs in 3D worldsUsage-basedFree
2NVIDIA ACETechnical teams with GPU budgetNot publishedUnknown
3InworldVoice-heavy characters on usage billingUsage-basedFree
4ElevenLabsVoice acting for dynamic linesSubscriptionFree
5Dialogue System for UnityUnity teams adding LLM chatAsset purchaseFree trial
6Yarn SpinnerWriters shipping branching dialogueFree and paid versionsFree
7inkText-first branching storiesOpen sourceOpen sourceUnited Kingdom
8articy:draft XLarge narrative teams on Windows or MacFree version plus licensesFreeGermany
9ScenarioStudios generating on-style assetsSubscriptionFree
10LayerArt teams mixing many modelsUsage-basedUnknownUnited States
11Ludo.aiIndie prototypes needing quick assetsSubscriptionFree trial

The 11 tools, reviewed

How this list is ordered: by our editorial judgment of fit for the reader named above, based on vendor documentation, pricing and license pages opened on the review date. Vendors can pay for more visibility; how that works is explained in our disclosure, linked above every ranking. No commission is earned on clicks or purchases, and we do not work as an affiliate.

  1. 1

    Convai

    Talking AI characters with Unity and Unreal plugins

    Why #1: Closest to a finished NPC product here: characters, engine plugins and a credit meter in one place, with a free tier to start on.

    Best for
    Voiced NPCs in 3D worlds
    Pricing
    Billing is credit-based: every conversation turn, connected session and processed file draws from the plan's credit balance, and the pricing page mentions free-tier access for developers.
    Free option
    Free
    Company
    Convai
    Platforms
    Unity SDK, Unreal Engine plugin, web plugins
    Key features
    AI characters, Convai Playground, Avatar Studio, Unity SDK, Unreal Engine plugin, Core API

    Convai is built around the character itself rather than the model underneath. You define who an NPC is in its Playground, drop the Unity or Unreal plugin into a project, and the character talks.

    Documentation also covers web plugins and a modding framework, which helps if a small team wants one vendor for prototype and shipping build. Because billing is metered per conversation turn, you can estimate cost from expected play sessions before committing.

    Where it falls short

    Plan amounts sit behind a JavaScript pricing page that we could not read, so cost per player cannot be checked here. Credits draw down on every turn and session, which makes budgeting harder for games with long, chatty play.

    Skip it if

    Teams that want fully scripted, offline dialogue should look at Yarn Spinner or ink, and those who need on-device inference should read the NVIDIA ACE entry.

    Sources checked for Convai
  2. 2

    NVIDIA ACE

    Speech, language and animation models for in-game characters

    Why #2: The deepest technical stack of the group, spanning speech, language and face animation, though it asks you to assemble the parts yourself.

    Best for
    Technical teams with GPU budget
    Pricing
    No price is shown on the ACE for Games page, which offers Get Started and Notify Me buttons instead.
    Free option
    Unknown
    Company
    NVIDIA
    Platforms
    Cloud and on-device; Unreal Engine 5 for Audio2Face-3D
    Key features
    Small speech, language and animation models for games, Cloud and on-device inference, NVIDIA In-Game Inferencing (NVIGI) plugins, Agentic loop and RAG guides for custom engines, Audio2Face-3D in Unreal Engine 5

    ACE bundles small speech, language and animation models that can run in the cloud or on the player's own GPU, plus In-Game Inferencing plugins that schedule model work next to rendering. The page names Total War: PHARAOH, PUBG Co-Player Characters and inZOI as partner examples, and offers agentic loop and retrieval guides for custom engines.

    A studio with graphics engineers gets the most control over latency and running cost here.

    Where it falls short

    This is a toolkit, not a character editor, so design, memory and dialogue tooling are yours to build. There is no price on the page, and on-device models only help players whose hardware can carry the load.

    Skip it if

    Small teams without engine programmers will move faster with Convai or Inworld, where the plumbing is already assembled.

