Altered AI

What Is Altered AI?

A voice actor needs to play five characters in a single session. A game developer wants a protagonist who sounds authentically American, a villain who sounds Mid-Atlantic, and an NPC who sounds Eastern European — without booking three separate recording sessions. A podcaster records a rough interview and wants to clean up their stumbling, uncertain delivery without re-recording anything. These are the kinds of problems that standard text-to-speech tools do not solve, because TTS starts from text, not from a human voice already performing.

Altered AI starts from performance. The London-based company — formally Altered Ltd, registered at Companies House under number 11564329 since September 2018 — built its product around speech-to-speech transformation: you speak, and the system changes what comes out. Voice morphing, accent conversion, voice cloning, and speech fluency enhancement are the core capabilities of Altered Studio, its desktop application.

The team is small, around 14 people, which means the product has clear edges to it. Enterprise SLAs are not a realistic expectation. The roadmap moves carefully. What Altered lacks in scale, it compensates for with focus: it has built genuine differentiation in a specific niche that larger voice AI companies tend to ignore, and it has done so from a UK base where UK GDPR and serious privacy commitments come with the territory.

Key Features

Real-Time Speech-to-Speech Voice Morphing

The core Altered Studio capability is live voice transformation. You speak into a microphone; the system converts your voice into a different voice in real time, preserving your performance — your timing, your emotion, your pacing — while replacing the voice itself. This is meaningfully different from text-to-speech: you are not generating speech from written words, you are performing speech and having it transformed.

The practical use cases are clear. Game developers fill placeholder voices before final casting is confirmed. For voice actors, it opens up multiple character interpretations from a single recording session. Content creators experimenting with persona differentiation find it equally direct to apply. The real-time aspect means you can monitor the transformation live and adjust your performance accordingly, rather than generating audio and listening back iteratively.

Voice Cloning

Altered allows you to clone a voice from audio samples and use that clone as a target for morphing. You speak, and the output sounds like the cloned voice performing what you said. This is useful for maintaining character consistency across long recording sessions, or for creating a branded voice that represents a product or IP. Cloning is available on Creator plans ($65/month) and above.

Voice cloning from recorded audio samples sits in a different technical category than real-time morphing — it requires an enrolment process — but Altered integrates both within the same Studio workflow. The result is a pipeline where you can record samples, create a clone, and then morph live performance into that clone in the same session.

Local Processing Mode

On paid plans, Altered Studio processes audio locally on your machine. Nothing is uploaded to Altered's servers. This is a significant privacy guarantee for users handling sensitive content — voice actors working on confidential productions, journalists with protected sources, therapists or clinicians using voice tools with patient audio. The free tier processes audio in the cloud with consent to anonymised R&D use; paid tiers flip this to complete local privacy.

For UK GDPR purposes, voice data qualifies as biometric data in many jurisdictions. Local processing means Altered is never in possession of your audio — the data never travels. This is a structural privacy advantage that most cloud-only voice AI platforms cannot offer.

Accent Conversion

Altered's accent conversion feature allows you to shift the regional accent of a recording without changing the voice identity. A speaker recorded in a strong regional British accent can be converted to a neutral RP accent; an American accent can be shifted toward General American or Mid-Atlantic. This is a rare capability at Altered's price point. Larger speech companies offer accent neutralisation for accessibility purposes; Altered positions it as a creative tool.

The results vary depending on accent distance — closely related accents convert more convincingly than distant ones — but for production use cases where the alternative is re-casting or extensive reshoots, even approximate conversion has significant value.

Euphonia Speech Fluency Models

This is Altered's most distinctive and ethically interesting feature. Euphonia models are trained to help users with voice disfluencies — stuttering, cluttering, dysarthria, or acquired speech conditions — produce smoother, more fluent-sounding output. A person who stutters can speak naturally and have the output transformed to remove the stutter, while retaining the character and identity of their voice.

Few commercial voice AI tools prioritise accessibility at this level. It positions Altered within a broader conversation about voice as an assistive technology, aligning it with tools like Acapela Group's My-Own-Voice service, albeit from a different technical angle. Euphonia is available across paid tiers.

Pricing

Altered Studio's pricing structure has four meaningful tiers. The free plan gives access to basic cloud-processed voice transformation for evaluation — useful to verify whether the product solves your specific problem. It processes audio in the cloud and requires consent to anonymised use for Altered's research and development.

The Starter plan at $5/month is an entry point, though it does not unlock local processing or advanced features. The Creator plan at $65/month is where the product becomes genuinely useful for professional workflows — it includes local processing, voice cloning, accent conversion, and commercial usage rights. At $65/month, it is expensive relative to basic TTS tools, but appropriate compared to the cost of human re-recording or additional casting sessions it displaces.

