01
Text-to-speech
Turn text into natural speech.
Generate WAV, PCM, MP3, or Opus audio with speed control, native voices, and familiar voice aliases.
POST /v1/audio/speech
Voice AI + live AI avatars
Speech-to-text. Text-to-speech. Live AI avatars.
Transcribe any conversation, generate natural speech, build realtime voice agents, and put a consented digital presenter on screen in one platform.
Base URL
https://api.raspy.app/v1
16 voicesNatural TTS
Word timestampsFast STT
WebSocketsRealtime speech
Live + renderedAI avatars
Voice AI + live avatars
Give software a voice, an ear, and a face.
Use the API surface your applications already understand. Raspy keeps speech, realtime agent, and avatar workloads under one authenticated, metered contract.
02
Speech-to-text
Turn every recording into useful text.
Transcribe audio with language detection, word timestamps, JSON, plain text, SRT, and VTT output.
POST /v1/audio/transcriptions
03
Live AI avatars
Put a digital presenter on screen.
Give a consented digital identity voice, captions, and a visual presence for live sessions or rendered presenter video.
POST /v1/avatar/renders
Realtime
Listen, speak, interrupt, continue.
Bounded WebSocket sessions stream PCM into partial and final transcripts, or stream text into cancellable audio with backpressure.
WS /v1/realtime/transcriptions
WS /v1/realtime/speech
Made for real work
From a single utterance to a complete show.
Use one capability or chain them together. The same contract supports agents, media pipelines, accessibility, and live production.
Conversations that hear and answer.
Stream caller audio to text, generate the response, then speak it back without changing SDKs.
Make every recording discoverable.
Create transcripts, word-level timing, subtitles, summaries, and searchable media archives.
Show up live or produce on demand.
Use a consented digital presenter for realtime sessions, narrated updates, training, and social video.
Power hosts, shows, and co-presenters.
Connect realtime speech and avatar outputs to streaming, recording, clipping, and distribution workflows.
OpenAI-compatible
Change the base URL. Keep your client.
Point an OpenAI audio client at Raspy, use your Raspy key, and keep familiar request shapes for speech and transcription.
- 1Get access
Use a metered key issued through Tokaroo.
- 2Set the base URL
Use
https://api.raspy.app/v1. - 3Call the capability
Generate speech, transcribe audio, or render an avatar.
raspy_quickstart.py
from openai import OpenAI
client = OpenAI(
base_url="https://api.raspy.app/v1",
api_key=RASPY_KEY,
)
speech = client.audio.speech.create(
model="raspy-tts-1",
voice="nova",
input="Your product can speak.",
)
speech.stream_to_file("voice.mp3")/v1/audio/speechTTS/v1/audio/transcriptionsSTT/v1/avatar/rendersVideo/v1/voicesVoicesHonest status
Use what is live. See what comes next.
Production claims match the capabilities reported by the running API.
Text-to-speechLive
Natural voices, four audio formats, voice aliases, and speed control.
Speech-to-textLive
Transcription, word timestamps, JSON, text, SRT, and VTT.
Realtime speechLive
Streaming TTS and STT with finalization, cancellation, and backpressure.
Avatar videoLive
Consented identity registry plus durable MP4, narration, captions, and thumbnail output.
GPU lip syncIn evaluation
A reviewed worker adapter exists; production activation still requires isolated GPU proof and licensing checks.
Voice enrollmentPlanned
Consent, revocation, provenance, and permission controls will arrive before cloning is exposed.
Questions
Built to fit the stack you already have.
Is Raspy only for Gwen and Water?
No. Raspy powers internal products and is also exposed as an authenticated API for other applications.
Can I use an existing OpenAI client?
Yes. Speech and transcription follow the familiar OpenAI audio surface. Change the base URL and key.
Are video avatars lip synced today?
The production baseline creates narrated, captioned MP4 presenter videos without lip sync. GPU lip sync remains clearly labeled until the production worker passes its gates.
How are identity and consent handled?
Avatar identities require source type, owner reference, consent attestation, and consent scope. They can be revoked or deleted through the API.
Voice is an interface
Make your product speak, listen, and show up.
Bring speech-to-text, text-to-speech, realtime agents, and live AI avatars into one experience.