    Sources checked for NVIDIA ACE
  3. 3

    Inworld

    Realtime voice models and LLM inference for consumer apps

    Why #3: Published, metered prices and a voice stack make it easy to model costs, but the pitch has moved from game characters to consumer voice.

    Best for
    Voice-heavy characters on usage billing
    Pricing
    The On-Demand plan costs USD 0 to start and bills per use (TTS-2 at USD 25 per 1M characters, open-model LLM inference from USD 0.07 per 1M tokens); a Creator plan is USD 25 per month.
    Free option
    Free
    Company
    Theai, Inc. dba Inworld AI
    Key features
    Text to speech, Speech to text, Speech to speech, LLM inference, Voice cloning and voice design, Commercial license on the on-demand plan

    Inworld now describes itself as realtime voice models plus LLM inference for consumer applications, and the pricing page spells out unit costs. That openness is rare in this niche.

    Pricing lists GDPR and SOC 2 Type II on every plan, and the on-demand plan carries a commercial license. For a studio that already has its own dialogue logic and wants a fast speech layer, it is a clear, costable pick.

    Where it falls short

    The homepage no longer leads with game characters, so NPC-specific tooling such as personalities or in-engine plugins is not visible in what we read. You will wire the voices into your own conversation logic.

    Skip it if

    If you need a ready-made NPC with an engine plugin, Convai is the closer fit; for the richest voice catalog, see ElevenLabs.

    Sources checked for Inworld
  4. 4

    ElevenLabs

    Large voice catalog with APIs, cloning and dubbing

    Why #4: Not an NPC brain, but the broadest published voice and sound tooling, with a free tier and prices anyone can read.

    Best for
    Voice acting for dynamic lines
    Pricing
    Free tier with 10k credits per month; Starter is $1 for the first month then $6; Creator $22; Pro $99; Scale $299; Business $990; Enterprise is custom.
    Free option
    Free
    Company
    ElevenLabs
    Platforms
    Web app; API with SDKs
    Key features
    Text to speech, Speech to text, Voice cloning, Dubbing, Music generation, Sound effects

    ElevenLabs supplies the voice, not the character logic. The homepage meta cites 17,500+ voices in 90+ languages, and the product list adds speech to text, voice cloning, dubbing, sound effects and music generation behind APIs and SDKs.

    For a narrative game that wants thousands of lines voiced without booking sessions, or an NPC pipeline that already has its own dialogue model, the tiered plans make the monthly bill easy to forecast.

    Where it falls short

    Credits are consumed per generation, so a game that voices lines live for many players needs a volume quote beyond the self-serve tiers. There is no game-engine plugin on the pages we read.

    Skip it if

    Studios that want the character, memory and engine plugin in one tool should read Convai; teams that only need scripted branching can skip paid voice altogether and use ink.

    Sources checked for ElevenLabs
  5. 5

    Dialogue System for Unity

    Unity dialogue asset with an OpenAI and Ollama addon

    Why #5: The practical bridge between traditional branching dialogue and language models, provided your project already lives in Unity.

    Best for
    Unity teams adding LLM chat
    Pricing
    No price appears on the opened pages; an evaluation version can be downloaded and the asset is sold through the Unity Asset Store.
    Free option
    Free trial
    Company
    Pixel Crushers
    Platforms
    Unity only
    Key features
    Dialogue System for Unity, Addon for OpenAI and Ollama models, Procedural Dialogue addon, Quest Machine addon, Evaluation version, Integrations with other assets

    Dialogue System for Unity is a conversation and quest toolkit, and Pixel Crushers extends it with an addon that calls OpenAI models and local models through Ollama. The addon page lists GPT-5, GPT-4o and Llama 3.2 and 3.3, Deepseek and Gemma.

    Authors can keep hand-written scenes where story matters and let a model handle idle chatter. The site quotes the teams behind Disco Elysium and Suzerain.

    Where it falls short

    Everything runs inside Unity, so Godot and Unreal projects are out. The OpenAI addon requires Dialogue System 2.2.58 or later, and any model usage fees come from the provider you connect, not from this asset.

