REVERIE · TTS

Text-to-speech

Base URL https://api.lucidweights.ai/v1/tts/. Convert text to speech with registered speakers. REST for batch synthesis, WebSocket for streaming.

NOTE

Send your API key in the x-api-key header for REST requests. WebSocket streaming also accepts an api_key query parameter. Speaker registry and external speaker management require a Dev API key.

Endpoints

Method	Path	Purpose
POST	/v1/tts/inference	Synthesize speech, base64 audio in JSON
POST	/v1/tts/file-inference	Synthesize speech, audio file response
WS	/v1/tts/stream	Real-time streaming synthesis
POST	/v1/tts/pronunciation/{speaker}/words	Add word pronunciation override
POST	/v1/tts/pronunciation/{speaker}/phonemes	Add phoneme override
DELETE	/v1/tts/pronunciation/{speaker}/words/{word}	Remove word override
DELETE	/v1/tts/pronunciation/{speaker}/phonemes/{phoneme}	Remove phoneme override
GET	/v1/tts/speakers	List registered speakers · Dev key
GET	/v1/tts/speakers/{name}	Get a speaker · Dev key
POST	/v1/tts/speakers	Register a speaker from reference audio · Dev key
PUT	/v1/tts/speakers/{name}	Update speaker audio or transcript · Dev key
DELETE	/v1/tts/speakers/{name}	Soft-delete a speaker · Dev key
POST	/v1/tts/speakers/{name}/restore	Restore a speaker · Dev key
GET	/v1/tts/speakers/{name}/preview	Generate a speaker preview · Dev key
GET	/v1/tts/external-speakers	List external speakers · Dev key
GET	/v1/tts/external-speakers/{name}	Get an external speaker · Dev key
POST	/v1/tts/external-speakers	Register an external speaker · Dev key
PATCH	/v1/tts/external-speakers/{name}	Update an external speaker · Dev key
DELETE	/v1/tts/external-speakers/{name}	Soft-delete an external speaker · Dev key
POST	/v1/tts/external-speakers/{name}/restore	Restore an external speaker · Dev key
POST	/v1/tts/prerecordings/refresh	Incrementally refresh cached audio
POST	/v1/tts/prerecordings/refresh/manual	Regenerate selected cached audio
POST	/v1/tts/prerecordings/sync	Sync cached audio across replicas

See Speaker registry, External speakers, and Prerecordings for request fields, responses, and access requirements.

Common request fields

Field	Default	Description
text	required	Text to convert to speech
speaker	required	Registered speaker name
language	default	ISO 639-1 language code
output_format	pcm_24000	Audio format (see below)
speed	1.0	Range 0.1–2.0
expressive_level	30	Range 0–100; optimal 25–65
custom_pronunciations	-	Per-request word/phoneme overrides
translate_text	false	Translate input before synthesis

Output formats

pcm_16000pcm_22050pcm_24000pcm_44100ulaw_8000mp3_22050_32mp3_44100_32mp3_44100_64mp3_44100_96mp3_44100_128mp3_44100_192

Voices and languages

Reverie synthesizes in 30 languages. The playground library has more than 250 voices, including featured English and Hindi voices. Search and filter there by name, language, gender, accent, or style. The selected speaker can still synthesize in any language it supports.

Featured voices

Start with these in the Playground, then browse the full library:

Kabir

kabir

Measured, narrative weight

Emily

emily

Warm, front-desk clarity

Riya

riya

Bright, service-ready

Serena

serena

Calm, support-desk poise

Marcus

marcus

Friendly, walkthrough tone

Supported languages

Pass the ISO 639-1 code as language. Each language has conversational speakers in the playground catalog.

ArabicarDanishdaGermandeGreekelEnglishenSpanishesFinnishfiFrenchfrHebrewheHindihiIndonesianidItalianitJapanesejaKhmerkmKoreankoLaoloMalaymsBurmesemyDutchnlNorwegiannoPolishplPortugueseptRussianruSwedishsvSwahiliswThaithTagalogtlTurkishtrVietnameseviChinesezh