dTelecom APIs

Stream speech or download a complete WAV.

Kokoro and Supertonic speech in 32 languages, with complete-document SSML and explicitly enabled watermarking.

One speech rate

$0.006 per 1,000 spoken Unicode code points for streaming and WAV. Output is mono 48 kHz PCM16. Streaming supports interruption; WAV jobs provide a private downloadable file.

Voices and speed

Kokoro offers 54 voices in eight languages, including Mandarin. Supertonic offers F1–F5/M1–M5 across 31 languages, including English and Russian. Together they cover 32 languages.

Explicit controls

Choose voice, language and speed. Speed defaults to 1.0; Kokoro plain text supports 0.5–2.0, Supertonic 0.75–1.5. M5 at 1.1 is an example preference, not a service-wide default.

Complete-document SSML

Negotiate dtelecom-ssml-v1 and send a complete <speak> document. Use timed pauses, pronunciation aliases/dictionaries, rate/volume, and language or voice changes within the same engine. Plain text remains the default. The subset excludes incremental XML, pitch, marks, external audio and cross-engine switching.

A document renders at most 60 seconds. Explicit breaks are at most 5 seconds each and 30 seconds combined; effective speed is 0.75–1.5. Billing uses expanded spoken text, excluding tags and breaks. See the complete SSML contract.

Optional AudioSeal watermark

Audio is unmarked by default. Explicitly request watermarking and check server availability and activation. AudioSeal B runs at 48 kHz after resampling and gain, before PCM16; the TTS pipeline applies no later cleanup. It adds buffering and processing, without a price multiplier. Detection is not a personal signature or cryptographic proof; short or transformed clips may not detect.

Text/SSML to WAV

Submit up to 10,000 source and expanded spoken characters in a 64 KiB JSON request. Poll, then download the private completed WAV, retained for 24 hours. The outer output bound is 30 minutes; the separate 60-second SSML document limit still applies. See WAV jobs.