    Skip it if

    Unreal or Godot developers should look at Yarn Spinner, which supports all three engines, or at Convai for its Unreal plugin.

    Sources checked for Dialogue System for Unity
  6. 6

    Yarn Spinner

    Screenplay-style dialogue language for three engines

    Why #6: Best when writers need to own the dialogue: plain-text scripts, three engines, and a long list of shipped games.

    Best for
    Writers shipping branching dialogue
    Pricing
    Free and paid versions exist for each engine, and the pages we opened show no price figures.
    Free option
    Free
    Company
    Yarn Spinner
    Platforms
    Unity, Godot (C# and GDScript), Unreal Engine
    Key features
    Screenplay-like dialogue language, Unity, Godot and Unreal support, Localisation, Voice-over support, Version control friendly, Browser trial

    Yarn Spinner lets writers draft branching conversations in a screenplay-like language, compile them, and call them from code with simple APIs. The homepage lists drafting, feedback, version control, localisation and voice-over among the chores it simplifies, and its showcase includes A Short Hike, NORCO and DREDGE.

    Godot and Unreal arrived beside Unity and Godot C#, so engine choice is less of a trap than with Unity-only assets.

    Where it falls short

    It produces authored, deterministic dialogue. No language-model chat, generated voice or character memory is described on the pages we read, so open-ended NPC talk needs another layer on top.

    Skip it if

    Teams chasing free-form, generated conversation should start with Convai or Inworld; Yarn Spinner will not improvise.

    Sources checked for Yarn Spinner
  7. 7

    ink

    Open source narrative scripting language with an editor

    Why #7: Zero cost and proven by inkle's own games, but strictly an authoring language rather than an AI character platform.

    Best for
    Text-first branching stories
    Pricing
    Free: ink, the Inky editor and the Unity plugin are released under the MIT license.
    Free option
    Open source
    Company
    inkle, United Kingdom
    Platforms
    Inky for Mac, Windows and Linux; Unity plugin; Unreal via Inkpot
    Key features
    Markup-first scripting language, Inky editor with live play pane, Export to JSON and web, Unity integration, Unreal integration through Inkpot

    ink treats text as the main material and slips logic in as markup, which keeps long branching scripts readable. inkle built it as middleware for its own releases and says millions of words have been written in it. Inky plays the story as you type and exports JSON or a web page, and the MIT license means no fee at any budget.

    It suits solo writers and small studios wanting a durable, engine-agnostic script format.

    Where it falls short

    Unity gets the official plugin; Unreal support comes from The Chinese Room's Inkpot container rather than inkle. There is no voice, generation or runtime AI, and no free-form conversation.

    Skip it if

    Anyone expecting a talking character should look at Convai or Inworld; ink only plays what a writer has authored.

    Sources checked for ink
  8. 8

    articy:draft X

    Visual database for storylines, characters and dialogue

    Why #8: Strong organization for large branching narratives, with engine exports, though it plans dialogue rather than generating it.

    Best for
    Large narrative teams on Windows or Mac
    Pricing
    A free version exists; Multi-User and Single-User licenses are sold on a pricing page whose amounts did not appear in the text we read.
    Free option
    Free
    Company
    Articy Software GmbH & Co. KG, Germany
    Platforms
    Windows and macOS
    Key features
    Nested flow view for branching, Characters and variables database, Localization, Unity and Unreal integrations, Version control and change history, API

    articy:draft X combines a nested flow view with a database of characters, variables and items, so quest designers and writers share one source of truth. The vendor lists integrations for Unity and Unreal, localization, flexible exports and an API, and describes version control with change history for remote teams.

    Where a game has hundreds of quests and thousands of lines, this kind of structure matters more than any single chat feature.

    Where it falls short

    It runs only on Windows and macOS. Amounts for single and multi-user licenses were not readable in what we opened, and no language-model or voice generation is described.

