Unreal Speech

What is Unreal Speech?

Unreal Speech is a text-to-speech API that converts written text into spoken audio, aimed at developers who need large volumes of narration at a low cost per character. Its homepage positions it directly against ElevenLabs on price.

The service works through a handful of HTTP endpoints. A short streaming call returns audio for up to 1,000 characters in about 300 milliseconds, a standard call handles up to 3,000 characters, and an asynchronous task endpoint accepts up to 500,000 characters, which the site says equals roughly 10 hours of audio. A WebSocket endpoint streams audio and word timings together for live captions or highlighting.

The current engine is described on the official site as powered by Kokoro TTS, and one directory lists version 8.0 as released in February 2025. The company says it is based in San Francisco and reports processing 7 billion characters a month with 99.9 per cent uptime. A separate browser Studio lets non-developers paste text, pick a voice and download an MP3.

Key Features

Per-word timestamps and a 10-hour single request set Unreal Speech apart, because they let reading apps sync text and audio without stitching many short clips together.

• Low-latency streaming: The /stream endpoint returns audio for up to 1,000 characters in about 0.3 seconds, according to the official site. That speed suits chat assistants and interactive apps where a reply has to start speaking almost at once.
• Long-form synthesis tasks: The /synthesisTasks endpoint accepts up to 500,000 characters in one asynchronous job. The company says this produces around 10 hours of audio, so a whole book chapter or long report can be converted in a single call.
• Word-level timestamps: Responses can include timings for each word or sentence alongside the audio. Reading apps and e-learning tools use these timings to highlight text as it is spoken or to generate captions.
• WebSocket streaming: A newer /streamWithTimestamps endpoint sends audio and word timings over one WebSocket connection. This avoids separate requests for sound and alignment data, which simplifies real-time players and live caption features.
• Voice and language range: The catalogue lists 48 voices across 8 languages, including US and UK English, Spanish, French, Italian, Portuguese, Japanese, Hindi and Mandarin Chinese. You choose a voice by ID in each request or from a dropdown in the Studio.
• Browser Studio: The Studio lets you paste up to 3,000 characters, pick a voice and generate an MP3 without writing code. Free users get 32 kbps downloads, while the Studio Pro tier offers 320 kbps files and 30,000 characters per voiceover.

Pros

Cost per character is the reason most teams try Unreal Speech, with overage rates published from $8 to $16 per million characters depending on the plan.

• Low unit cost: Published overage rates run from $16 per million characters on Basic down to $8 on Enterprise. High-volume publishers converting articles or books to audio can keep running costs predictable as usage grows.
• Useful free tier: The free plan includes 250,000 characters a month, which the site estimates at about six hours of audio. Developers can build and test a full prototype before paying anything, provided they add the required attribution.
• Built for long text: A single asynchronous request can cover up to 500,000 characters. Audiobook and article-to-audio projects therefore avoid splitting content into hundreds of small calls and reassembling the results afterwards.
• Timing data included: Word and sentence timestamps come with the audio rather than as a paid add-on. This helps teams building read-along, accessibility and language-learning features that need text highlighting.
• Unused characters roll over: The homepage states that paid plans carry unused characters into later months, while the free plan resets monthly. Seasonal publishers with uneven demand waste less of what they pay for.

Cons

Voice quality divides users: several reviews on one directory describe some voices as robotic or flat, even though others praise how realistic they sound.

• Mixed voice quality: User comments on There’s An AI For That include complaints about lifeless delivery and odd pronunciations in British English voices. Test your chosen voice on real content before committing to a large paid plan.
• Steep jump between tiers: After Basic, the next plan is Plus at $499 a month, with Pro at $1,499 and Enterprise at $4,999. Teams that outgrow Basic but need far less than 42 million characters may pay for capacity they never use.
• Introductory Basic price: The $4.99 Basic price is a discount for the first six months, against a listed regular price of $49. Budget for the higher figure if you plan to stay on Basic long term.
• No voice cloning: The homepage states that voice cloning is not currently offered. Brands that need a custom or signature voice will have to look at providers such as ElevenLabs instead.
• Thin privacy detail: The privacy policy covers account and usage data but does not say how long submitted text or generated audio is kept. Organisations handling sensitive scripts should ask the company directly before sending content.

