Voice cloning for indie games: NPC dialogue and multilingual localization

November 2, 2026 · 10 min read · By Resona Editorial Team · Resona Blog

A lot of indie games ship without voice acting. Usually, it's not because the developers don't want it. Recording dialogue takes time, coordination, and money that a small team can't easily spare.

Even a single recording session means finding an actor, preparing a quiet space, recording multiple takes, and editing the results. A friend of mine once spent a whole weekend recording shopkeeper lines for a demo. Two days later the quest got rewritten, and that audio is now a folder on a drive nobody opens. Then there's localization. Add a few languages, and you're looking at separate actors and recording sessions for each one.

Voice cloning doesn't remove all of that work, but it can make the process much more manageable. For a small team — or a solo developer — it offers a practical way to give NPCs distinct voices, revise dialogue without scheduling another session, and carry a character's voice into multiple languages.

Here's what that workflow can actually look like.

What changes for a two-person team

The traditional process is fairly straightforward, but it has a lot of moving parts. You cast an actor, schedule a session, find somewhere quiet to record, work through multiple takes, and clean up the audio afterward.

Then the script changes.

For a game with a few voiced characters, that's inconvenient. (If you've never had to ask a friend to re-record six lines because you spotted a typo, this all sounds trivial. It is not.) For a game with dozens of NPCs and branching dialogue, revisions can quickly become one of the biggest parts of the audio workload.

With voice cloning, you can create a reusable voice from a clean sample of speech. Depending on the service and the voice, that might only require a short recording. Once the voice is set up, new dialogue can be generated without bringing the actor back into the studio for every script revision.

There's also the option of starting with an existing voice library. If you don't need a completely custom voice, you can browse a library of more than two million ready-made voices by characteristics such as age, accent, or style and assign them to different characters.

For some small projects, that's often enough to get a basic voice pipeline running without treating audio production as a separate department.

NPC dialogue without a recording studio

One of the most useful parts of synthetic voices is the revision loop.

Suppose a line works on paper but sounds awkward once it's placed in a scene. With traditional recording, changing it means booking another session. With a cloned voice, you can edit the line and generate a new version while keeping the same character voice.

This becomes especially useful with branching dialogue.

Imagine a guard with three states: calm, alert, and alarmed. You can write variations for each state, use delivery cues such as [calm], [tense], or [panicked], and compare the results in context.

If one version doesn't land, you only need to change that line. The other dialogue stays untouched.

The same approach makes it easier to give NPCs different voices. A blacksmith, a quest giver, and a background character don't need to share the same default voice simply because recording a large cast isn't practical.

Players rarely register each individual voice on a conscious level, but variety helps a game world feel less repetitive — especially in dialogue-heavy games.

Multilingual localization without starting over each time

Localization creates a different kind of problem.

Translating dialogue is only one part of localizing voiced content. Traditionally, every additional language means new actors, new recording sessions, and another round of editing and direction.

A multilingual voice system can simplify that process. If the same voice model supports 83 languages, you can translate the script and generate localized dialogue without rebuilding the entire recording pipeline for each market.

That can help preserve continuity between versions of the game. A companion character can retain recognizable vocal characteristics whether the player is hearing the English, German, French, Spanish, or Japanese version.

One thing worth knowing before you commit: accents and pronunciation don't automatically become culturally native just because the text changes languages.

A voice cloned from an English sample often carries characteristics of that original voice when speaking Japanese or another language. For a robot, a fantasy setting, or a deliberately stylized character, that's usually fine. But if a game is closely tied to a specific culture or setting, accent and language performance should be treated as creative decisions, not something to leave entirely to the generation system.

In those cases, recording suitable source material in the target language usually produces better results.

The financial difference can still be significant. Traditional dubbing costs tend to grow with every language because each version requires people and studio time. With a voice generation workflow, the cost is more closely tied to how much audio you generate.

That makes multilingual voice acting more realistic for projects that wouldn't otherwise have the budget for it.

Who you can't clone

The first voice I ever cloned was my own, which felt like the safest possible test subject. That's the spirit of this whole section: don't clone someone's voice without their explicit permission and the appropriate rights to use it.

That includes celebrities, voice actors, and anyone else whose voice you don't have permission to reproduce.

For game development, there usually isn't much reason to build around impersonation anyway. The more interesting opportunity is creating voices for original characters.

Instead of copying a specific person, you can start with the qualities you're looking for — for example, a warm voice, a mid-range register, or a slightly breathy delivery — and create a voice for that character.

The goal should be to build a cast that belongs to your game, rather than borrow the identity of someone outside it.

If you're planning to use voice cloning commercially, it's worth understanding the rights and consent side before recording anything. Our guide to voice cloning and consent basics covers some of the issues to consider.

Where Resona fits

For developers working with a limited budget, pricing is part of the production decision.

Resona is paid-only and uses a credit-based model rather than a subscription. Credit packs start at $9, with $1 equal to 50 credits, and unused credits don't expire.

Cloning a voice costs 100 credits — roughly $2 — and that's a one-time cost for each voice rather than a charge for every line.

Speech generation is priced at 4,200 credits per million UTF-8 bytes. For typical dialogue, the actual cost depends on the length of your script and how many revisions you generate.

The console shows the cost before generation, so you can check how much a batch of dialogue will use before running it. That's useful when you're working through a larger localization or NPC dialogue pass.

Because the system is pay-as-you-go, the cost scales with the amount of audio you generate. A small game with a handful of voiced characters doesn't have to pay for capacity intended for a much larger production.

If you want to check the numbers against your own script, open the console and paste a few lines in. Recording a twenty-second sample still takes longer than generating an hour of dialogue, and I don't expect that sentence to ever stop being slightly absurd.

Related reading

Sources