MyVocal AI
MyVocal AI clones a voice from about one minute of audio, then reuses that voice for text-to-speech across 98 documented languages and for AI song covers. A developer API is published; paid pricing is not.
What is MyVocal AI?
MyVocal AI is a synthetic voice platform built around one idea: you clone a voice once, then reuse it everywhere. Three products sit on top of that single cloned voice — Voice Clone, Text to Speech and AI Cover — and the same voice identity carries across all of them.
Cloning starts from a short sample. The publisher asks for at least one minute of clean audio, recommends five, and says five to ten minutes gives the best result. You either record directly in the browser or upload existing files, with no constraint on what is said. Analysis and training are announced at 30 to 60 seconds, after which generation runs in real time. A separate endpoint even lets you design a voice from scratch, with no audio at all.
Language coverage is the platform's strongest verifiable claim. The house model, MyVocal V3, is documented for 98 languages, each listed individually with its code in the public API reference; only Kirghiz and Lingala are explicitly excluded. Older model identifiers remain available for English, Spanish and Portuguese variants. A cloned voice keeps its characteristics when it switches language, which is what makes the tool useful for localisation rather than mere dubbing. Emotion recognition infers intent from the text and reproduces non-verbal sounds such as "um" and "mhm".
AI Cover applies the same voice to singing. You pick or upload an instrumental track and receive a cover in one to three minutes. Training for singing wants isolated vocal stems, because the model does not separate music from voice.
Everything is reachable through a documented REST API — authentication by access key, streaming text-to-speech, callbacks, voice management, history, and a published OpenAPI specification.
Usage is metered in characters per month, with roughly two hours of audio per 100,000 characters. A permanent free plan with a daily quota is active from signup; paid tiers exist and add faster dedicated generation servers, but no price is published anywhere on the public site. The service has been owned and operated since 26 April 2024 by ProVisionary Intelligence Limited, a Hong Kong company, and is delivered as a web application only.
What it does
- Clone a voice from a recording or an uploaded audio file
- Generate speech from text in 98 documented languages
- Create a sung cover of a track using a cloned voice
- Give a conversational agent or application a natural voice through the API
- Build character voices for games, animation and storytelling
- Produce voice-overs for YouTube videos, podcasts and social posts
- Design a brand-new voice without supplying any audio sample
When to use MyVocal AI / When not to
A quick filter to help you decide if MyVocal AI is the right fit.
When to use MyVocal AI
- Content creators producing voice-overs for YouTube videos, podcasts and social media without recording every take themselves
- Musicians and hobbyists who want to sing a cover in their own cloned voice, in any language
- Developers embedding speech generation in an app or giving a conversational agent a natural voice through the REST API
- Localisation and translation teams that need the same vocal identity to carry across dozens of languages
- Game and animation studios building distinct character voices without booking a studio session
When not to use MyVocal AI
- Buyers who must budget before signing up, since no paid price is published anywhere on the site
- Procurement and compliance teams needing a DPA, a subprocessor list or a postal address before purchasing
- Anyone wanting to turn a full song with backing music into a cover, as the model does not separate vocals from instrumentals
- Users who expect a native mobile app, since only a web application and an API are offered
- Organisations requiring a vendor that states its own GDPR compliance in writing rather than staying silent on it
How to use MyVocal AI
A typical end-to-end flow, from setup to results.
