AI voices in games and adverts: how character voices are made in 2026

In 2026, AI voices in games and adverts are mostly used alongside human voice actors rather than instead of them. Studios use them for temporary dialogue during development, short lines like combat barks, fast localisation and, increasingly, characters that respond live to players. Brands use AI voiceover for ads to produce many versions quickly. The rules have tightened too: union agreements now require consent and payment for digital replicas of a performer’s voice, and new disclosure rules mean audiences increasingly get told when a voice is synthetic.

This guide explains how AI character voices are actually made, where the debate with voice actors stands, and what small creators should know before using them.

How studios use AI voices in games

Temporary voiceover during development

The least controversial use is also one of the most common. Games change constantly during development, and recording a professional actor for every draft line isn’t practical. Text to speech lets designers hear a scene, test pacing and catch problems early. Final lines are then recorded by actors once the script settles.

NPC barks and incidental lines

Barks are the short, repeated lines characters shout in the middle of play: “Reloading!”, “Over there!”, “Enemy down”. A game can need thousands of them. Embark Studios publicly said in 2023 that The Finals uses text to speech for contestant barks and commentator lines, built from the recorded voices of real actors, so it could add new lines in hours rather than months. The studio has since said it used a similar approach in ARC Raiders, while continuing to pay actors for recording sessions, and it has re-recorded some AI lines with human actors because, in its CEO’s words, “there is a quality difference”.

Localisation

Translating a game into a dozen languages means recasting and re-recording every line. AI voices can make smaller languages viable for studios that couldn’t otherwise afford a full dub, and some tools can keep a character’s voice recognisable across languages. This is also where voice actors’ concerns are sharpest, because localisation has traditionally been steady work for performers in many countries.

Dynamic dialogue and live characters

The newest use is characters that speak lines no one wrote in advance. In May 2025, Fortnite added a Darth Vader who could talk back to players in real time. According to Epic Games, his replies came from Google’s Gemini model and his voice from an ElevenLabs model recreating the late James Earl Jones, with the support of the Jones family. It was a striking demonstration of what’s possible, and it quickly became a test case for the union rules described below.

How brands use AI voiceover for ads

Advertising has different pressures from games. A single campaign might need dozens of cuts: different lengths, products, prices, regions and languages. AI voiceover for ads helps with:

• Versioning at scale. Swapping a price, date or product name without booking another session.
• Localisation. Producing regional versions of the same advert quickly.
• Drafts and animatics. Letting clients hear a script before the final voice is recorded.
• Small budgets. Giving small businesses a professional-sounding voiceover they couldn’t otherwise afford.

When an advert uses a recognisable performer’s voice, or a digital replica of a specific actor, consent and payment rules apply, which brings us to the debate.

AI voice actors and the debate with performers

Voice actors have been among the most vocal critics of generative AI, because their voice is both their instrument and their identity. The concern isn’t AI in general so much as being cloned without permission, or being replaced by a copy of themselves without fair pay.

Key developments with SAG-AFTRA

SAG-AFTRA, the US performers’ union, has set much of the agenda:

• January 2024: SAG-AFTRA announced an agreement with Replica Studios that let members license digital replicas of their voices for video games, with consent, minimum pay and limits on how long a replica could be used before further consent and payment.
• July 2024 to July 2025: SAG-AFTRA’s video game performers went on strike from 26 July 2024, with AI protections a central issue. A deal was reached in June 2025, and members ratified the 2025 Interactive Media Agreement in July 2025 with 95.04% voting in favour. It includes consent and disclosure requirements for AI digital replicas and lets performers suspend consent for new AI-generated material during a strike.
• 2025 Commercials Contracts: the union’s new advertising contracts, put to members in spring 2025, require notice and written consent before a digital replica is created or used, and set extra pay when a replica performs work the actor would otherwise have done in person.
• May 2025: SAG-AFTRA filed an unfair labour practice charge against Llama Productions, an Epic Games subsidiary, over the AI Darth Vader voice in Fortnite, arguing the company used AI to replace bargaining-unit work without bargaining with the union. The estate’s permission didn’t settle the union’s concern, which shows that consent from one person and the wider question of displaced work are separate issues.

