Is Lingotion legally safe? Yes, Lingotion has licenses for every single byte of AI training data, and all AI models are trained from scratch. Lingotion can provide compliance documentation for all training data with licenses that allow the use of data for training generative AI. Additionally, Lingotion has direct agreements with the actors to guarantee compliance with data protection laws.
Is Lingotion ethical? Yes, Lingotion works together with actors using a revenue-sharing model where 30% of Lingotion revenue is paid as royalties to the actors providing the AI training data.
Is Lingotion using internet scraping, open models, or synthetic data from open models? No, every single byte of training data is sourced properly through licenses that allow the use of data for training commercial generative AI models for the purpose of generating acting.
Do actors need to join live sessions? No. Recordings are done in your own home-studio.
How do I know what to record? All text is generated through our custom software, designed to guide and streamline your recording process efficiently.
Do I need professional equipment to record? Yes. The work requires professional audio gear, such as a Røde mic (or similar), a dedicated interface (e.g. Focusrite), and an audio-booth.
Where do I record my lines?All recordings are done directly on the Lingotion web-portal.
How often are royalties paid out? Are there minimum thresholds? What does the reporting schedule look like? Royalties are reported and paid out annually. There is a minimum payout threshold of 300 EUR.
Where and how is the data stored? All data is securely stored in Google Cloud infrastructure and is accessible only by the company. We follow strict access and data protection protocols to ensure the highest levels of privacy and security.
Will my voice still be available if Lingotion doesn’t survive as a company? Yes. Any voices that have already been sold will continue to exist and be usable. Even in the unlikely event that Lingotion ceases to operate, all voice data remains protected under the Terms of Service.
Can I invite other actors? Yes, referrals are welcome! Lingotion is often looking for diverse voices and accents.
Do actors need to join live sessions? No. Recordings are done in your own home-studio.
How do I know what to record? All text is generated through our custom software, designed to guide and streamline your recording process efficiently.
Do I need professional equipment to record? Yes. The work requires professional audio gear, such as a Røde mic (or similar), a dedicated interface (e.g. Focusrite), and an audio-booth.
Where do I record my lines?All recordings are done directly on the Lingotion web-portal.
How often are royalties paid out? Are there minimum thresholds? What does the reporting schedule look like? Royalties are reported and paid out annually. There is a minimum payout threshold of 300 EUR.
Where and how is the data stored? All data is securely stored in Google Cloud infrastructure and is accessible only by the company. We follow strict access and data protection protocols to ensure the highest levels of privacy and security.
Will my voice still be available if Lingotion doesn’t survive as a company? Yes. Any voices that have already been sold will continue to exist and be usable. Even in the unlikely event that Lingotion ceases to operate, all voice data remains protected under the Terms of Service.
Can I invite other actors? Yes, referrals are welcome! Lingotion is often looking for diverse voices and accents.
Is it just a voice cloning TTS (Text to Speech)? No, it is much more than just a voice cloning TTS. Lingotion captures the uniqueness of an actor and clones exactly how the actor performs a character, including the entire emotional range, dialects, pronunciations, and most importantly, all the flaws. It is in the flaws that we become human and unique, and compared to standard AI solutions that output human averages, the Lingotion solution captures and clones the unique differences that make a character human.
Can you clone casted actors? Yes, but the process is different from standard voice cloning. Lingotion is committed to ethical AI practices, which means we require agreement with the actor before cloning their acting to ensure their rights are protected. Additionally, Lingotion supports a wide range of emotional expressions, which requires a more extensive recording process.
Can it clone an actor doing a specific type of character? Yes, that is one of the central functions of Lingotion. It clones the uniqueness of an actor performing a character, where each emotional expression, pronunciation, dialect, and even flaws like mispronunciation are cloned exactly.
Does Lingotion generate the dialogue script? No, Lingotion only generates the acting and never the script. It is up to the game designer to select the method of creating the dynamic script using, for example, templates or procedural generation.
