Disney Granted AI Voice Patent For Costumed Characters

Disney has been granted a US patent for an AI system that writes a reply in a character's voice when a guest speaks to a performer in costume, then plays that reply through a speaker built into the costume. Patent US 12,784,732 B2 was granted to Disney Enterprises, Inc. on 3rd November 2026 and uses a large language model to write each reply, with a camera inside the costume reading the performer's real facial expression to set the emotional tone of the voice.

How Does Disney's Talking Character Patent Work?

According to the patent, performers playing Disney characters are often kept from using their own voices to avoid inconsistency and what Disney calls brand dilution, which leaves them relying on poses, gestures and miming. Pre-recorded lines in a brand-approved voice are the existing workaround, but the patent takes this further.

The patent sets out the process as a flowchart. It works like this.

1. A guest speaks to the performer, who is wearing the character's costume or mask.
2. Microphones and sensors capture what the guest said and how the performer reacted, and send a description of both to Disney's computing platform.
3. The system looks up the character's communication traits in a character database, covering things like the character's persona, how they speak, their favourite vocabulary and any unusual expressions they're known for.
4. A large language model uses the guest's words and those traits to write a reply in real time.
5. A separate AI model turns the written reply into audio in the character's voice, using the performer's facial expression to pick the emotional tone.
6. The audio is sent back and played through a speaker in the costume.

The patent defines real-time as a reply arriving without an unnatural-seeming delay, which is important because any delay would leave the performer nodding enthusiastically for several seconds. It also says the language model may be trained on the specific character using reinforcement learning, so the replies sound like that character rather than a generic chatbot in a Mickey suit. Each step can run automatically. The patent says a human editor or system operator could review or even modify the replies, but describes that as optional.

How Does The Performer's Facial Expression Change The Character's Voice?

This is what sets Disney's system apart from a talking speaker in a costume. The performer's face is read by an inward-facing camera, which the patent compares to a selfie camera, and that expression is fed into the voice model along with the written reply. The patent gives examples of how this plays out.

Examples taken from Disney patent US 12,784,732 B2.

The patent also describes mechanical actuators that could produce facial expressions on the costume or mask itself, so in some versions the character's face could move along with the voice. If a child has just dropped their churro on the floor, a beaming Mickey cheerfully asking whether they're having a great day won't help, but a frown from the performer inside the costume, which the patent links to disappointment in the voice, could bring both the character's tone and face down to match the moment.

What Sensors Are In Disney's Patented Character Costume?

Quite a few. The patent lists cameras, automatic speech recognition, RFID, facial recognition, object recognition, environmental sensors and eye tracking that follows the performer's eye movement. There are also microphones, a module that detects the stress, rhythm and intonation of the guest's speech, and a speech-to-text module. The cameras could include RGB, depth-sensing and infrared types.

The system can also take in what's going on around the performer. The patent mentions background conversations, music, announcements, the day of the week and time of day, the weather, and scheduled events nearby. In theory, that gives a character enough context to comment on the rain or a parade starting soon.

Will Disney's AI Characters Remember You?

Possibly. The patent describes an optional interaction history database that stores previous conversations between a guest and a character, and the language model can use that history when writing a new reply. It suggests the database could hold hundreds, thousands or even millions of interaction histories.

On privacy, the patent says Disney's preferred approach is to keep personally identifiable information out of the stored history, and it states that the system doesn't need to keep details such as age, gender, race or ethnicity. In some versions, it would only hold a set number of the most recent interactions. Facial recognition appears in the list of possible sensors, but the patent doesn't spell out how the system would know which guest it's talking to.

Which Characters Could Use Disney's AI Voice Patent?

The patent covers cartoon characters, video game avatars, fictional humans from books, film or television, fictional non-human characters and historical figures. It says the character database would typically store tens, hundreds or thousands of character profiles. Each profile can include a character archetype. The patent uses the hero and the villain as examples of opposite traits. The profile can also record the character's persona, such as optimism or pessimism and how assertive they are.

When Did Disney File The AI Character Voice Patent?

Disney Enterprises, Inc. filed the application on 20th May 2024, and it was published as US 2025/0356839 A1 on 20th November 2025. The patent was granted on 3rd November 2026, almost two and a half years after filing. The inventors named are Alif Khalfan of Redwood City, California and Malcolm E. Murdock of Los Angeles, California. The granted patent has 19 claims and 4 drawing sheets, and you can read the full patent on the USPTO website.

Check Walt Disney World offers for UK guests

What Does Disney's AI Character Voice Patent Mean For UK Guests?

For anyone planning a trip, nothing changes yet. A patent protects an idea and doesn't commit Disney to building anything. This one doesn't name a theme park or give a launch date, and no specific Disney character is mentioned. The patent also says nothing about which languages the system would speak, which would be relevant at Disneyland Paris.

If Disney does put the system into use, the patent points to these outcomes for guests.

• A character in costume could answer your questions in their own recognisable voice instead of miming
• Replies would be written by an AI language model as you talk
• The character's mood would follow the performer's face, so the person inside the costume still shapes the conversation
• A returning guest could get a reply that draws on an earlier chat with the same character

If the idea of making small talk with an AI-voiced Mickey fills you with dread, I wouldn't worry about it yet. There's still a real performer inside the costume, and the patent uses their facial expression to shape each reply, with the AI writing and voicing the words. Disney doesn't have to use everything it patents, and a granted US patent stops competitors in the US from building the same system whether Disney uses it or not. Whether guests will want to chat with a character they know is being voiced by AI is something only a meet and greet will answer, if Disney ever puts this into use.

If you want to keep an eye on deals while Disney works out what to do with it, our Disneyland Paris offers guide and Walt Disney World deals guide for UK guests are kept up to date with the latest offers.

Join the FREE Mickey From The UK Email Club

Click Here To Join The FREE Mickey From The UK Email Club

You may not see all our content on social media - signing up to our Email Club means you won't miss out.

See more from Mickey From The UK in Google Search

Click Here To Add Us As A Preferred Source On Google

Adding us as a preferred source means you'll see more of our news, guides, and reviews when you search.

Disney AI Character Voice Patent: Your Questions Answered

Has Disney patented AI that lets costumed characters talk?
Yes - Disney Enterprises, Inc. was granted US patent 12,784,732 B2 on 3rd November 2026. It covers an AI system that writes a reply in a character's voice when someone speaks to a performer impersonating that character, then plays it through the performer's costume or mask.

When did Disney apply for the AI character voice patent?
Disney filed the application on 20th May 2024. It was published as US 2025/0356839 A1 on 20th November 2025 and granted on 3rd November 2026.

How does Disney's AI character voice system work?
A large language model takes a description of what the guest said, along with the character's stored communication traits, and writes a reply in real time. A second AI model then turns that reply into audio in the character's voice, using the performer's facial expression to set the emotional tone.

Does the patent say which Disney parks or characters will use it?
No - the patent doesn't name any theme park or launch date, and no specific Disney character is mentioned in it.

Does the performer's facial expression change the character's voice?
Yes - the patent gives the example of a smile producing happiness in the character's voice. A smirk would come across as smugness and a frown as disappointment.

Will Disney's AI characters remember guests?
Yes - the patent describes an optional interaction history database that stores past conversations between a guest and a character, which the language model can use when writing a reply. The patent says the system doesn't need to keep personally identifiable information such as age, gender, race or ethnicity.

Who invented Disney's AI character voice patent?
The patent names Alif Khalfan of Redwood City, California and Malcolm E. Murdock of Los Angeles, California as the inventors.