    Skip it if

    Small prototypes with a few conversations are better served by ink or Yarn Spinner, which carry no license cost.

    Sources checked for articy:draft X
  9. 9

    Scenario

    Custom-trained art models plus audio generation

    Why #9: Useful beside an NPC system for portraits and assets, though it generates content, not conversations.

    Best for
    Studios generating on-style assets
    Pricing
    Starter is $15 per month, Pro $45 and Max $75, with Enterprise on custom pricing; 50 free daily credits need no card, and annual billing saves 33 percent.
    Free option
    Free
    Company
    Scenario
    Platforms
    Web app; MCP and REST API with TypeScript and Python SDKs; integrations with Claude Code, Codex and Cursor
    Key features
    Custom model training, Image, video, 3D and audio generation, Workflows and node canvas, Agents (Genie, Node Agent), MCP and REST API, Skills

    Scenario trains custom models on a team's own art and runs image, video, 3D and audio generation through workflows and agents. The about page names Scopely, Epic Games, Supercell and Tripledot Studios.

    In an NPC pipeline it covers character portraits, props and ambient audio rather than dialogue. Plans are public, a free daily allowance lets a lead try it, and MCP and REST access allow pipeline automation.

    Where it falls short

    It does not handle speech, personalities or conversation state, so it cannot make an NPC talk. Custom model training is listed from the Pro tier upward, not on Starter.

    Skip it if

    If the need is a character that answers players, move up to Convai or Inworld; Scenario only supplies art around them.

    Sources checked for Scenario
  10. 10

    Layer

    Asset generation for game studios across 500+ models

    Why #10: Wide model choice and no seat fees, yet tier amounts are not readable and the output is assets, not NPC speech.

    Best for
    Art teams mixing many models
    Pricing
    Usage-based with unlimited members at $0 per seat; tier amounts were not in the opened text, and the pricing page links to a calculator and to sales.
    Free option
    Unknown
    Company
    Layer, United States
    Platforms
    Web app; REST API, TypeScript SDK and MCP for Claude, Claude Code and other agents
    Key features
    Image generation and editing, Video generation and lip sync, 3D asset creation and rigging, Music and sound effects, Voice cloning, REST API, SDK and MCP

    Layer says it was founded in July 2023 in San Francisco and counts 400+ studios as users. Its homepage lists 139 image models, 174 video models and 48 3D models, plus music, sound effects and voice cloning, reachable through a REST API, TypeScript SDK and MCP.

    For NPC work it can produce character art, rigged 3D bodies and voice lines, and the zero-seat price helps larger crews.

    Where it falls short

    Because pricing is usage-based and the tier numbers sit behind a calculator, you cannot compare cost with other entries here. It offers no conversation engine of its own.

    Skip it if

    Teams that primarily need talking characters should choose Convai or Inworld, and use Layer only for the surrounding art.

    Sources checked for Layer
  11. 11

    Ludo.ai

    Generator for sprites, UI, music and 3D game assets

    Why #11: An inexpensive way to fill a prototype with art and sound, listed last because it creates assets rather than dialogue.

    Best for
    Indie prototypes needing quick assets
    Pricing
    Indie is $15 per month billed annually ($180 a year, 3,000 credits), Pro $35 per month annually, Studio $300 per month annually; start free with 30 credits.
    Free option
    Free trial
    Company
    Ludo.ai
    Platforms
    Web app; API, MCP and CLI access
    Key features
    Sprite generation, 2D assets and styles, Image editor, Audio generation, 3D assets, Video pipeline

    Ludo.ai produces sprites, animations, icons, UI, textures, music, 3D models and video meant to drop into an engine. The homepage offers 30 free credits, and Pro and Studio include unlimited image generation in the web app, which suits solo developers iterating on looks.

    API, MCP and CLI calls always draw from credits, so scripted pipelines meter differently from manual use.

    Where it falls short

    Nothing on the opened pages covers dialogue, voice acting or character behavior. The Indie plan has only 3,000 credits per year and a single seat.