The Professional plan at $200/month adds higher usage limits and advanced customisation, suited for teams or high-volume production pipelines. Enterprise pricing is custom, with API access unlocked at that tier for teams building Altered's capabilities into larger workflows. Annual billing reduces costs meaningfully — Altered offers quarterly and annual billing options with discounts over monthly rates.

A 7-day trial is available for the Creator plan, which is long enough to run a meaningful test of the morphing and cloning workflow on real production material.

EU Compliance & Privacy

Altered AI is a UK company, which means UK GDPR — the post-Brexit parallel to EU GDPR — governs its data handling. UK GDPR is functionally equivalent to EU GDPR in most respects, though the two are now separate legal frameworks. From a compliance perspective, EU organisations working with Altered are dealing with a third-country data processor, which requires appropriate contractual safeguards (such as a Data Processing Agreement and Standard Contractual Clauses).

In practice, the local processing mode on paid plans is the most meaningful compliance feature Altered offers. When voice audio never leaves your device, the question of where Altered hosts its servers becomes largely academic. For the free tier, audio is processed on Altered's cloud infrastructure — the terms require consent to anonymised use for model training and improvement.

Voice cloning data is biometric data under GDPR and UK GDPR. Altered's terms require users to have appropriate consent from any individual whose voice is being cloned. Compliance tooling for managing this consent is not built into Studio — developers and producers need to manage it externally.

For EU organisations with strict data sovereignty requirements, Altered's non-EU status is a genuine limitation compared to providers like Gradium or Acapela Group. Local processing helps. A legal basis review is advisable before deploying Altered in regulated EU contexts.

Who It's Best For

Voice actors and audio producers who need to experiment with character voices, fill placeholder audio, or cover multiple roles without multiple sessions. Speech-to-speech morphing solves a specific production problem that TTS cannot address.

Game developers and narrative content creators working on projects that require voice consistency across long recording sessions or iterative creative development. Voice cloning from samples makes character voice maintenance tractable at small team scale.

Users with voice disfluencies who want to communicate more fluently without losing the identity and character of their natural voice. Euphonia is a serious assistive feature, not a gimmick.

If you are primarily building developer APIs for production applications, want an EU-based data processor, or need an enterprise-grade SLA with a dedicated support team, Altered is not the right fit. Its strength is in creative professional tools, not infrastructure. ElevenLabs has better API developer experience; Gradium has lower latency for real-time agents; Acapela has deeper EU compliance credentials and enterprise history.

The Verdict

Altered AI occupies a specific and defensible niche: speech-to-speech transformation for creative professionals who perform rather than type. The real-time morphing, local processing mode, and Euphonia accessibility features are genuine differentiators in a crowded voice AI market. The limitations are equally real — 14 people, no EU base, a steep Creator plan price, and thin integrations outside the desktop application itself. For its target users, those trade-offs are acceptable. For teams that need EU-native compliance, API-first architecture, or enterprise reliability, Altered is a specialist tool rather than a platform.

Frequently Asked Questions

Is Altered AI GDPR compliant?

Altered AI is a UK company subject to UK GDPR, which is a post-Brexit equivalent to EU GDPR with largely the same protections. On paid plans, local processing mode keeps all voice audio on your device — nothing is uploaded to Altered's servers. EU organisations should review their data processing agreements since the UK is now a third country under EU GDPR.

Can Altered AI run locally without sending data to the cloud?

Yes, on paid plans ($65/month and above). Local processing mode runs voice transformation entirely on your machine, with audio never leaving your device. The free tier uses cloud processing and requires consent to anonymised data use for model training. Local mode is the most meaningful privacy guarantee Altered offers.

How does Altered AI compare to ElevenLabs?

Altered AI specialises in speech-to-speech voice morphing — transforming live spoken performance into a different voice in real time. ElevenLabs focuses on text-to-speech: generating high-quality audio from written text. They solve different problems. If you perform audio, Altered is the better fit. If you generate audio from text, ElevenLabs is more capable.

Does Altered AI have a free plan?

Yes. The free plan gives access to basic cloud-processed voice transformation. To unlock local processing, voice cloning, accent conversion, and commercial usage rights, you need the Creator plan at $65/month. A 7-day trial of the Creator plan is available without committing to a full subscription.

What is Euphonia in Altered AI?

Euphonia is Altered's assistive voice model that helps users with voice disfluencies — including stuttering, cluttering, or dysarthria — produce smoother, more fluent-sounding output while retaining their natural voice identity. It processes speech in real time and applies fluency enhancement before output. Few commercial voice AI tools have invested in this type of accessibility capability at this level of specificity.