- Create an account with an email login link or a Google sign-in
- Choose between recording a sample directly on the site and uploading existing audio files
- Supply at least one minute of clean, low-noise speech; five to ten minutes gives the best fidelity
- Wait roughly 30 to 60 seconds while the voice is analysed and trained
- Open Text to Speech, paste your text and pick your cloned voice or a premade one
- Select the output language among the 98 documented codes, or let the model detect it
- Generate, then listen and download the resulting audio
- For a cover, open AI Cover, select a song or upload your own instrumental, and generate
- Find past generations in History and manage trained voices in My Voice
- For programmatic use, take an access key and call the REST endpoints, with streaming and callbacks available
Pros & Cons
Pros
- 98 languages are listed one by one in the public API reference, so the multilingual claim can actually be checked
- One cloned voice serves speech, singing and conversational agents without retraining
- A permanent free plan with a daily quota lets you evaluate the output before any commitment
- The API is unusually complete for a tool of this size: streaming, callbacks and an OpenAPI specification
- The privacy policy states that voice recordings are not used to train AI models
- Personal data is hosted in the United Kingdom and the EEA, with adequacy and standard contractual clauses cited
- A low entry barrier: one minute of audio and roughly a minute of processing
Cons
- No paid price is published anywhere — no pricing page exists and the subscription policy links to one that cannot be reached
- No postal address appears on any page, and the publisher is named only inside the terms of service
- A single email address is the only contact channel; there is no contact page or form
- The security certifications on display belong to Amazon Web Services, not to MyVocal AI itself
- No data processing agreement and no formal subprocessor list are published
- The site contradicts itself on languages, advertising eight in its FAQ against 98 in the API documentation
- Editorial upkeep is poor: the FAQ still refers to "Speakperfect.ai", and two features announced in 2024 have left no trace
Pricing & Plans
A permanent free plan is available and is applied by default on registration, with a daily generation quota rather than a time-limited trial. Paid subscriptions exist and are described by the subscription policy and the FAQ, which mention a larger character allowance and access to dedicated, faster generation servers. However, no price point is published on the public site: there is no pricing page, the subscription policy refers to one that cannot be reached without an account, and no amount appears in the sitemap or in the page source. Consequently no entry price and no currency can be stated here. Billing is metered in characters per month and processed through Stripe.
- Free plan — applied by default at registration
- permanent
- limited by a daily generation quota that resets each day
- Paid subscription — referenced by the subscription policy and the FAQ
- adding a larger monthly character allowance and dedicated faster generation servers
- but neither named nor priced publicly
- Additional character packs — purchased separately
- consumed only once the subscription allowance is exhausted
- carried over between periods
- and frozen if the account is downgraded to the free plan
Data, GDPR & hosting
A consolidated view of how MyVocal AI handles your data.
GDPR overview
MyVocal AI never states that it is GDPR compliant. That silence, rather than any denial, is why this record marks compliance as no. The only occurrence of "GDPR" on the site sits in a list of certifications attributed to Amazon Web Services on the data security page, and describes the host, not the publisher. The substance is nonetheless partly there: the privacy policy, effective 26 April 2024, grants access, rectification, erasure, portability, withdrawal of consent, objection and restriction, plus the right to complain to a supervisory authority. Personal data is stored in the United Kingdom and the EEA, and transfers rely on the UK adequacy decision and standard contractual clauses. No data protection officer and no Article 27 EU representative are named; requests go to a single support address.
Who owns the data?
You keep ownership of the audio you upload, and you must warrant that you hold the rights to it. By submitting it you grant ProVisionary Intelligence Limited a royalty-free licence to use, modify and process that content in order to build the AI voice models. The company and its licensors own the service itself, the resulting AI models and all underlying technology, and you receive only a limited, revocable, non-exclusive licence to use them. Any feedback you send becomes the company's exclusive property. On termination you may request destruction of your models and content, though the vendor keeps rights over aggregated, anonymised metadata.
Reuse rights
Cloned voices may be used commercially, but only if you genuinely hold the rights to the voice — your own, or one you have explicit permission and a licence to use. The vendor states plainly that cloning is lawful only with the consent of the person concerned, and it performs no technical check of its own. AI song covers are treated more cautiously: usage rights follow the original song's copyright, and the publisher itself recommends personal use unless you have secured a licence. Voice-overs for YouTube are described as acceptable in most cases, subject to your plan licence and the platform's own rules. Responsibility for the legality of both uploads and generated output rests entirely with you. On the vendor's side, voice recordings are declared not to be used for training AI models; account, usage and payment data serve service delivery, personalisation, marketing, analytics, fraud prevention and legal compliance.
