# Inworld AI

> Source: https://aiwiki.ai/wiki/inworld_ai
> Updated: 2026-10-30
> Fact-checked: 2026-10-30
> Categories: AI Companies, AI in Gaming, Conversational AI, Speech & Audio AI, Voice AI
> License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) - attribute to "AI Wiki (aiwiki.ai)"
> Cite as: AI Wiki. "Inworld AI." aiwiki.ai, 30 Oct 2026. https://aiwiki.ai/wiki/inworld_ai
> From AI Wiki (https://aiwiki.ai), the free encyclopedia of artificial intelligence. Reuse freely with attribution.

Inworld AI is an American [artificial intelligence](https://aiwiki.ai/wiki/artificial_intelligence) company headquartered in Mountain View, California, that describes itself as a research lab and inference provider for consumer AI, selling realtime voice models and model inference to developers.[^1][^5] Founded in 2021 by Ilya Gelfenbeyn, Kylan Gibbs, and Michael Ermolenko, the company initially focused on creating AI-driven non-player characters (NPCs) for video games and virtual environments through its Character Engine.[^2] It has since retired its legacy Character Studio, the web tool for authoring those characters, and directs developers to its Agent Runtime instead.[^50] Inworld then moved into voice [AI infrastructure](https://aiwiki.ai/wiki/ai_infrastructure), launching a standalone [text-to-speech](https://aiwiki.ai/wiki/text-to-speech_models) API in 2024, the TTS-1 models in June 2025, the Inworld Runtime orchestration engine in August 2025 and a public inference service in October 2026.[^51][^52][^53][^54] Its flagship model, Realtime TTS-2, entered research preview in May 2026 and general availability on 31 August 2026.[^39][^4] Inworld says TTS-2 launched at first place on the [Artificial Analysis](https://aiwiki.ai/wiki/artificial_analysis) Controlled Voice leaderboard; on 30 October 2026 it ranked seventh there and eighth on Artificial Analysis's main text-to-speech leaderboard.[^5][^55][^56] Inworld lists Xbox, NVIDIA, Ubisoft, Niantic, NBCUniversal and Streamlabs among partners that have built with it, and consumer apps such as Talkpal, Bible Chat and Status by Wishroll among its voice customers.[^3][^1]

<figure class="w-56">
<img src="/images/site-2026-10-28-inworld_ai-logo-5f37583c2bc4-tile.png" alt="Inworld AI logo" width="480" height="146" />
<figcaption>Inworld AI logo, from the official website <a href="https://inworld.ai/">inworld.ai</a>. Trademark of its owner.</figcaption>
</figure>

## History

### Founders and the API.AI legacy

The founding team shares deep roots in conversational AI infrastructure. Ilya Gelfenbeyn co-founded API.AI, a developer platform for [natural language understanding](https://aiwiki.ai/wiki/natural_language_understanding) and [speech recognition](https://aiwiki.ai/wiki/speech_recognition), and served as its chief executive officer.[^6] API.AI grew to serve millions of developers before [Google](https://aiwiki.ai/wiki/google) acquired it in September 2016; following the acquisition, Google rebranded the platform as Dialogflow, one of the most widely deployed conversational AI frameworks in the industry.[^6] Gelfenbeyn also built Speaktoit (later branded Assistant.ai), an independent [voice assistant](https://aiwiki.ai/wiki/voice_assistant) that attracted more than 40 million users.[^6] After the acquisition, he joined Google's developer ecosystem programs before departing to co-found Inworld.

Michael Ermolenko led AI engineering at API.AI before and after the Google acquisition, accumulating direct experience building large-scale conversational pipelines.[^7] Kylan Gibbs took a different path: he worked as a product manager at [Google DeepMind](https://aiwiki.ai/wiki/google_deepmind) and earlier as a consultant at Bain and Company.[^8] Gibbs also co-founded FlowX, a startup later acquired before his departure to join Gelfenbeyn and Ermolenko.[^8] Gibbs became chief executive officer in January 2024, with Gelfenbeyn transitioning to executive chairman.[^9][^2]

John Gaeta, an Academy Award-winning visual effects designer best known for pioneering the "bullet time" technique in The Matrix, joined as Chief Creative Officer effective June 2022.[^10] According to Contrary Research, Gaeta held that role until April 2024 and then became a strategic advisor, a role he still held as of June 2025.[^2]

### Founding vision (2021)

Gelfenbeyn, Gibbs, and Ermolenko founded Inworld AI in 2021 with a specific thesis: virtual environments, including video games, online worlds, and nascent metaverse platforms, were rapidly increasing in the amount of time users spent inside them, yet those environments remained socially shallow.[^2] NPCs in most games still relied on branching dialogue trees written by hand, offering limited interactivity and no capacity to respond dynamically to player actions or language. The founders believed that [generative AI](https://aiwiki.ai/wiki/generative_ai) had reached a threshold where AI-driven characters could replace scripted dialogue with contextual, personality-consistent conversation.

The company's initial product, the Character Engine, was built around what Inworld called the "Character Brain" and "Contextual Mesh" and orchestrated multiple [machine learning](https://aiwiki.ai/wiki/machine_learning) models intended to mimic the full range of human communication.[^11] Inworld's website in November 2023 described three parts: "Real-time AI" (low-latency interactions, orchestrated across multiple LLMs), the Character Brain (multimodal expression covering conversation, voice, animations and emotions) and the Contextual Mesh (custom knowledge, safety guardrails and narrative controls meant to keep characters "in-character and in-world").[^51] Developers configured a character's personality by describing it in natural language rather than writing explicit dialogue logic.[^51]

In November 2021, the company announced a $7 million seed round co-led by [Kleiner Perkins](https://aiwiki.ai/wiki/kleiner_perkins) and CRV, with participation from Meta, to fund the initial development of the Character Engine.[^12]

### Early adoption and the Disney Accelerator (2022)

Inworld joined the Disney Accelerator program in 2022, gaining access to resources and strategic relationships within Disney's entertainment portfolio.[^13] At the 2022 Disney Accelerator Demo Day, the company presented a "Droid Maker" prototype built in collaboration with ILMxLAB (now ILM Immersive), the immersive storytelling studio within Lucasfilm.[^14] The demo allowed users to assemble and converse with interactive Star Wars droids, illustrating how AI characters could extend major intellectual property into interactive experiences.[^14]

During the same period, community modders began integrating Inworld's API into popular titles including The Elder Scrolls V: Skyrim, Stardew Valley, and Grand Theft Auto V.[^15] These unofficial mods attracted significant attention online and demonstrated consumer appetite for dynamic NPC conversation, while also functioning as organic demonstration of the platform's versatility. The Stardew Valley integration, for example, gave 33 villagers AI "brains" through which players could hold real-time conversations rather than triggering scripted lines.[^15]