If it is real actors doing the acting, can you mix manual recordings and Lingotion AI-generated acting? Yes, that is one of the use cases for Lingotion. Lingotion clones an actor performing a character with the entire emotional expression as close as possible to real recordings. This means you can hire the actor and mix generated acting with prerecorded manual recordings, creating a seamless blend between the two methods.
How many emotions are there? There are currently 33 emotions for each character, with 59 emotions on the roadmap.
Is it possible to mix emotions? Emotion mixing is on the roadmap with a planned release in 2026.
Do you support non-verbal sounds such as grunts, etc.? Non-verbal control characters are on the roadmap with a planned release in 2026.
How are the control characters edited? Control characters are emojis.
Can you support languages other than English? Yes. However, the current release only supports English. We will add more languages over time based on market demand.
What game engines are supported? Unity and Unreal are supported with SDKs provided by Lingotion, but Lingotion also provide integration with custom engines.
Does Lingotion use any cloud services? No. Lingotion runs entirely on the player’s device, fully integrated within the game engine. It does not use any cloud infrastructure.
Can it run on both GPU and CPU? Yes. You can choose whether the Lingotion Engine runs on the CPU or GPU.
Can it run in real time on a CPU using only a few cores? Yes, the Lingotion AI model is a proprietary solution that is super-efficient both in terms of memory usage and compute performance.
Is Lingotion LLM-based? No. LLMs are large autoregressive models that often require too much memory and GPU performance to run concurrently with game graphics. Lingotion uses a proprietary non-autoregressive AI architecture to achieve the necessary performance efficiency.
Is there any way to synchronize the audio with external events to create a trigger for a position in the text? Yes. Since the timing depends on the character and emotion, Lingotion provides text-position-to-sample-position synchronization. The SDK can provide the exact sample position when a word in the input text is spoken.
What does the input look like? It is a JSON structure that specifies language, emotion, and text for a dialogue or part of a dialogue, similar to a script.
Will it affect graphics rendering if it runs on the GPU? Only minimally. Lingotion is designed to minimize its impact on real-time graphics performance.
What platforms are supported? Lingotion currently supports PC, iOS, Android, Quest, and macOS. Support for other platforms is on the roadmap and can be provided upon request.
Can I start testing now? Yes, the Lingotion SDK is publicly available on GitHub, and Lingotion provides two starter characters that are free to download and test. The starter characters only incur a cost if you publish a game using them.
Can you support custom in-house engines? Yes, we provide integration and maintenance for custom engines.
How large are the AI models? The AI models come in five sizes depending on the hardware capabilities: X-Small, Small, Medium, Large, and X-Large. The sizes range from 3.4 million parameters for X-Small to 33 million parameters for X-Large. All models can run on CPU. X-Large is typically intended for high-end PCs and consoles, Medium is for mid- to low-end PCs and older high-end phones, and X-Small is for older budget phones with no GPU support and very poor CPU performance.
How much memory does an character model require? The current model size per character is 13 MB–132 MB depending on model size. The roadmap includes further reduction of character model sizes.
Can you run multiple syntheses concurrently? Yes, the SDK is thread-safe, and it is possible to run multiple syntheses in parallel due to the super-efficient performance of the Lingotion model.
Will it synthesize the entire sentence before getting the audio? No, to reduce latency, Lingotion starts streaming audio before the entire sentence has been synthesized.
How long is the latency from synth call to receiving audio? It depends on the hardware capabilities, but it is usually in the range of
100 ms–500 ms. For very low-performance devices, it can exceed 1000 ms.
How is the business model? You purchase an actor clone per game title for a one-time fee per title. For smaller studios with limited budgets, there is a reduced one-time fee option with a game-title revenue-sharing model.
Is there any per-line, per-character, or per-usage cost? No—there are no such costs. The Lingotion acting engine runs entirely on the player’s device, so there are no usage fees of any kind.
Do actors receive royalties? Yes, from the fees that the studio pays to Lingotion. Lingotion acts as an agent for the actors. Rather than paying the actor directly for their time, studios license a clone of the actor from Lingotion for use in a specific title. This ensures actors are compensated fairly through royalty agreements.
Join our Discord community and get support here:
