CloudPolly

CloudPolly is a self-hosted text-to-speech platform built to be run as a business. It turns written text into natural-sounding audio — audiobooks, podcasts, YouTube and news narration, e-learning and tutorial content, marketing spots, customer support prompts, and applications that talk — and it ships with the accounts, plans, gateways and dashboards needed to charge other people for doing it.

The synthesis itself comes from five vendors: ElevenLabs, Amazon Web Services, Microsoft Azure, Google Cloud Platform and IBM Cloud. You register with one of them or with all of them, your own keys, your own billing with the vendor. With every vendor connected the catalogue runs to over 1300 lifelike voices across more than 146 languages and dialects, so a platform can serve markets far outside its own. Only the languages and voices of the vendors you have activated appear in the app, which also makes it a legitimate way to start on one vendor and add the rest later.

Voices, and the control over them

• Standard and Neural voices — alongside standard TTS, the Neural (NTTS) engines deliver a clear step up in speech quality, including Google's WaveNet voices.
• Speaking styles — where the vendor supports it, neural voices take a delivery style rather than a flat read: a Newscaster style (AWS/Azure) tuned for news narration, and a Conversational style (Azure) suited to two-way and telephony use.
• Voice effects — separate effect combinations for standard and for neural voices, with the applicable set shown against each voice.
• Near real-time synthesis — text comes back as audio without a queue to babysit, and streaming output can be optimised for playback.

SSML without learning SSML

The editor exposes the SSML tag set as controls, so an operator's customers get fine-grained delivery without hand-writing markup. The full list of tags available for the selected voice is shown in the app, since support varies by vendor and voice.

• Rate, pitch and loudness — adjust the speed, tone and volume of any passage.
• Emphasis — lean on the words that carry the sentence.
• Pronunciation — say digits, dates, abbreviations and awkward words the way they are meant to be said.
• Word and phrase replacement — substitute what is spoken without touching the source text.
• Mute and beep out — silence or bleep any part of a sentence.
• Clear effects — strip every applied effect and start the passage again.
• Listen to a selection — preview only the highlighted text instead of re-rendering the whole piece.

Sound Studio and long-form work

• Mix up to 20 voices in a single task — dialogue, interviews and multi-narrator pieces come out as one file, in MP3, OGG or WAV, subject to each vendor's format support.
• Any voice in the catalogue, in one task — the voice picker is not limited to a single vendor per job.
• File merge — the Sound Studio joins audio files in MP3, OGG, WAV and WEBM, whether they were synthesised in one pass or assembled from several.
• Up to 60,000 characters per task — chapters and full articles go through in one pass.
• Text file upload — drop a file in rather than pasting into a box.
• Direct-to-bucket synthesis — large text can be synthesised straight into your Amazon S3 bucket.
• Projects — group results per book, channel or client instead of scrolling one long list.

Output and storage

• Formats — MP3 (AWS, Azure, GCP, IBM, ElevenLabs), OGG (AWS, Azure, GCP, IBM), WAV (GCP, IBM) and WEBM (Azure), with the format options enabled dynamically for whichever voice is selected.
• Storage targets — the local server, Amazon S3 or Wasabi, so audio does not have to accumulate on the app disk.
• Sharing and downloads — results are shareable to social media or downloadable directly, and download rights in the user panel are a switch you control.

The business layer

• Plans — monthly subscriptions, yearly plans, prepaid packs and a free tier, with a "most popular" marker on the plan you want chosen.
• 8 payment gateways — PayPal, Stripe, Razorpay, Paystack and Mollie for subscriptions and prepaid; Braintree and Coinbase for prepaid; and offline bank transfer for both. All of them, and every SaaS feature, are covered by the Regular License.
• Crypto — Coinbase takes Bitcoin, Bitcoin Cash, Ethereum, USD Coin, Litecoin, Dogecoin and Dai on prepaid plans.
• Coupons and promo codes — discount codes for prepaid plans.
• Affiliate and referral system — referral invites, earnings and payout requests, with admin notification when a payout is asked for.
• Finance dashboard — monthly and yearly income at a glance, plus estimated spend on the cloud TTS vendors split by standard and neural characters per vendor, which is the figure that tells you whether a plan is priced above cost.
• Invoices and currency — system currency drives the signs and formatting everywhere, with invoice defaults set from admin.
• Google AdSense — ad slots on the front page for platforms monetising free traffic.

What the operator controls

• Voice curation — enable or disable individual voices, or flip all standard or all neural voices for a whole vendor at once; rename voices to suit your brand; voice avatars and demo samples included.
• Free-tier limits — set the character limit for free users independently of subscribers, and decide whether neural voices are within their reach at all.
• Defaults — pick the language and voice that greet a visitor on the front end and a customer in the panel.
• Content managers — blog, FAQ page, use cases, customer reviews, and Terms and Privacy pages, all editable from the admin panel; notifications, support and the voice catalogue can be hidden entirely.
• Customisation — custom CSS and JS for the front end, footer image, vendor logos shown or hidden, custom URL redirection, registration country default, SMTP with tested sending, social OAuth, and admin login that still works while the app is in maintenance mode.
• Per-user usage — synthesis and listen-mode totals on each customer's profile with usage charts, and the same totals surfaced in the admin user list.
• Operations — database backup, one-click auto update, and support tickets with optional email notification to the admin.

The public side

The front end is a landing page rather than a bare tool: a live listen mode, with autoplay, that lets a visitor hear the product without registering, sample voices, prices pulled dynamically from the plans you configured, voices grouped by country, use cases, customer reviews, FAQ and a blog with SEO-friendly titles and metadata. It is fully responsive, though it is a web application and not a native mobile app.

How it got here

Seven releases, each one documented. v1.1 added the referral system, Wasabi storage, bank transfer and the front-end layout with its blog and live listen mode. v1.2 brought Mollie, Braintree, Razorpay, Paystack and Coinbase. v1.3 added project management, auto update, per-voice enable and disable, voice renaming, the FAQ page and 109 new Azure and IBM voices. v1.4 added another 94 Azure and GCP voices with samples for all of them. v1.5 introduced the Sound Studio, 20-voice mixing, text file upload, the 60,000-character limit and the redesigned front page. v1.6 added yearly and free-tier plans, WEBM merge support, and 60 more Azure and AWS voices across 9 new Azure languages. v1.7 added ElevenLabs covering 29 languages, 72 new Azure neural voices with Armenian and Basque, and 102 new GCP Neural2, Studio and neural voices.

Requirements worth knowing

Built on PHP 7.4 and Laravel 8.4, with comprehensive documentation and six months of included support. At least one cloud vendor account is required — any combination works, and only the activated vendors' voices appear in the app, so reaching the full catalogue means registering with all of them. TTS usage is billed by the vendors at their own published rates, which is exactly what the estimated-spend dashboard is there to track.

Ideal for

• Publishers and content teams narrating articles, newsletters and posts at volume.
• Audiobook and podcast producers who need multi-voice reads out of one task.
• E-learning and training operations localising course audio across dozens of languages.
• Founders launching a paid text-to-speech service with billing, plans and affiliates already built.
• Agencies producing voiceover for video and marketing without booking a studio.
• Accessibility projects giving written content a spoken form for readers who need one.