Game characters and talking avatars get lumped together, but they're different jobs: one lives inside a game engine, the other on a web page or in a training scenario. Game teams need dialogue, speech and animation that run in real time. Trainers and product teams usually want a talking face they can embed or share. Settle that first, then test two tools with one script.
Plenty of teams start with Convai because it puts talking AI characters inside virtual worlds, with plugins for the big game engines and uses in XR training and social spaces. If you're comparing it or moving on, it helps to see that the field has split. Some tools still build characters for games, some build digital humans for training and service, and a growing group sells live talking avatars by API. The nine below cover all three directions.
Nine options side by side
Start here if you just want the shape of the field. The groups and the A to Z list further down explain who each one is for.
| Tool | Kind of tool | Who it's aimed at |
|---|---|---|
| Anam | Real-time avatar API | Developers adding a live avatar to a product |
| Beyond Presence | Enterprise real-time avatars | Companies sharing an avatar agent by link |
| D-ID | AI avatars and agents | Turning existing video into something interactive |
| Inworld AI | Voice AI and game dialogue platform | Voices and live dialogue in games |
| NVIDIA ACE for Games | Game character AI toolkit | Studios building in-game characters |
| Simli | Real-time avatar API | Trying talking characters before building |
| Soul Machines | Digital people and AI assistants | AI assistants that need a face |
| Tavus | Real-time video agents | Conversational video agents, no code needed |
| UneeQ | Digital humans for training | Roleplay practice for staff |
Where the talking-character field has moved
Names you remember from a couple of years of game demos may not mean what they did. Charisma.ai, once a familiar name in interactive AI characters, now redirects to Charismatic AI Studio, which describes itself as an AI film studio. Inworld's homepage now leads with voice products rather than character tools, with text to speech and speech recognition up front and games and media as one offering among several.
So the first thing to test is simple: open each vendor's current homepage and read what it says it does today. If a tool you'd shortlisted now talks mostly about film, voice or customer service, ask directly whether character work in a game engine is still something they support and plan to keep supporting. A plugin that stops being updated is far harder to deal with than a shortlist that's one name shorter.
Game characters, digital humans and avatar APIs
Characters inside a game
NVIDIA ACE for Games is the clearest match for in-engine characters, with models for speech, intelligence and animation built to run on the player's device. Inworld AI fits teams whose main need is voices and live dialogue for games and narrated media.
Digital humans for training and service
Soul Machines builds AI assistants with a face through its Studio product. UneeQ has narrowed its focus to roleplay training and coaching, which makes it a close match for XR or simulation teams building practice scenarios for staff.
Video agents and avatars
Tavus lets teams build conversational video agents with no code or through an API. D-ID combines video generation with an avatar agent product and a real-time streaming API.
Real-time avatar APIs
Anam, Simli and Beyond Presence all offer real-time avatars you call from your own product. The differences sit in how you start: one leans on its own avatar video models, one lets you talk to sample characters on its homepage, and one lets you embed an agent or share it by link.
A ten-line scene for two tools
Write one short scene your character must handle, around ten lines of dialogue with one question it can't answer. Then run it through two tools from the group that matches your project, and judge each on the same points:
- How long the character takes to start talking after the user speaks
- Whether lip movement and expression stay in step with the voice
- What it does with the question it shouldn't answer
- How hard it was to get the scene running in your engine or on a page
- Whether you can change the script without rebuilding everything
All nine, A to Z
Alphabetical, and in no particular order beyond that. Each line reflects how the maker describes its own product.
Anam
Best for: Developers who need real-time interactive avatars by API, built on the vendor's own avatar video models, with low latency as its stated focus.
Beyond Presence
Best for: Enterprise teams building avatar agents in a studio, then embedding them on a site or sharing them with a simple link.
D-ID
Best for: Content teams with existing video who want, in the vendor's words, to transform passive content into interactive AI experiences, with a streaming API too.
Inworld AI
Best for: Game and media teams needing voices and live dialogue, with text to speech, speech to text and speech to speech on one developer platform.
NVIDIA ACE for Games
Best for: Middleware and game developers wanting small speech, intelligence and animation models for in-game characters, tuned to run on the player's own device.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Simli
Best for: Builders who'd like to talk to sample characters on the homepage before reading the docs and wiring real-time avatars into their own app.
Soul Machines
Best for: Brands wanting an AI assistant with a face, built through the vendor's Studio product for digital people.
Tavus
Best for: Teams that want conversational video agents, which the vendor calls PALs, built without code or deployed through its developer API.
UneeQ
Best for: Learning and development teams running repeatable roleplay with a digital human, so staff can practise real workplace conversations and build confidence.
Engine or screen? Advice before you build
Decide early whether you need a character that lives inside your engine or a face that lives on a screen. Those are different products now, and a lot of wasted evaluation time comes from mixing them. Keep your scene, your knowledge files and your voice choices in your own hands, so moving between tools later means reconnecting rather than rewriting.
For the rest of a game project, my guide to AI tools for game development covers more of the build, and if you need art as well as dialogue, see my list of AI game asset generators.
Do you make AI character, voice or avatar tools? Tool makers can ask to be considered for a featured place on this list.
Want your tool on this list? Featured spots and reviews here are paid partnerships. Here's how to partner with me.
Frequently asked questions
What's the difference between an AI character and an AI avatar?
An AI character usually lives inside a game or virtual world and reacts to the player. An AI avatar is usually a talking face on a web page or in an app. Several tools here sit on one side or the other, so match them to where your character will appear.
Can these tools run inside Unity or Unreal Engine?
Some are built for game developers, others are delivered through a web API or embed. Ask each vendor what engine support they offer today and test a small scene before you commit.
Which kind suits XR training?
Tools focused on digital humans and roleplay practice are the closest fit, because they're built around repeatable conversations. Test with a real training scenario, not a demo script.
Do I need to write code?
Not always. Some tools offer no-code builders or a shareable link, while others are APIs for developers. Pick based on who'll maintain the character after launch.
How do I keep a character from saying something wrong?
Give it a tight script and a limited knowledge source, then test the awkward questions on purpose. Check what controls each tool offers for limiting answers.