    Skip it if

    Studios that need an NPC to hold a conversation should start with Convai; Ludo.ai fits the art side only.

    Sources checked for Ludo.ai

What this list shows

Of the eleven tools ranked, five publish price figures we could read (Inworld, ElevenLabs, Scenario, Ludo.ai and the free ink), while NVIDIA ACE, Pixel Crushers, Layer and articy:draft X leave amounts out of the pages we opened. Convai's plans exist but render through JavaScript. Three are free or open source in some form: ink under MIT, Yarn Spinner with a free tier, and articy:draft X with a free version.

Only articy (Germany) and Layer (founded in San Francisco) state a location on the pages we read.

The list also shows how loose the label is. Only four entries produce live, model-driven speech; four more author conversations; three generate assets. Inworld now describes itself as voice and inference for consumer applications, so its place reflects what its pages say today.

The practical trade-off is control versus improvisation: authored dialogue is predictable and cheap, generated dialogue is flexible but metered and harder to moderate. Charisma.ai was left out because its web address now opens a film production studio page. Ready Player Me could not be opened.

How to choose

1. Write down what players may say

List five example exchanges you want in the game. If every one fits a fixed choice, an authoring tool such as Yarn Spinner or ink is enough and costs nothing in usage fees. If players will type or speak free text, you need a model-backed option like Convai or NVIDIA ACE, and you should add moderation to your plan.

2. Match the tool to your engine

Open the plugin list for your engine first. Unity teams can use almost anything here; Unreal teams should check Convai, NVIDIA ACE, Yarn Spinner and ink; Godot teams have Yarn Spinner. Install the free or trial version, and see how long it takes to get one character talking in a blank scene.

3. Price one hour of play

Multiply spoken lines per hour by your expected players, then apply each vendor's unit price: characters for speech, tokens for language, credits for turns. Compare that figure to a flat license. Where the vendor shows no price, ask for a written quote before building on the service.

Frequently asked questions

What is the best AI NPC and dialogue platform in 2026?

Convai is the best fit for most teams, because it packages characters, voices and Unity and Unreal plugins with a free tier. NVIDIA ACE is better when you have graphics engineers and want models on the player's GPU. Inworld is the better choice when you only need a voice layer with clearly published unit prices.

Can an NPC use a language model without a cloud service?

Yes, in some setups. NVIDIA ACE describes cloud and on-device models, and the Pixel Crushers OpenAI addon says it works with local models run through Ollama, such as Llama 3.2 and 3.3 and Gemma. Local inference depends on the player's hardware and adds memory use.

Is branching dialogue still worth using next to AI?

Yes. Yarn Spinner and ink give writers exact control over plot-critical scenes, and ink is free under the MIT license. Many studios keep story beats authored and add model-driven chatter for minor characters, which Dialogue System for Unity supports through its addon.

Which tools here work in Godot?

Yarn Spinner is the one entry with Godot support in both C# and GDScript. Convai lists Unity, Unreal and web plugins instead. For other options, check that a vendor documents a Godot integration before committing, and consider ink's plain JSON export.

How much do AI voices for game characters cost?

Published figures vary. Inworld lists TTS-2 at USD 25 per 1M characters. ElevenLabs runs from a free 10k-credit tier through $22 Creator and $99 Pro plans up to Enterprise quotes.

Convai bills credits per turn, and NVIDIA ACE shows no price.

Do asset generators like Scenario or Layer make NPCs talk?

No. Scenario, Layer and Ludo.ai generate art, 3D, music and sound effects, and some list voice cloning, but they do not run a conversation. Use them to produce portraits and props, then connect a conversational tool for speech.

What should I check before shipping AI dialogue to players?

Confirm the commercial license, data handling terms and moderation controls. Inworld states GDPR and SOC 2 Type II on its plans. Read each vendor's terms on generated content, and plan fallbacks for when a service is slow or offline.

Talk it through

Rankings are a starting point. The discussion thread for this ranking is open to anyone with a GA Forum account.