### Expansion into AAA and enterprise partnerships (2023 to 2024)

At the Game Developers Conference (GDC) in 2024, Inworld appeared alongside several major industry names. Ubisoft presented NEO NPCs, a research prototype built using the Inworld Engine that showed NPCs capable of environmental awareness, real-time reaction and animation, conversation memory, and collaborative decision-making with other characters.[^16] The demo represented Ubisoft's first public-facing generative AI NPC prototype and was developed in conjunction with NVIDIA's Avatar Cloud Engine (ACE) technology, with NVIDIA Audio2Face providing real-time facial animation.[^16]

[Microsoft](https://aiwiki.ai/wiki/microsoft)'s Xbox division announced a multi-year co-development partnership with Inworld in November 2023 to build AI-assisted tools for game narrative creators.[^17] One output of this collaboration, the Narrative Graph, allows developers to upload source material and generate branching narrative structures that visualize story logic.[^18] Another tool, Project Explora, extended narrative AI into broader game design workflows.[^19] At GDC 2024, Inworld and Microsoft also demonstrated a text-based prototype called Mists of Aurora that illustrated the Narrative Graph's capabilities.[^18]

NVIDIA integrated Inworld's character intelligence into its Covert Protocol demo at GDC 2024, a social-simulation experience built on Unreal Engine 5 where players acted as private detectives completing objectives through conversation with AI digital humans.[^20] The demo combined [NVIDIA Riva](https://aiwiki.ai/wiki/nvidia_riva) for speech recognition and NVIDIA Audio2Face for facial animation alongside Inworld's inference pipeline, and Inworld committed to releasing the demo's source code.[^20]

### Pivot to voice AI platform (2024 to 2025)

From 2024 Inworld moved from a character-creation platform for games toward real-time voice AI infrastructure. In May 2024 its website announced that "Inworld's AI voices are now available in a standalone text-to-speech API".[^51] Inworld's own account is that, as partners moved from demos to live titles, it "hit the same wall" of keeping voice, timing and consistency reliable at scale, and that consumer apps in health and fitness, learning and social later approached it with the same problems.[^3][^53]

Inworld announced the Inworld Runtime on 13 August 2025, describing it as an AI runtime for consumer applications that grew out of infrastructure it had built for its own partners.[^53] The Runtime is a C++ graph engine that orchestrates LLMs from multiple providers ([OpenAI](https://aiwiki.ai/wiki/openai), [Anthropic](https://aiwiki.ai/wiki/anthropic), Google, Mistral), text-to-speech, speech-to-text, memory and tools.[^53][^3] Inworld's TTS also integrates with voice agent frameworks including LiveKit, [Pipecat](https://aiwiki.ai/wiki/pipecat), Vapi, and NLX.[^22]

### Consumer AI focus (2026)

In June 2026 Inworld said it was cutting prices "in half or more for most developers across the whole stack".[^57] On 30 September 2026 it announced that it had acquired Ultravox, a platform for building real-time voice agents; GeekWire reported that Ultravox, a Seattle startup founded in 2022 as Fixie, was bought for an undisclosed sum.[^58][^59] On 14 October 2026 Inworld opened Inworld Inference, which it calls "the first inference service designed for consumer AI", to every developer.[^54]

## Funding

Inworld raised a $7 million seed round in November 2021 co-led by Kleiner Perkins and CRV with participation from Meta.[^12] In March 2022 it raised $10 million in a round led by Bitkraft Ventures with participation from The Venture Reality Fund; angel investors included Twitch co-founder Kevin Lin, Animoca Brands co-founder Yat Siu and Oculus co-founder Nate Mitchell.[^60] Intel Capital later put the pre-seed and seed rounds at nearly $20 million in total.[^23] The company announced a $50 million Series A in August 2022 co-led by Section 32 and Intel Capital, with participation from [Founders Fund](https://aiwiki.ai/wiki/founders_fund), First Spark Ventures, Kleiner Perkins, BITKRAFT Ventures, CRV, Microsoft's M12 fund, Micron Ventures, LG Technology Ventures, SK Telecom [Venture Capital](https://aiwiki.ai/wiki/venture_capital), The Venture Reality Fund, and HTC.[^23] The round brought total funding to approximately $70 million at the time.[^23]

In August 2023, Inworld announced new funding from Lightspeed Venture Partners that brought its valuation to over $500 million and its total funding to more than $100 million; the announcement called it the best-funded startup at the intersection of AI and gaming.[^24] GamesBeat reported the round as $50 million, and TechCrunch reported that Lightspeed would invest more than $50 million at a $500 million post-money valuation.[^25][^44] The round included participation from [Stanford University](https://aiwiki.ai/wiki/stanford_university), Samsung Next, Microsoft's M12 fund, First Spark Ventures (co-founded by former Google CEO Eric Schmidt), and LG Technology Ventures.[^24]

Inworld's website states that it has raised more than $125 million.[^1] In October 2026, GeekWire, citing PitchBook, reported that Inworld had raised $119.8 million since 2021 and was valued at $515 million.[^59] The company's investor base has included [Lightspeed Venture Partners](https://aiwiki.ai/wiki/lightspeed_venture_partners), Kleiner Perkins, Founders Fund, CRV, Intel Capital, Meta, Microsoft M12, Samsung Next, Stanford University, LG Technology Ventures, and Bitkraft Ventures, among others.[^1][^23][^24]

## Character Engine for games (2021-2025)

Inworld's original business was the Character Engine and the web-based Character Studio for authoring game characters. Archived copies show Studio online in June 2025 and redirecting to Inworld's newer developer platform by September 2025.[^61] Inworld's documentation now states that "the legacy Inworld Character Studio has been retired" and that Studio's characters, scenes, knowledge, triggers, player profiles and voice studio are no longer available; those workflows moved to the Inworld Portal and the Agent Runtime.[^50]

### NPC dialogue and behavior

Inworld's original application was enabling game developers to build NPCs with dynamic dialogue and behavioral autonomy. Traditional NPC dialogue relies on pre-authored trees: writers compose every possible exchange, and the game engine navigates those branches based on player choices. This approach scales poorly with narrative complexity, produces rigid interactions that players can exhaust, and cannot respond coherently to player input that deviates from anticipated paths.