Who is Using Unreal Speech?

Reading and audio-publishing apps make up most of the named users, with the homepage quoting the chief executive of Listening.com on a 75 per cent cost saving.

• Read-aloud apps: Services such as Listening.com convert papers and articles into audio at scale. One directory also lists Readwise and Matter as users, though the official site does not confirm them.
• App developers: Developers add voice output to chatbots, assistants and mobile apps through the streaming endpoint. Low latency keeps spoken replies feeling responsive.
• Audiobook and course creators: Publishers and educators turn long manuscripts and lessons into narration with the asynchronous task endpoint. Timestamps help sync slides or highlighted text.
• Video and social creators: Creators use the browser Studio to produce quick voiceovers without code. They paste a script, choose a voice and download an MP3 for editing.

Pricing

Unreal Speech offers a free API tier and paid plans from $4.99 a month for the first six months (official pricing page, October 2026).

The Studio page separately advertises a Pro voiceover plan at $10 a month after a 3-day trial, which appears to be distinct from the API plans. Futurepedia lists the same API figures but warns that its pricing may be out of date.

Disclaimer: the introductory Basic discount and overage rates can change, so recheck the pricing page when your first six months end.

What Makes Unreal Speech Unique?

A single request of up to 500,000 characters, returned with word timings, is the clearest difference between Unreal Speech and most low-cost TTS APIs. Many rivals cap requests at a few thousand characters, which forces developers to split and stitch long documents.

Pricing is the other difference. The company compares a $49 plan against a much higher ElevenLabs bill for similar volume and claims to be 11 times cheaper. That gap matters less for a short ad script and much more for a library of articles or books, where character counts climb quickly.

Compatibilities and Integrations

REST and WebSocket endpoints are the main way in, and the documentation shows request samples in more than 30 languages and variants, including Python, JavaScript, Go and PHP.

• API: Four endpoints cover streaming, standard, long-form and WebSocket synthesis. Character limits range from 1,000 to 500,000 per request.
• Code samples: The Postman-based documentation generates snippets for Python, Node.js, Java, C#, Ruby, Swift and other languages. No-code connectors were not listed on the pages we read.
• Studio: The browser Studio needs only a sign-in and exports MP3 files. Free downloads are 32 kbps and Pro downloads are 320 kbps.
• Output data: Audio returns with optional word or sentence timestamps. These can drive captions, text highlighting or audio search inside your own app.

Unreal Speech Tutorials

Getting first audio from the API takes six steps and a few minutes once you have an account.

1. Sign up at unrealspeech.com and copy your API key from the dashboard.
2. Open the API documentation and pick an endpoint that fits your text length.
3. Send a test request with a short sentence and a voice ID from the catalogue.
4. Play the returned audio and try other voices and speed settings.
5. Switch to /synthesisTasks for long documents and poll for the finished file.
6. Add the attribution link if you stay on the free plan, or upgrade before launch.

How We Rated It

Unreal Speech scores 3.7 out of 5 because its low per-character pricing and long requests outweigh mixed voice quality and thin privacy detail.

Summary

Unreal Speech suits developers and publishers who need large amounts of narration with word timings at a low cost per character. Its main limitation is voice quality, which some users find flat, and it offers no voice cloning or custom brand voice. Privacy terms say little about retention of submitted text, so check with the company before sending confidential material. We think it is a sensible budget choice for read-aloud and audiobook pipelines rather than for premium character work. Start on the free tier, convert a real chapter with two or three voices, and compare listener feedback before upgrading.