Data retention & training
Hosting summary
Personal data is processed and stored on servers located in the United Kingdom and the European Economic Area. AI models and voice data are held on Amazon Web Services infrastructure. For users outside the EEA, transfers are said to rely on the UK adequacy decision and on standard contractual clauses. The publisher describes automatic encryption at the physical layer across the AWS network before data leaves secured facilities, service-to-service TLS, and HTTPS with SHA-2 compliant cipher suites. Access points are protected by two-factor authentication with role segregation and periodic risk review. One caveat matters: the compliance list shown on the data security page — PCI-DSS level 1, SOC 1, SOC 2, FedRAMP, GDPR, HIPAA/HITECH — is presented as Amazon Web Services' certification portfolio, not as MyVocal AI's own. Third-party cookie and analytics providers named on the site, including Google Analytics, Fullstory and Intercom, process data in the United States.
Things to keep in mind
Risks and trade-offs to weigh before adopting MyVocal AI.
- Cloning someone else's voice requires their explicit consent; the publisher says so but performs no technical verification, leaving impersonation entirely to your conscience
- The voices are described as nearly indistinguishable from real human speech, which makes misleading a listener — or enabling voice fraud — a genuine risk
- You grant the publisher a royalty-free licence over the audio you upload, and any feedback you send becomes its exclusive property
- Deletion of your models and content is available on request after termination, not automatically, and aggregated anonymised metadata is retained regardless
- Song covers carry copyright exposure that the tool does not clear for you: the publisher itself advises personal use unless you hold a licence
- Security badges on the site describe Amazon Web Services' certifications, and reading them as MyVocal AI's own would overstate the vendor's assurances
- Leaning on a cloned voice for routine narration can quietly erode your own recording practice and the judgement that comes with it
Setup & Integrations
Technical difficulty
Very low for ordinary use. The tool runs entirely in the browser with no installation; you sign in with an email link or a Google account, then record or upload a sample. The only real requirement is one to five minutes of clean, low-noise audio that you have the right to use. Integration is moderately easy for developers: REST endpoints authenticated by an access key, streaming output, callbacks and a published OpenAPI specification. Two practical constraints apply — file-related endpoints are not supported through plain direct calls, and AI Cover training expects isolated vocal stems rather than full mixes.
Deployment
Supported languages
Behind MyVocal AI
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
How much audio do I need to clone a voice?
How many languages does MyVocal AI actually support?
How long does cloning take, and how fast is generation?
Is there a free plan, and how much do paid plans cost?
Can I use a cloned voice commercially?
Are my voice recordings used to train the models?
Where is my data stored?
Is there an API?
How do I contact the publisher?
Is there a minimum age?
Should you pick MyVocal AI?
MyVocal AI is a real, working product rather than a landing page, and the evidence for that is unusually concrete: a public API reference that names 98 languages one by one, documented endpoints, an OpenAPI specification, and stated timings and quotas. The single-voice design is the genuinely useful idea here — clone once, then speak, sing and drive a conversational agent with the same vocal identity, in almost any language. For a content creator, a small studio or a developer wanting speech in an application, the permanent free plan makes evaluation cheap and immediate.
The reservations are commercial and editorial rather than technical. No paid price is published anywhere, which makes budgeting impossible before signup and is the single biggest obstacle for anyone buying on behalf of an organisation. There is no postal address, no contact page and no formal data processing agreement, and the publisher — ProVisionary Intelligence Limited, a Hong Kong company operating under Singapore law — is named only inside the terms of service. The security certifications shown on the site belong to Amazon Web Services, not to MyVocal AI, and the site never claims GDPR compliance for itself even though it grants the corresponding rights. Upkeep is patchy too: the FAQ still names a different product, the advertised language count contradicts the documentation, and two features announced in 2024 have left no trace since.
Try it on the free quota, judge the output on your own voice, and treat the absence of published pricing and legal paperwork as something to settle with the publisher before you depend on it.
- Choosing a selection results in a full page refresh.
- Opens in a new window.