The Character Engine replaced static trees with generated dialogue, orchestrating multiple [large language model](https://aiwiki.ai/wiki/large_language_model)s and other models, and gave NPCs memory and recall, emotional intelligence and goals and motivations of their own.[^11][^51] In July 2023 Inworld added Contextual Mesh features (configurable safety, relationships, player profiles and a "4th Wall" setting) that it said confined characters' knowledge of the world, the player and their relationships, "preventing potential hallucinations".[^26]

The engine supported multi-agent conversations in which two to five AI characters talked among themselves or with the player in a group; when the player addressed the group, Inworld selected the character that responded.[^62]

### Unreal Engine and Unity integration

Inworld distributed SDKs for Unity, Unreal Engine, the web, Node.js and a native C++ development kit, plus a Godot SDK that its documentation described in 2024 as "built and maintained by the developer community", supporting Godot 4.2.1 on Windows and Mac.[^30] For Unreal, Inworld offered an "AI NPC Engine" plugin on the Unreal Engine Marketplace (listed February 2023; titled version 1.5 in its last listing); by September 2025 the listing had not been migrated to Fab and was available only from buyers' Epic Games Launcher vaults.[^27] The Unity SDK was offered in full and core versions.[^28]

Runtime-based toolkits replaced these SDKs. Inworld released the Unreal AI Runtime SDK on 29 October 2025, with over 100 models including LLMs, STT and TTS under a single API key, built-in observability and templates for Character, Metahuman and Lipsync workflows, and opened early access to a Unity AI Runtime SDK the same day.[^29] As of October 2026, Inworld's Studio migration guide points developers to Character templates in the Unity AI Runtime SDK and the Unreal Agent Runtime SDK, and to a Node.js voice agent template.[^50]

### Notable gaming customers and deployments

**NetEase Games:** Inworld worked with NetEase's Team Miaozi studio to build fully AI-controlled NPCs that respond in real time in playable game builds. NetEase also integrated Inworld's character AI into Cygnus Enterprises, an action RPG, as a generative AI-powered companion named PEA (Personal Electronic Assistant).[^31]

**Niantic:** The creator of Pokemon Go used Inworld AI to power Wol, an [augmented reality](https://aiwiki.ai/wiki/augmented_reality) experience inspired by the redwood trees of Muir Woods that lets visitors converse with an interactive Northern Saw-whet owl character representing the redwood ecosystem.[^32] The project, created in partnership with Liquid City and 8th Wall, used more than 30 generative AI models and won a 2024 Webby Award for Best Use of Augmented Reality, AI, Metaverse, or Virtual.[^32]

**Ubisoft:** As described above, Ubisoft's NEO NPC prototype showcased at GDC 2024 used Inworld's engine.[^16] The prototype explored NPCs with environmental awareness, real-time emotional reactions, and inter-character strategic collaboration.

**Xbox (Microsoft):** A multi-year co-development agreement produced the Narrative Graph tool and Project Explora, AI-assisted authoring tools for game writers.[^17][^18]

**NVIDIA:** The Covert Protocol demo at GDC 2024 combined NVIDIA ACE hardware acceleration with Inworld's character AI and was used at industry events to illustrate next-generation social simulation.[^20]

**Indie and community titles:** Inworld-powered mods for Skyrim, Stardew Valley, and Grand Theft Auto V attracted community attention.[^15] The title Vaudeville, an indie murder mystery game from Bumblebee Studios, gained Steam traction in 2023 using Inworld's dialogue system after streamers including MoistCr1TiKal, HasanAbi, and Videogamedunkey featured it.[^33]

**Death by AI:** An AI-native Discord game from Playroom, a six-person startup within Little Umbrella. According to Inworld's case study, it had 700,000 daily users three days after launch, 10 million players in its first month and 20 million three months later; after switching from OpenAI and ElevenLabs, Playroom integrated an Inworld Custom LLM Service API, and its CEO is quoted as saying "We found a path to profitability with Inworld".[^34]

**Status by Wishroll:** A social simulation game in which players live as anyone they imagine on a simulated social network. Inworld's case study says Status passed 500,000 daily active users after its February 2025 full release and cut AI costs by more than 95%.[^35] Inworld also says Status reached one million users in 19 days.[^1]

## Voice AI products

### Inworld TTS-1 and TTS-1-Max

In June 2025, Inworld launched Inworld TTS-1 through its API and released the larger TTS-1-Max as experimental; the models were later renamed Realtime TTS 1 and Realtime TTS 1-Max.[^52][^36] Both are transformer-based [autoregressive model](https://aiwiki.ai/wiki/autoregressive_model)s built on LLaMA backbones and trained using a sequential pipeline of pre-training, [supervised fine-tuning](https://aiwiki.ai/wiki/supervised_fine-tuning), and [reinforcement learning](https://aiwiki.ai/wiki/reinforcement_learning) alignment.[^37]

TTS 1-Max employs LLaMA-3.1-8B as its speech [language model](https://aiwiki.ai/wiki/language_model) (SpeechLM) backbone, yielding approximately 8.8 billion parameters.[^37] The architecture uses X-codec2, an audio codec that merges acoustic and semantic information into a single codebook of 65,536 tokens.[^37] The model's vocabulary was expanded from the LLaMA base of 128,256 tokens to 193,856 tokens, incorporating audio tokens and special control tokens.[^37] The audio decoder converts generated tokens back into 48 kHz waveforms via a super-resolution module.[^37]

TTS 1, the smaller variant, was designed for real-time synthesis and on-device use cases, with approximately 1.6 billion parameters.[^37] Both models support zero-shot voice cloning from a few seconds of reference audio, without fine-tuning or additional training.[^36]

The training pipeline involved large-scale pre-training on over one million hours of raw audio mixed with text data, supervised fine-tuning on 200,000 hours of high-quality filtered audio-text pairs, and reinforcement learning alignment using Group Relative Policy Optimization (GRPO).[^37]

In blind human evaluation at launch, TTS 1-Max achieved win rates of 59.1% against ElevenLabs, 60.9% against [Cartesia](https://aiwiki.ai/wiki/cartesia), 55.3% against TTS 1, and 60.7% against OpenAI TTS-1-HD.[^36] These benchmarks reflected pairwise comparisons where human raters chose their preferred sample without knowing which model produced it.

