Create voiceovers from text, arrange multivoice projects, and make a voice from consenting speaker samples.
Used this tool? Rate it
Used this tool? Rate it
Voicemaker is an online text-to-speech service for making downloadable voice recordings from written text. It combines a voice catalog with controls for delivery, a multivoice project editor and tools for creating a voice from samples supplied by a consenting speaker. Its website promotes more than 2,000 AI voices across 130 languages and accents; those are catalog-wide figures, not a promise that every model or individual voice covers every language.
The product is intended for people producing speech as well as developers integrating speech generation. Projects can be converted in the web interface, while a separate usage-based developer entry lists APIs for text to speech, speech to text and speech to speech. Available voices, conversion lengths, team access and credit allowances differ by plan. For accounts without a paid subscription, free and trial output has narrower rights than output generated during a paid subscription.
The model guide positions Turbo for low-latency conversational use, while High-Res emphasizes emotional delivery and pacing. It lists over 30 Turbo languages and a response time under 100 ms. Expressive is described as a prompt-driven model for changing tone and mood, including effects such as laughter or whispers. These are vendor descriptions rather than an independently measured guarantee of voice quality or latency. Model-level language coverage is narrower than the overall catalog and varies by family.
The speech-to-speech guide limits conversion to Pro+ and custom cloned voices. The controls also include speed, pitch and volume. A pronunciation editor is promoted for making names and other recurring terms more consistent. Voicemaker provides previewable voices, so different delivery options can be heard in the product rather than inferred from the catalog's total voice count.
Voice cloning uses uploaded samples or audio recorded from the device. The instructions say a voice can be built from up to 25 files totaling under 10 MB, with at least one minute of audio. They ask for details such as the voice name and description, and the submitted recording must belong to a speaker who has expressly agreed to the creation and use of that voice. The guide says that processing usually takes less than five minutes, but does not promise that time for every request.
The cloning guide says its background-music removal can remove only minimal background audio. That is a practical limit on the samples used to create a clone, not a general claim that all noise can be cleaned. The service terms distinguish consent to train an account-specific cloning voice from the separate consent needed to place training audio into a research dataset for general voice technology.
A Multi Editor project can combine clips using different voices, including voices in different languages. The instructions describe generating a clip, adding it to a track and combining the selected clips for preview and export. They specify up to 100 audio files in this process; this is a limit on audio files, not a promise of 100 simultaneously available speakers or voices.
The export guide lists MP3, WAV, OGG, AAC and OPUS as selectable formats, with sample rates from 8,000 to 48,000 Hz for telephony and higher-quality production. The pricing comparison separately connects some export-quality options to particular plans. Format availability is therefore a different question from whether a given tier includes a higher bitrate or sample rate.
The documented project flow is useful for dialogue or a script divided among speakers:
Voicemaker presents its voice catalog for video narration and other speech projects. A creator can turn a written script into downloadable audio, adjust delivery and choose a voice for a target language.
A developer can use the separately listed API entry for programmatic text-to-speech, speech-to-text or speech-to-speech workflows. The model guide presents Turbo as a low-latency option for conversational AI, IVR and chatbots; this is a stated positioning, not a benchmark measured by Toolso. API access is separately described as usage-based, so the web subscription table alone does not specify the cost of an integration.
Teams and Business are designed for shared production rather than just a larger individual allowance. Embedded plan data lists a shared workspace, member controls and usage reporting; actual plan availability depends on the selected tier. The subscription terms make the account responsible for authorized users and for removing access when appropriate.
A team handling shared audio projects may value the administrative controls offered with team plans, including member management and team credit controls. An organization seeking negotiated availability commitments, a data processing addendum or dedicated infrastructure needs to distinguish those contract-specific provisions from the standard self-service terms. The latter say they do not themselves include a service-level agreement.
The service is less suitable where a buyer needs guaranteed acceptance by a particular distributor or rights organization: the terms expressly decline to warrant acceptance by third-party platforms and audiobook or advertising programs. The terms also state that the service is not designed or licensed for life-critical, emergency or medical-diagnostic use where an inaccurate result could cause severe harm.
Voicemaker's public entry is a web service with an online editor, project storage and downloadable audio. Its terms cover the website and APIs among other service interfaces. The developer entry advertises REST APIs for speech workflows. Cloud file history is advertised as syncing across devices. A separate API account or billing arrangement should not be assumed to match the exact credits shown on an individual web subscription.
The public pricing page presents Free, Starter, Creator and Pro individual tiers, plus Teams, Business and a custom Enterprise offering. Prices and allowances vary by billing option and settings. At the November 12, 2026 capture, a Starter entry showed $5 and 200,000 credits per month, while Creator showed $10; the static comparison listed 500,000 monthly Creator credits but embedded card data listed 400,000. Those two figures are inconsistent, so neither should be treated as a verified checkout allowance for a selected account.
| Access | Publicly described boundary |
|---|---|
| Free | Limited generations and 250 characters per conversion; personal use in the plan comparison |
| Starter | 200,000 monthly credits and 3,000 characters per conversion in the comparison |
| Creator | 5,000 characters per conversion; the captured page gives conflicting monthly credit figures |
| Pro | Monthly rollover for one billing cycle; credit capacity depends on the offered selection |
| Teams and Business | Shared administration and team credit controls; eligible unused credits can roll over for up to three months |
| Enterprise | Custom commercial terms and options shown for larger or security-sensitive deployments |
The homepage promotes “Unlimited Free Standard TTS”. The pricing FAQ says every plan has a credit limit and rejects unlimited TTS. The statements do not explain how their scopes reconcile.