These agreements apply to signatory companies and union members, mostly in the US. Rules differ elsewhere, but the same principles are spreading: consent, compensation and control over how a voice is used.

Ethics, consent and disclosure

Whether you’re a large studio or a solo developer, a few principles hold up well:

Consent comes first

Never clone or imitate a real person’s voice without their clear, informed permission. That includes celebrities, voice actors you admire and people you know. Reputable providers make consent a condition of their cloning tools, and some require a recorded consent statement.

Pay people fairly

If an AI voice is built from an actor’s recordings, they should be paid for that use, and know where and how long it will be used.

Be open with your audience

Disclosure is becoming a requirement rather than a courtesy:

• Steam updated its rules in January 2026 so that developers must disclose AI-generated content that players actually see or hear, including audio, while behind-the-scenes productivity tools no longer need disclosing.
• The EU AI Act’s transparency rules have applied since 2 August 2026. Among other things, they require deployers to clearly disclose audio, image or video deepfakes that are artificially generated or manipulated.
• Some TTS providers, such as OpenAI, require you to tell listeners that the voice they’re hearing is AI-generated.

This isn’t legal advice. If you’re publishing commercially, check the rules in your markets and your provider’s terms.

Tips for indie developers and small creators

AI character voices can give a small team a fully voiced game or polished advert on a tiny budget. A few practical tips:

1. Read the licence before you start. Check whether your plan allows commercial use, whether attribution is required, and whether there are limits on games, adverts or broadcast. Free tiers are often non-commercial. Our comparison of text to speech providers covers licensing for the main APIs.
2. Use AI voices where they shine. Placeholder dialogue, barks, tutorials, menu narration and quick ad variants are a natural fit. Emotional story moments usually still benefit from a human performance.
3. Budget for actors where it matters. Many indie games mix AI temp lines during development with a human cast for main characters.
4. Pick distinct voices. Give each character a different age, accent or pitch so players can tell who’s talking without subtitles.
5. Write for the ear. Short lines, clear punctuation and phonetic spellings for invented names make a huge difference. Our guide on how to make text to speech sound natural goes into detail.
6. Keep records. Note which tool, voice and plan you used for each line, so you can answer licensing or disclosure questions later.
7. Disclose clearly. A line in your credits or store page is simple and builds trust.

If you’re making short videos rather than games, our creator’s guide to AI voiceovers for YouTube, TikTok and Shorts covers that workflow.

Trying voices quickly

If you just want to audition voices or produce a few clips, you don’t need an API. MyVoice’s paid plans let you download audio clips in any of its 278 AI voices across 15 languages, and you can hear every voice on our homepage before installing. As with any tool, check the terms that apply to your project before publishing.

FAQ

Do video games use AI voice actors?

Some do, usually alongside human actors. Common uses are temporary dialogue during development, short barks, localisation and live conversational characters. Many studios still record main characters with human performers.

Is it legal to use AI voices in adverts?

Generally yes, if you have the right licence from your voice provider and you aren’t imitating a real person without permission. Union contracts add consent and payment rules when a real performer’s digital replica is used, and disclosure rules apply in some markets, such as the EU.

What did the SAG-AFTRA video game agreement say about AI?

The 2025 Interactive Media Agreement, ratified in July 2025 after a strike that began in July 2024, requires consent and disclosure for AI digital replicas of performers and lets performers suspend consent for new AI-generated material during a strike.

Do I have to disclose AI voices in my game?

On Steam, yes, if players hear AI-generated audio: Valve’s rules require disclosing generative AI content that appears in the game. Other platforms and regions have their own rules, so check where you publish.

← Back to blog