At launch, supported languages in production included English (all major accent variants), Mandarin Chinese, Korean, Dutch, French, and Spanish. Japanese, German, Italian, Polish, and Portuguese were available in experimental status.[^36] Inworld discontinued TTS-1 and TTS-1-Max on 15 June 2026 and routes requests for them to newer models.[^63]

### Realtime TTS 1.5 and 1.5-Max

In January 2026, Inworld updated the TTS line with the 1.5 generation, introducing what the company described as approximately 30% greater expressiveness over TTS 1.[^38] The 1.5 generation came in two variants:

- **Realtime TTS 1.5 Mini**: Optimized for minimum latency, with P90 time-to-first-audio under 130 milliseconds, according to Inworld's release notes.[^64] Priced at $5 per million characters.[^38]
- **Realtime TTS 1.5 Max**: Higher expressiveness with P90 time-to-first-audio under 250 milliseconds. Priced at $10 per million characters.[^38]

Both variants expanded language support to 15 languages, adding Hindi, Arabic, and Hebrew to the lineup.[^38][^64] As of October 2026, Inworld's documentation lists both 1.5 models as deprecated.[^63]

### Realtime TTS-2

In May 2026, Inworld launched Realtime TTS-2 as its new flagship voice model, available initially as a research preview through the Inworld API and Inworld Realtime API; it reached general availability on 31 August 2026.[^4][^39] TTS-2 represents an architectural advance beyond the TTS 1.x generation by incorporating closed-loop audio context: the model takes the audio of the prior turns of a conversation as input, not just the text of the current turn.[^4] Inworld says this lets TTS-2 pick up the user's tone, pacing and emotional state and adjust its delivery.[^39]

Developers can steer TTS-2 using natural language instructions rather than discrete emotion tags, for example a bracketed direction such as "[speak sadly, as if something bad just happened]", in the way developers prompt an LLM.[^4] Inline non-verbal markers such as [laugh], [sigh] and [breathe] are placed in the text where the sound should occur.[^39]

TTS-2 maintains a consistent voice identity across more than 100 languages with on-the-fly language switching within a single generation.[^39] Inworld reports sub-200ms first-chunk latency for TTS-2.[^39] Three stability modes ship with TTS-2: Expressive, Balanced (default), and Stable.[^39] In August 2026 Inworld added Realtime TTS-2 Flash; its documentation lists 20 ms P90 time to first audio byte, measured server-side, against 100 ms for TTS-2, and support for 200+ languages and locales for both models.[^64][^63]

At launch Inworld listed availability through [Cloudflare](https://aiwiki.ai/wiki/cloudflare), [DeepInfra](https://aiwiki.ai/wiki/deepinfra), GMI Cloud, LiveKit, Stream and VoiceRun, and quoted executives from Latitude, k-ID, Isekai Zero, Luvu and Talkpal.[^39] The launch post said that the previous model, Realtime TTS 1.5, already ranked first on the Artificial Analysis Speech Arena, ahead of Google and ElevenLabs.[^39][^4]

### Voice cloning

Voice cloning is a central capability across Inworld TTS generations. As of October 2026 the platform offers two modes:

**Instant cloning**: Zero-shot cloning, available to all users through the Inworld Portal or API, from as little as 3 seconds of audio up to a 30-second limit (15 seconds for voices cloned before 15 August 2026).[^65] Inworld made zero-shot cloning free for all users when it launched TTS-1.[^36]

**Professional cloning**: In beta as of October 2026. It trains a clone from up to 50 audio samples totaling at least 10 minutes, and training typically takes 30 to 60 minutes.[^65] The On-Demand pricing tier includes two professional voice clones.[^49]

Inworld states that all synthesized audio from its platform carries an imperceptible watermark identifying it as AI-generated, and that it has safeguards against cloning voices without explicit consent.[^36]

Inworld's cross-lingual voice cloning feature, introduced with TTS-2, preserves a speaker's voice identity when switching output languages, allowing a single voice definition to serve global deployments without the speaker needing to record in each target language.[^39]

### Inworld Runtime

Inworld announced the Inworld Runtime, which its documentation also calls the Agent Runtime, on 13 August 2025.[^53][^50] Implemented in C++ with SDKs for Node.js and Unreal Engine, the Runtime is a graph-based engine that connects LLMs, speech-to-text, text-to-speech, memory or knowledge, and tools into a single pipeline.[^3]

Key capabilities described at launch include:

- **Provider-agnostic LLM access**: Models from Anthropic, Google, Mistral and OpenAI, plus open models such as DeepSeek, Llama and Qwen through Groq, Tenstorrent and Fireworks AI.[^53]
- **Remote configuration without redeployment**: Remote configuration, A/B variants through a Graph Registry, and built-in telemetry covering latency paths, first-audio timing and lip-sync cadence.[^3]
- **Scaling**: Inworld says the same graph scales "from 10 test users to 10 million concurrent users", and offers custom on-premise deployments for enterprises.[^53]
- **Observability**: Dashboards tracking end-to-end latency and output quality metrics, enabling production monitoring and cost optimization.[^29]

Inworld described Runtime pricing as "entirely usage-based with no upfront costs", and cited Wishroll and Bible Chat among early users.[^53] The Unreal version of the Runtime was in early access in mid-October 2025, and the Unreal AI Runtime SDK went live on 29 October 2025 alongside early access for Unity.[^3][^29]

### Speech-to-text and full stack

Alongside TTS, Inworld offers Realtime STT, which it says supports more than 30 languages.[^1] Its Realtime API connects STT, a developer-chosen LLM and TTS in one spoken conversation over a single WebSocket or WebRTC connection, handles turn detection and interruptions, and follows the OpenAI Realtime protocol.[^66] The TTS API also offers an OpenAI-compatible endpoint.[^40] Inworld positions its speech-to-speech offering as a voice agent platform with TTS included, in contrast to bring-your-own-components frameworks such as LiveKit Agents, Vapi, Pipecat, Retell and NLX.[^21]

## Key customers

Beyond game studios, Inworld's customer base as of 2025 and 2026 spans:

**[Language learning](https://aiwiki.ai/wiki/language_learning)**: Talkpal, which teaches over 80 languages to more than 5 million users, uses Inworld's Realtime TTS and voice cloning; Inworld reports that a four-week A/B test showed a 40% reduction in TTS costs, a 7% increase in feature usage and a 4% lift in user retention.[^41] The application serves multiple language pairs including English, German, and French.[^41]

**Companion and social apps**: Bible Chat, which Inworld says has about 800,000 daily active users, cut TTS costs by 85% after switching to Inworld while scaling its voice features from roughly 2 million to 20 million characters per week.[^42][^1]

**Live streaming**: Logitech G's Streamlabs integrated Inworld's AI alongside NVIDIA into an Intelligent Streaming Assistant for content creators, unveiled at CES 2025 and offering a 3D sidekick, producer functionality, and technical assistant capabilities.[^43]

**Media and entertainment**: Inworld lists NBCUniversal among media partners that have built with it.[^3]

**Brand and children's experiences**: Inworld's 2023 funding announcement named Alpine Electronics and LG Uplus among companies with Character Engine NPC experiences.[^24] In August 2023 Alps Alpine North America added two AI characters, designed with Inworld AI, Trigger XR and Stage3 Agency, to its Alpine Park experience on Roblox.[^67] In June 2023 LG Uplus launched an English version of its children's metaverse Kidstopia with generative AI NPCs developed in collaboration with Inworld.[^68]

## Comparison with competitors

In voice synthesis, Inworld's own overview names [ElevenLabs](https://aiwiki.ai/wiki/elevenlabs), Cartesia, Deepgram and AssemblyAI among competing providers.[^1] The table below uses each vendor's own published figures as of 30 October 2026. Vendors measure latency differently, so the latency figures are not directly comparable.

