Unreal Speech

Unreal Speech is a text to speech API for developers who need generated audio at volume and at a low unit cost. It converts written text into spoken audio over a simple HTTP interface, begins streaming audio in roughly 300 milliseconds and accepts requests long enough to produce about ten hours of audio in a single job, which suits turning articles, books and long documents into narration. The service exposes several endpoints: a streaming endpoint that returns short passages immediately, a standard synthesis endpoint, an asynchronous task endpoint for long jobs, and a websocket that streams audio and timing data together. Timestamps are returned per word or per sentence with start and end times and character offsets, which makes it straightforward to highlight text in sync with playback inside a reading application. The catalog covers 48 voices across 8 languages, with controls for speed, pitch, bitrate and output format, and the vendor states that the demo runs on the Kokoro model. A browser studio, code samples and an SDK are provided alongside the API. Pricing is monthly and scales with character volume. A free tier gives 250,000 characters, about six hours of audio. Starter is $10 a month for 500,000 characters, Basic is $49 a month for 3 million characters, Plus is $499 a month for 42 million characters, Pro is $1,499 a month for 150 million characters and Enterprise is $4,999 a month for 625 million characters. Volume above a billion characters is quoted individually through a high volume inquiry form, and the published figures convert characters to audio at roughly 750 characters a minute.