The plan comparison describes credits as a shared balance across tools and models. Default voices and FlashX cost one credit per character in the rate list, Pro V2 and the listed Turbo variants cost two, and Expressive or High-Res ProPlus voices cost four. The model guide gives conflicting Turbo multipliers: 3× in its summary and about 2× later on the same page. Speech-to-speech is listed at 100 credits per second and speech-to-text at 10 credits per second.
The FAQ says Chinese, Japanese and Korean characters each cost two credits because of their processing requirements. It does not explain how that rate combines with model-specific rates. The current pricing FAQ says credits are counted on conversion rather than downloads. A January 2026 v1.8 changelog says changing volume, speed or pitch under Voice Settings V2 did not use credits if text and voice were unchanged; current applicability is unverified. The site provides an illustrative estimate of about ten hours from 500,000 credits, while saying the actual duration varies with voice, language and speaking speed.
An add-on is listed at $20 for one million credits, valid for one year and available on paid plans. A separate clone-slot add-on is listed at $2 per year for an eligible plan.
The terms say monthly subscriptions receive a new allowance each billing cycle. Annual Starter, Creator and Pro receive their annual credits upfront, whereas annual Teams and Business receive monthly allocations from the yearly total. Starter and Creator have no rollover on either monthly or yearly plans. Pro monthly can carry unused credits for one cycle; Teams and Business can carry eligible credits for up to three months, subject to the stated annual-renewal reset. Unused and carried credits are cleared if the subscription expires without renewal.
For card subscriptions processed through Stripe, auto-pay is enabled by default. Turning it off prevents the next renewal but does not end or refund the current period. The refund policy says PayPal and Razorpay plans expire at the end of their duration without automatic renewal.
The refund policy limits standard eligibility to a first purchase requested within five days and excludes discounted or promotional purchases, renewals and upgrades. If eligible, the listed schedule gives a full refund when no credits have been used and deducts $2, $4, $6 or $8 across successive 50,000-credit bands. Above 200,000 used credits, the policy says no refund applies. Cancellation alone is not described as a partial refund of the remaining subscription.
ElevenLabs offers its own text-to-speech and voice-creation products, including tools for creating expressive spoken audio. Murf offers a speech editor with controls over voice delivery and also presents a developer API. Each service has its own plans and rights terms.
Voicemaker's terms say a customer retains rights in submitted materials and the generated audio, while Voicemaker retains ownership of its service and licensed voices. The service receives a limited license to host and process customer content for operating and securing the service. This allocation does not eliminate the customer's responsibility to have rights in submitted material or consent from any speaker whose voice is cloned.
The terms say automated output can contain errors, may not be unique and may resemble output produced for other users. They do not guarantee acceptance by third-party platforms, distributors or rights organizations.
The terms state that customer content and output are not used to train Voicemaker's general or publicly available voice models. They allow processing for the requested results and for operating, securing and billing the service, as well as de-identified operational metrics. The account-specific clone is trained on the submitted voice samples. The terms require separate explicit consent before training audio is added to general research datasets, and say that declining this research use does not remove the ability to create the account's clone.
The public security policy uses two descriptions that are not identical: its introduction says text and audio are not used to train or improve models without explicit consent, while its project-information section describes using anonymized project information for model training. The pricing FAQ’s captured HTML gives a broader no-training description for input text and generated audio; its collapsed answer was not interactively verified. These statements have different scopes and cannot be collapsed into a single unconditional promise.
With customer permission, the cloning terms allow employees and contractors to listen to samples for quality checks, and allow employees to create non-defamatory internal quality-assurance utterances. The same section describes access and sharing restrictions for the clone, with exceptions under the privacy policy. A team using samples should distinguish these stated terms from an independent audit of how its particular data is processed.
The terms set a minimum age of 16. If law requires a higher age for Voicemaker to lawfully provide the service, including personal data use, without parental consent, the minimum age is that higher age.
The privacy policy describes automatic erasure of typed text and converted audio after 60 minutes depending on user settings. The terms address a different event: after termination or account closure, customer content and account data are deleted from production systems within thirty days, subject to legal, security and transaction-record exceptions and encrypted backup cycles. The security policy separately describes deletion of deleted projects from servers and backups within thirty days. These provisions concern different triggers and categories and should not be treated as a single universal retention clock.
Working storage is supplied during an active subscription, but the terms explicitly say Voicemaker is not a long-term archival service and may delete unused or expired files. They also say the self-service terms do not include an SLA; availability commitments apply only if included in a separate written agreement.
A cached company review profile of uncertain age showed 4.3 out of 5 from 41 reviews and warned that the sample may not be representative. Direct and browser-based source checks were blocked. That score summarizes a company profile, not a controlled test of an individual voice model or proof of any particular product limitation.
The terms permit commercial use and distribution of voices created on paid plans, while excluding resale of the service's offerings. Free and trial output is limited to personal, internal-evaluation or other noncommercial use unless the account has a paid subscription. Output lawfully downloaded during a paid subscription may continue to be used commercially after it ends, subject to the terms.
Yes. The terms define a consenting speaker as someone who has explicitly agreed both to the use of their recordings to train the clone and to the generation and use of their voice as described in the terms. They prohibit submitting another speaker's recording without that consent.
No. Rollover depends on the plan: the terms allow one cycle for Pro monthly and up to three months for Teams and Business, with the stated reset at annual renewal. Starter and Creator have no rollover on monthly or yearly plans, and Pro yearly has none. Remaining unused credits are cleared if the subscription expires without renewal.