| Feature | Inworld AI | ElevenLabs | Cartesia |
|---|---|---|---|
| Flagship TTS models | Realtime TTS-2, Realtime TTS-2 Flash[^63] | Eleven v4, Eleven v4 Turbo[^69] | Sonic 3.6[^70] |
| Vendor-stated latency | 100 ms (TTS-2) and 20 ms (TTS-2 Flash), P90 time to first audio byte measured server-side[^63] | Median time to first speech of about 150 ms (v4 Turbo)[^69] | First audio in under 90 ms[^70] |
| Languages | 200+ languages and locales[^63] | More than 90[^69] | 44[^70] |
| [Voice cloning](https://aiwiki.ai/wiki/voice_cloning) | Instant (3 to 30 seconds of audio); professional cloning in beta[^65] | Instant from 10 seconds of audio; professional voice clones[^69] | Instant from 10 seconds of audio; pro voice cloning on the Startup plan and above[^70][^71] |
| Pricing | $25 per million characters (TTS-2, On-Demand), as low as $5 at Enterprise scale, as of 30 October 2026[^49] | $80 (v4) and $40 (v4 Turbo) per million characters via API[^72] | Monthly credit plans at 1 credit per character, e.g. $49 per month for 1.25 million credits (Startup)[^71] |
| [Artificial Analysis](https://aiwiki.ai/wiki/artificial_analysis) text-to-speech leaderboard, 30 October 2026 | 8th (Realtime TTS-2)[^56] | 1st (Eleven v4 Turbo) and 2nd (Eleven v4)[^56] | 4th (Sonic 3.6)[^56] |

### Character.AI

Character.AI is a consumer-facing AI companion platform that allows end users to chat with fictional, celebrity-inspired, or user-created AI characters. It is not primarily a developer platform and does not provide game engine SDKs or real-time TTS APIs. The two companies serve largely non-overlapping markets: Character.AI targets recreational chat users, while Inworld targets developers building interactive products.

### Replika

Replika is an AI companion application focused on emotional support, personal wellness, and relationship simulation. Like Character.AI, Replika is a consumer product rather than developer infrastructure and does not offer B2B APIs or game engine integrations. The two companies have distinct business models: Replika monetizes subscriptions with individual end users, while Inworld monetizes API consumption with developers and studios.

## Use cases beyond gaming

Inworld's website lists games and media as one of five markets, alongside social and companion apps, learning and education, health and wellness, and voice agents for support, sales and operations.[^5] Reported uses include:

**Language learning**: Talkpal uses cloned voices to give each AI teacher a distinct persona.[^41]

**Customer service and sales agents**: Inworld markets its voice stack for support, sales and operations agents, with its speech-to-speech product aimed at voice agent builders.[^5][^21]

**Interactive entertainment outside games**: Experiences like Niantic's Wol augmented reality product in Muir Woods illustrate how location-based and mixed-reality applications can use AI characters to create contextual, conversational interactions in physical spaces.[^32]

**Content creation tools**: The Xbox Narrative Graph co-development illustrates how AI character infrastructure can assist writers and designers as a creative tool rather than a runtime component, generating narrative structures for human review.[^18]

## Reception

Inworld received press attention beginning with its Disney Accelerator participation in 2022, its August 2023 funding round and its GDC 2024 showcase.[^13][^24][^16] TechCrunch's August 2023 coverage of the financing said that, among the startups prototyping AI tools to generate NPC dialogue, "one of the more promising is Inworld".[^44] In November 2025 Modular, whose platform powers the model, said Inworld TTS 1 Max held the first position on the Artificial Analysis speech leaderboard.[^46]

Developer sentiment toward generative AI in games has been mixed. In the Game Developers Conference's 2024 State of the Game Industry survey, conducted in October 2023 with more than 3,000 respondents, 21% expected generative AI to have a positive impact on the game industry, 57% a mixed impact and 18% a negative impact, while 84% said they were somewhat or very concerned about the ethics of using it.[^73][^47]

The expansion into voice cloning also touches on performers' concerns: as of May 2025, a strike by SAG-AFTRA video game voice actors had run for about a year, with generative AI the key sticking point.[^48]

## Limitations

Reported limitations include:

**Inference cost at scale**: AI-native games generate large inference volumes. Death by AI used 1.2 billion tokens in its first month, and its developer found its original OpenAI and ElevenLabs setup financially unsustainable three days after launch.[^34] Inworld's case studies describe custom engineering for such customers.[^34][^35]

**Hallucination and lore breakage**: A 2024 survey of LLMs in games notes that LLM-driven characters are prone to hallucinations that do not fit the intended scenario, a risk that can be reduced by giving the model the current state of the environment as well as the conversation history.[^74] This is a general challenge of using probabilistic [generative model](https://aiwiki.ai/wiki/generative_model)s in constrained fictional contexts; Inworld's Contextual Mesh controls were intended to reduce it.[^26]

**Platform changes**: Inworld has retired or deprecated several products that developers built on, including the legacy Character Studio, the TTS-1 models (discontinued 15 June 2026) and the TTS 1.5 models (deprecated).[^50][^63] Its public GitHub repository for the Node.js voice agent template, which the Studio migration guide still recommends, was archived on 28 July 2026.[^75]

**Adoption timeline in AAA gaming**: Ubisoft described its 2024 NEO NPC work as "only a prototype", with "still a way to go before it can be implemented in a game".[^45] In the GDC 2024 survey, 21% of AAA developers said their companies had banned generative AI tools, compared with 9% of indie developers.[^47] Inworld's published game case studies, such as Death by AI and Status, come from small teams.[^34][^35]

## See also

- [ElevenLabs](https://aiwiki.ai/wiki/elevenlabs)
- [Cartesia](https://aiwiki.ai/wiki/cartesia)
- [Voice AI](https://aiwiki.ai/wiki/voice_ai)
- [Character.AI](https://aiwiki.ai/wiki/character_ai)
- [Adding Generative AI Features to Games](https://aiwiki.ai/wiki/adding_generative_ai_features_to_games)

