The company many developers knew for talking game characters now leads its homepage with realtime voice models and inference for consumer apps. So the alternative you need follows from what you're moving: a character engine for game studios, a voice model or voice agent platform for voice apps, or a digital human when users need a face. No single platform here covers all three.
If you built on Inworld AI for game characters, you may have noticed its homepage now talks about text to speech, speech to text and speech to speech instead. That leaves studios and app teams replacing different halves of the same product. The nine platforms below are grouped by the job each one does best, so you can replace the part you used without rebuilding the rest.
Four kinds of platform, sorted by job
Realtime voice and conversational AI has split into layers, and most vendors now sell one layer rather than the whole thing.
Voice agents you build and deploy
ElevenLabs lets teams build, test and deploy voice and chat agents for a website, an app or a call centre. Vapi is a platform for developers to build, test and deploy voice AI agents, and LiveKit offers an open source framework plus a cloud platform for voice, video and physical AI agents. These are where to look when the agent itself is the product.
Voice models underneath
Cartesia builds realtime voice AI, from text to speech and speech to text through to managed voice agents. Hume AI works with labs and enterprises to evaluate and improve emotionally intelligent voice AI. These sit lower in the stack, closer to the models than to a finished app.
Digital humans with a face
Soul Machines lets teams create and deploy face-to-face AI agents that appear as digital humans. UneeQ uses digital humans for AI roleplay training and coaching, with feedback after each session, so it's aimed at learning teams rather than games.
Characters for games and virtual worlds
The character engine in this list is covered in the next section, alongside the change that sends many developers looking in the first place.
Characters and voice now live in different places
The product you're leaving now sells voice models and inference for consumer apps, not a game-character product, at least not on its homepage. So anyone who came for game characters has to compare voice-model vendors and character engines as two separate decisions.
Convai, by contrast, still sells conversational characters for games, XR and virtual worlds, with plugins for Unity and Unreal Engine. Meanwhile Deepgram's page announces that Daily and Pipecat are joining it, so the voice infrastructure side is shifting too. The first thing to test is whether the plugin or SDK you'd rely on is current for your engine version, and whether the voice layer you pick will still talk to it after any of these changes settle.
It's also worth writing down which parts of your current build depend on the old vendor directly. Dialogue, memory, voice and animation may each need a different replacement, and it's far easier to plan that up front than to discover it halfway through a sprint.
Nine platforms in one table
| Tool | Kind of platform | Who it suits |
|---|---|---|
| Cartesia | Realtime voice models | Realtime speech in and out |
| Convai | Game and virtual-world characters | Unity and Unreal Engine games |
| Deepgram | Voice AI APIs | Speech APIs under one platform |
| ElevenLabs | Voice agents and voice generation | Website, app and call centre agents |
| Hume AI | Emotion-aware voice AI | Labs and enterprise voice teams |
| LiveKit | Open source agent framework | Developers wanting an open source base |
| Soul Machines | Digital human agents | Face-to-face AI agents |
| UneeQ | Digital human training | Roleplay training and coaching |
| Vapi | Voice agent platform | Builders shipping voice agents |
Every platform, alphabetically
In no particular order beyond the alphabet. Each line is a short description based on what the vendor says about its own platform.
Cartesia
Best for: Teams that need realtime text-to-speech, speech-to-text or managed voice agents from a single voice AI provider.
Convai
Best for: Game, XR and virtual-world developers adding conversational AI characters through plugins for Unity and Unreal Engine.
Deepgram
Best for: Developers who want speech-to-text, text-to-speech and Voice Agent APIs together; the vendor calls it the whole voice AI stack in one platform.
ElevenLabs
Best for: Building, testing and deploying voice and chat agents for a website, app or call centre, agents the vendor says feel human.
Hume AI
Best for: Labs and enterprises evaluating and improving emotionally intelligent voice AI.
LiveKit
Best for: Developers who'd rather start from an open source framework, with a cloud platform for voice, video and physical AI agents.
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.
Soul Machines
Best for: Face-to-face AI agents that appear as digital humans, with the vendor promising deeply human emotion.
UneeQ
Best for: Training teams running AI roleplay and coaching with digital humans, with feedback after each session.
Vapi
Best for: Builders who want one platform to build, test and deploy voice AI agents.
Splitting the decision: advice for studios and app teams
Split the decision in two. Choose the character or agent layer for the job your players or users see, then choose the voice layer for how it sounds and how fast it responds. Teams that try to replace everything with one vendor often end up rebuilding later, so keep the pieces loosely joined and test each one on its own.
If you're building games, my guide to the best AI tools for game development covers the wider toolchain, and these AI game asset generators help with the art side. For quick prototypes, there's a list of each free AI game generator worth trying too.
One line of dialogue, two platforms
Here's a small test that tells you a lot. Write one short exchange your character or agent has to handle, such as a player asking for directions or a customer asking to change a booking. Pick two platforms from the same group above and build that single exchange in each, with the same prompt, the same voice style and the same expected reply.
Then listen. How long is the pause before it speaks? Does the voice hold up when the player interrupts? Can you change the line without starting over? Run it on the device your users will have, not just your desk machine, and keep a note of what you had to work around. Whichever one needs the fewest workarounds is the one to take into a bigger build.
Do you make voice AI or conversational character 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 a voice model and a voice agent platform?
A voice model turns text into speech or speech into text. A voice agent platform wraps models into an agent that listens, decides and replies, so you build less yourself.
Can I still build talking game characters with these tools?
Yes. One platform here sells conversational characters with Unity and Unreal Engine plugins, and you can pair voice models with your own character logic if you'd rather build it.
Do I need a digital human or just a voice?
Only use a face if users need to see one, such as in training or face-to-face service. For games and phone agents, a voice is usually enough.
Is open source an option for voice agents?
One platform here offers an open source framework with a cloud platform beside it, which suits teams that want to see and change the code they run on.
How should a small studio test these platforms?
Build one short exchange in two platforms, test it on real devices, and judge the delay, the voice and how easy each one makes editing before you commit to a bigger build.
