Unreal Speech
A developer-oriented text-to-speech API with low-latency streaming, word-level timestamps, and batch synthesis of up to 500K characters, billed in tiers by monthly character volume.
About Unreal Speech
Unreal Speech is a developer-oriented text-to-speech (TTS) API. It offers /stream, /speech, and /synthesisTasks endpoints, supporting low-latency streaming of up to 1,000 characters, synchronous synthesis of up to 3,000 characters, and asynchronous batch synthesis tasks of up to 500,000 characters respectively; it can also stream audio along with word-level timestamps via a websocket endpoint. According to the official homepage, its streaming audio latency is 300ms, and it supports audio requests of up to 10 hours. Voices cover male and female speakers in American English, Chinese, Spanish, French, Hindi, Italian, and more. Pricing is tiered by monthly character volume, from a free 250K characters up to 625M characters on the Enterprise plan; custom plans for very large usage are also available by contacting sales.
Best for
Best use cases
Adding real-time voice replies to chatbots (via the /stream endpoint)
Batch-synthesizing audio for audiobooks, news, blogs, or other long conversational text
Generating audio with word-level timestamps for subtitles or text highlighting
Synthesizing speech in Chinese, English, and other languages
Compare before choosing
The official homepage claims its pricing is 11x cheaper than Eleven Labs
The official comparison center offers guides comparing it against leading TTS APIs such as Amazon Polly and ElevenLabs, allowing you to compare audio samples, features, and plans
Features
Low-latency streaming output
Use the /stream endpoint to synthesize text of up to 1,000 characters, suitable for latency-sensitive scenarios such as chatbots.
Synchronous synthesis
Use the /speech endpoint to synthesize text of up to 3,000 characters, returning a publicly accessible URL for the audio and timestamps.
Batch asynchronous synthesis
Use the /synthesisTasks endpoint to submit tasks of up to 500,000 characters, and retrieve audio and timestamp URLs via the task ID.
Word-level timestamps
Connect to the streamWithTimestamps endpoint via websocket to stream audio and word-level timestamps simultaneously.
Multilingual voices
Offers male and female voices in multiple languages, including American English, Chinese, Spanish, French, Hindi, and Italian.
Pros
- Offers a free quota and multiple character-based billing tiers, with free-plan characters resetting monthly
- Covers a range of interfaces from 1,000-character streaming to 500,000-character batch synthesis
- Provides word-level timestamps, suitable for subtitles or lip-sync scenarios
Pricing Plans
Free
- 250K characters
- 6 hours of audio
Basic
- 3M characters
- 67 hours of audio
- Discount for the first 6 months
Plus
- 42M characters
- 933 hours of audio
Pro
- 150M characters
- 3K hours of audio
Enterprise
- 625M characters
- 14K hours of audio
* Pricing information is for reference only. Please refer to the latest announcement on Unreal Speech's official website.