## References

[^1]: Inworld AI, "What is Inworld AI? Realtime Voice AI Research Lab", Inworld AI, 2026-02-19 (updated 2026-03-12). https://inworld.ai/resources/what-is-inworld-ai. Accessed 2026-10-30.
[^2]: Contrary Research, "Report: Inworld AI Business Breakdown & Founding Story", Contrary Research, updated 2025-07-11. Archived copy: https://web.archive.org/web/20260521071214/https://research.contrary.com/company/inworld-ai. Accessed 2026-10-30.
[^3]: Inworld AI, "The new AI infrastructure for scaling games, media, and characters", Inworld AI Blog, 2025-10-15. https://inworld.ai/blog/new-ai-infrastructure-scaling-games-media-characters. Accessed 2026-10-30.
[^4]: Asif Razzaq, "Inworld AI Launches Realtime TTS-2: A Closed-Loop Voice Model That Adapts to How You Actually Talk", MarkTechPost, 2026-05-05. https://www.marktechpost.com/2026/05/05/inworld-ai-launches-realtime-tts-2-a-closed-loop-voice-model-that-adapts-to-how-you-actually-talk/. Accessed 2026-05-24.
[^5]: Inworld AI, "Inworld AI | AI built for people" (homepage), Inworld AI, 2026. https://inworld.ai/. Accessed 2026-10-30.
[^6]: Bret Kinsella, "Ilya Gelfenbeyn, Founding CEO of API.ai (Acquired by Google and Now Dialogflow): Voicebot Podcast Ep 197", Voicebot.ai, 2021-02-28. https://voicebot.ai/2021/02/28/ilya-gelfenbeyn-founding-ceo-of-api-ai-acquired-by-google-and-now-dialogflow-voicebot-podcast-ep-197/. Accessed 2026-05-24.
[^7]: Crunchbase, "Inworld AI: Founder and Investor Profile", Crunchbase, 2026. https://www.crunchbase.com/organization/inworld-ai. Accessed 2026-05-24.
[^8]: Crunchbase, "Kylan Gibbs Co-Founder Inworld AI Profile", Crunchbase, 2026. https://www.crunchbase.com/person/kylan-gibbs. Accessed 2026-05-24.
[^9]: Wellfound, "Inworld AI: Founder, Leadership & Team", Wellfound, 2025. https://wellfound.com/company/inworld/people. Accessed 2026-05-24.
[^10]: Business Wire, "Visual Effects Giant John Gaeta Joins Inworld AI as Chief Creative Officer", Business Wire, 2022-06-01. https://www.businesswire.com/news/home/20220601005484/en/Visual-Effects-Giant-John-Gaeta-Joins-Inworld-AI-as-Chief-Creative-Officer. Accessed 2026-05-24.
[^11]: Inworld AI / Lightspeed, "Building With Inworld: The Character Engine for AI NPCs", Lightspeed Venture Partners, 2023. https://lsvp.com/stories/inworld-ai-npcs-character-engine/. Accessed 2026-05-24.
[^12]: GlobeNewswire, "Inworld AI Announces $7 Million in Seed Funding from Kleiner Perkins, CRV and Meta for Metaverse AI Development Platform", GlobeNewswire, 2021-11-08. https://www.globenewswire.com/news-release/2021/11/08/2329206/0/en/Inworld-AI-Announces-7-Million-in-Seed-Funding-from-Kleiner-Perkins-CRV-and-Meta-for-Metaverse-AI-Development-Platform.html. Accessed 2026-05-24.
[^13]: PR Newswire, "The Walt Disney Company Explores the Future of Immersive Experiences at 2022 Disney Accelerator Demo Day", PR Newswire, 2022-11-10. https://www.prnewswire.com/news-releases/the-walt-disney-company-explores-the-future-of-immersive-experiences-at-2022-disney-accelerator-demo-day-301675134.html. Accessed 2026-05-24.
[^14]: Eric Hal Schwartz, "Inworld AI Shows Off Star Wars 'Droid Maker' After Completing Disney Accelerator", Voicebot.ai, 2022-11-11. https://voicebot.ai/2022/11/11/inworld-ai-shows-off-star-wars-droid-maker-after-completing-disney-accelerator/. Accessed 2026-05-24.
[^15]: Catherine Lewis, "Horrifying Stardew Valley mod gives NPCs AI-powered 'brains'", GamesRadar+, 2024. https://www.gamesradar.com/horrifying-stardew-valley-mod-gives-npcs-ai-powered-brains/. Accessed 2026-05-24.
[^16]: Dean Takahashi, "Ubisoft debuts GenAI-powered Neo NPCs and gameplay prototype", GamesBeat, 2024-03-19. https://gamesbeat.com/ubisoft-neo-npcs-nvidia-inworldai-gdc/. Accessed 2026-05-24.
[^17]: Microsoft / Xbox, "Xbox and Inworld AI partner to empower game creators with the potential of Generative AI", Microsoft Game Dev, 2023-11-06. https://developer.microsoft.com/en-us/games/articles/2023/11/xbox-and-inworld-ai-partnership-announcement/. Accessed 2026-05-24.
[^18]: Alessio Palumbo, "Inworld AI Blows Up at GDC 2024 with Ubisoft's NEO NPCs, Xbox's Narrative Graph Tool and More", Wccftech, 2024-03-22. https://wccftech.com/inworld-ai-blows-up-at-gdc-2024-between-ubisofts-neo-npcs-xboxs-narrative-graph-tool-and-more/. Accessed 2026-05-24.
[^19]: Microsoft, "Project Explora", Microsoft Game Dev, 2024. https://developer.microsoft.com/en-us/games/products/project-explora/. Accessed 2026-05-24.
[^20]: NVIDIA, "NVIDIA Digital Human Technologies Bring AI Game Characters To Life", NVIDIA GeForce News, 2024-03-18. https://www.nvidia.com/en-us/geforce/news/nvidia-ace-gdc-gtc-2024-ai-character-game-and-app-demo-videos/. Accessed 2026-05-24.
[^21]: Inworld AI, "Voice Agent Platforms with Built-In TTS: 2026 Architecture Guide", Inworld AI, 2026-04-30. https://inworld.ai/resources/voice-agent-platforms-with-built-in-tts. Accessed 2026-10-30.
[^22]: Inworld AI, "Integrations", Inworld AI Documentation, 2026. https://docs.inworld.ai/docs/tts/integrations. Accessed 2026-05-24.
[^23]: Intel Capital, "Inworld AI Closes $50M Series A to Populate Video Games and the Metaverse with Interactive AI-Driven Characters", Intel Capital, 2022-08-23. https://www.intelcapital.com/inworld-ai-closes-50m-series-a-to-populate-video-games-and-the-metaverse-with-interactive-ai-driven-characters/. Accessed 2026-05-24.
[^24]: Business Wire, "Inworld AI, the Leading Character Engine, Raises New Funding From Lightspeed, Stanford, Microsoft's M12 Fund, First Spark (Eric Schmidt), and More, Bringing Valuation to Over $500 Million", Business Wire, 2023-08-02. https://www.businesswire.com/news/home/20230802502983/en/Inworld-AI-the-Leading-Character-Engine-Raises-New-Funding-From-Lightspeed-Stanford-Microsofts-M12-Fund-First-Spark-Eric-Schmidt-and-More-Bringing-Valuation-to-Over-$500-Million. Accessed 2026-05-24.
[^25]: Dean Takahashi, "Inworld AI raises $50M round at $500M valuation for AI game characters", GamesBeat, 2023-08-02. https://venturebeat.com/games/inworld-ai-raises-new-round-at-500m-valuation-for-ai-game-characters/. Accessed 2026-05-24.
[^26]: Kylan Gibbs, "Feature Release Week 2: Our Improved Contextual Mesh", Inworld AI Blog, 2023-07-27. Archived copy: https://web.archive.org/web/20240612204452/https://inworld.ai/blog/improved-contextual-mesh. Accessed 2026-10-30.
[^27]: Epic Games, "AI NPC Engine v.1.5 - Dialogue & Behavior for Unreal Engine - Inworld", Unreal Engine Marketplace, listed 2023-02-02. Archived copy (2025-09-10): https://web.archive.org/web/20250910225003/https://www.unrealengine.com/marketplace/en-US/product/inworld-ai-characters-dialogue. Accessed 2026-10-30.
[^28]: Inworld AI, "Unity", Inworld AI Documentation. Archived copy (2025-03-19): https://web.archive.org/web/20250319145913/https://docs.inworld.ai/docs/tutorial-integrations/Unity/. Accessed 2026-10-30.
[^29]: Inworld AI, "Unreal AI Runtime: Build AI-powered games faster", Inworld AI Blog, 2025-10-29. https://inworld.ai/blog/introducing-unreal-ai-runtime-sdk. Accessed 2026-10-30.
[^30]: Inworld AI, "Godot SDK", Inworld AI Documentation. Archived copy (2024-02-24): https://web.archive.org/web/20240224111947/https://docs.inworld.ai/docs/tutorial-integrations/godot/. Accessed 2026-10-30.
[^31]: Inworld AI, "How NetEase integrated Inworld AI in Cygnus Enterprises, Part 1", Inworld AI Blog, 2023. https://inworld.ai/blog/netease-cygnus-enterprise-integrated-ai-npcs-inworld. Accessed 2026-05-24.
[^32]: Niantic Labs, "Meet Wol: Niantic's first mixed reality experience powered by 8th Wall", Niantic Labs, 2023. https://nianticlabs.com/news/meetwol. Accessed 2026-05-24.
[^33]: Tomek Tomczyk, "A Vaudeville Early Access Postmortem", Game Developer, 2023. https://www.gamedeveloper.com/design/vaudeville-pre-mortem. Accessed 2026-05-24.
[^34]: Inworld AI, "How Inworld Helped Death by AI Reach Profitability at 20M Players", Inworld AI Blog, 2025-08-01. https://inworld.ai/blog/how-inworld-helped-the-ai-game-death-by-ai-with-20-million-players-reach-profitability. Accessed 2026-10-30.
[^35]: Inworld AI, "How Status Scaled to 500K Users at 95% Lower Cost", Inworld AI Blog, 2025-09-10. https://inworld.ai/blog/wishroll-status-cutting-ai-costs-by-95-percent. Accessed 2026-10-30.
[^36]: Inworld AI, "Introducing Inworld TTS", Inworld AI Blog, 2025-08-15. https://inworld.ai/blog/introducing-inworld-tts. Accessed 2026-05-24.
[^37]: Inworld AI Research, "TTS-1 Technical Report", arXiv:2507.21138, 2025. https://arxiv.org/abs/2507.21138. Accessed 2026-05-24.
[^38]: GlobeNewswire, "Inworld AI Launches TTS-1.5, a Groundbreaking Voice AI model for the Next Wave of Realtime AI Applications", GlobeNewswire, 2026-01-21. https://www.globenewswire.com/news-release/2026/01/21/3222492/0/en/Inworld-AI-Launches-TTS-1-5-a-Groundbreaking-Voice-AI-model-for-the-Next-Wave-of-Realtime-AI-Applications.html. Accessed 2026-05-24.
[^39]: Inworld AI, "Realtime TTS-2: A new frontier voice model that feels as it sounds", Inworld AI Blog, 2026-05-05. https://inworld.ai/blog/realtime-tts-2. Accessed 2026-10-30.
[^40]: Inworld AI, "Realtime TTS API", Inworld AI, 2026. https://inworld.ai/tts-api. Accessed 2026-05-24.
[^41]: Inworld AI, "Talkpal AI scales to 5 million language learners with Realtime TTS", Inworld AI Blog, 2026-01-13. https://inworld.ai/blog/talkpal-ai-scales-to-5-million-language-learners-with-inworld-tts. Accessed 2026-10-30.
[^42]: Inworld AI, "Voice AI for AI Companions", Inworld AI, 2026-04-02 (updated 2026-04-05). https://inworld.ai/resources/voice-ai-for-ai-companions. Accessed 2026-10-30.
[^43]: Business Wire, "Inworld AI Collaborates with Logitech G's Streamlabs and NVIDIA to Power the Brains Behind Intelligent Streaming Assistant", Business Wire, 2025-01-06. https://www.businesswire.com/news/home/20250106653761/en/Inworld-AI-Collaborates-with-Logitech-Gs-Streamlabs-and-NVIDIA-to-Power-the-Brains-Behind-Intelligent-Streaming-Assistant. Accessed 2026-05-24.
[^44]: Kyle Wiggers, "Inworld, a generative AI platform for creating NPCs, lands fresh investment", TechCrunch, 2023-08-02. https://techcrunch.com/2023/08/02/inworld-a-generative-ai-platform-for-creating-npcs-lands-fresh-investment/. Accessed 2026-10-30.
[^45]: Ubisoft, "How Ubisoft's New Generative AI Prototype Changes the Narrative for NPCs", Ubisoft News, 2024-03-19. https://news.ubisoft.com/en-us/article/5qXdxhshJBXoanFZApdG3L/how-ubisofts-new-generative-ai-prototype-changes-the-narrative-for-npcs. Accessed 2026-10-30.
[^46]: Modular, "TTS 1 Max (powered by Modular Platform) Ranked #1 Speech Model on Artificial Analysis", Modular Blog, 2025-11-07. https://www.modular.com/blog/tts-1-max-ranked-1-speech-model-on-artificial-analysis. Accessed 2026-10-30.
[^47]: Business Wire, "Game Developers Conference's 2024 State of the Game Industry Survey Shows Developer Concern About Layoffs, Generative AI Usage, Game Engine Pricing Changes and Return to Office Policies", Business Wire, 2024-01-18. https://www.businesswire.com/news/home/20240118843393/en. Accessed 2026-05-24.
[^48]: Eliot Lefebvre, "SAG-AFTRA's voice actor strike continues with generative AI being the key sticking point", Massively Overpowered, 2025-05-07. https://massivelyop.com/2025/05/07/sag-aftras-voice-actor-strike-continues-with-generative-ai-being-the-key-sticking-point/. Accessed 2026-10-30.
[^49]: Inworld AI, "Pricing". https://inworld.ai/pricing. Accessed 2026-10-30.
[^50]: Inworld AI, "Character Studio deprecation", Inworld AI Documentation, updated 2026-10-26. https://docs.inworld.ai/character-studio-deprecation. Accessed 2026-10-30.
[^51]: Inworld AI, homepage, archived copies of 2023-11-14 (https://web.archive.org/web/20231114185247/https://inworld.ai/) and 2024-05-17 (https://web.archive.org/web/20240517114405/https://inworld.ai/). Accessed 2026-10-30.
[^52]: Inworld AI, "Introducing Inworld TTS", Inworld AI Blog. Archived copy (2025-06-25): https://web.archive.org/web/20250625172840/https://inworld.ai/blog/introducing-inworld-tts. Accessed 2026-10-30.
[^53]: Dean Takahashi, "Inworld Runtime is 1st AI runtime for consumer apps", GamesBeat, 2025-08-13. https://gamesbeat.com/inworld-runtime-is-1st-ai-runtime-for-consumer-apps/. Accessed 2026-10-30.
[^54]: Inworld AI, "Introducing Inworld Inference", Inworld AI Blog, 2026-10-14. https://inworld.ai/blog/introducing-inworld-inference. Accessed 2026-10-30.
[^55]: Artificial Analysis, "Controlled Voice Leaderboard - Top AI Voice Cloning Models", Artificial Analysis, viewed 2026-10-30. https://artificialanalysis.ai/text-to-speech/leaderboard/controlled-voice. Accessed 2026-10-30.
[^56]: Artificial Analysis, "Text to Speech Leaderboard - Top AI Speech Models" (provider voice), Artificial Analysis, viewed 2026-10-30. https://artificialanalysis.ai/text-to-speech/leaderboard/provider-voice. Accessed 2026-10-30.
[^57]: Inworld AI, "Cost is the wall in front of consumer AI. We are taking it down.", Inworld AI Blog, 2026-06-10. https://inworld.ai/blog/consumer-ai-cost-pricing. Accessed 2026-10-30.
[^58]: Inworld AI, "We have acquired Ultravox to Advance Realtime Voice AI", Inworld AI Blog, 2026-09-30. https://inworld.ai/blog/inworld-acquires-ultravox. Accessed 2026-10-30.
[^59]: Lisa Stiffler, "Seattle voice AI startup Ultravox, formerly known as Fixie, acquired by Inworld", GeekWire, 2026-10-02. https://www.geekwire.com/2026/seattle-voice-ai-startup-ultravox-formerly-known-as-fixie-acquired-by-inworld/. Accessed 2026-10-30.
[^60]: Rachel Kaser, "Inworld AI raises $10M to develop AI-powered virtual characters for games", GamesBeat, 2022-03-03. https://gamesbeat.com/inworld-ai-raises-10m-to-develop-ai-powered-virtual-characters-for-games/. Accessed 2026-10-30.
[^61]: Inworld AI, "Inworld Studio" (studio.inworld.ai), archived copies of 2025-06-16 (https://web.archive.org/web/20250616151108/https://studio.inworld.ai/) and 2025-09-09, redirecting to platform.inworld.ai (https://web.archive.org/web/20250909131858/https://studio.inworld.ai/). Accessed 2026-10-30.
[^62]: Inworld AI, "Multi-Agent", Inworld AI Documentation. Archived copy (2024-12-10): https://web.archive.org/web/20241210113903/https://docs.inworld.ai/docs/tutorial-basics/multi-agent/. Accessed 2026-10-30.
[^63]: Inworld AI, "TTS models", Inworld AI Documentation, 2026. https://docs.inworld.ai/tts/tts-models. Accessed 2026-10-30.
[^64]: Inworld AI, "TTS release notes", Inworld AI Documentation, 2026 (entries dated 2026-01-21 to 2026-09-11). https://docs.inworld.ai/release-notes/tts. Accessed 2026-10-30.
[^65]: Inworld AI, "Instant voice cloning" and "Professional voice cloning", Inworld AI Documentation, 2026. https://docs.inworld.ai/tts/instant-voice-cloning and https://docs.inworld.ai/tts/professional-voice-cloning. Accessed 2026-10-30.
[^66]: Inworld AI, "Realtime API overview", Inworld AI Documentation, 2026. https://docs.inworld.ai/realtime/overview. Accessed 2026-10-30.
[^67]: Alps Alpine North America, "Alpine Continues to Invest in Alpine Park with the First AI Character Integration on Roblox", Alpine press release, 2023-08-15. https://www.alpine-usa.com/company/press/2023/alpine-continues-to-invest-in-alpine-park-with-the-first-ai-character-integration-on-roblox. Accessed 2026-10-30.
[^68]: Ahn Se-jun, "생성형AI 탑재한 LGU+ '키즈토피아' 글로벌 무대 밟는다" (LG Uplus Kidstopia with generative AI goes global), inews24, 2023-06-25 (in Korean). https://www.inews24.com/view/1606021. Accessed 2026-10-30.
[^69]: ElevenLabs, "Eleven v4: Our most expressive text-to-speech AI model yet", ElevenLabs Blog, 2026-09-28. https://elevenlabs.io/blog/eleven-v4. Accessed 2026-10-30.
[^70]: Cartesia, "Sonic AI voice: real-time TTS API with emotion", Cartesia, 2026. https://cartesia.ai/sonic. Accessed 2026-10-30.
[^71]: Cartesia, "Pricing: free plan, TTS, STT, and voice agents", Cartesia, 2026. https://cartesia.ai/pricing. Accessed 2026-10-30.
[^72]: ElevenLabs, "ElevenLabs API Pricing", ElevenLabs, 2026. https://elevenlabs.io/pricing/api. Accessed 2026-10-30.
[^73]: Game Developers Conference, "2024 State of the Game Industry" report (survey conducted 11-29 October 2023), GDC, 2024-01-18. https://reg.gdconf.com/state-of-game-industry-2024 (PDF copy: https://web.engr.oregonstate.edu/~mjb/cs557/2024StateOfTheGameIndustry.pdf). Accessed 2026-10-30.
[^74]: Roberto Gallotta et al., "Large Language Models and Games: A Survey and Roadmap", IEEE Transactions on Games, 2024 (arXiv:2402.18659). https://arxiv.org/abs/2402.18659. Accessed 2026-10-30.
[^75]: Inworld AI, "voice-agent-node: Inworld Runtime Node.js Voice Agent Template" (repository archived 2026-07-28), GitHub. https://github.com/inworld-ai/voice-agent-node. Accessed 2026-10-30.