Lugemi documentation

Language intelligence for speaking agents

Lugemi is first-party voice, speech, and translation infrastructure — Echo neural speech, Baobab dialect-aware MT, Atlas reasoning, and a complete OpenAPI surface for Africa-first and global corridors.

Explore full API reference →OpenAPI interactive docsDownload openapi.json

How Lugemi works

Lugemi provides AI language infrastructure: text-to-speech, speech-to-text, translation, voice cloning, conversational agents, and governance. Use it in four ways, suited to different audiences.

Voices

Speech personas as own:* IDs (for example own:en-gh-female) or clone:{id}. Accent Identity packs bind country, culture, and lifestyle to Echo voices.

Models

Echo Voice / Echo Listen, Baobab Translate, Atlas Reason, and Vector embeddings control quality, latency, and language coverage. See the live matrix at /models.

Usage

Metered characters (TTS/MT), tokens (chat), and STT minutes against workspace plans. Create lg_live_ or lg_test_ keys in /keys.

Choose your path

Pick the surface that matches how you ship.

Lugemi Creative

No-code studio for speech, dubbing, voice tools, and localization workflows in the browser.

Open Creative →

Lugemi Agents

Build text chatbots and spoken voice agents via API — copy-paste examples for builders.

Builders guide →

Lugemi API

REST + OpenAPI for every capability, with @lugemi/sdk (TypeScript) and MCP for agent IDEs.

API reference →

Builders

Ship native accent catalogs, connectors, and production own:* voices under your product.

Open Builders →

Meet the models

First-party Lugemi models — not reseller wrappers. Availability follows GET /v1/models/live.

Text-to-speech

Lugemi Echo Voice

Neural own:* speech for African and global corridors — batch + stream.

  • own:* catalog + clone:{id} with consent
  • Cultural English and L1 accents
  • POST /v1/tts/synthesize · /v1/audio/speech
  • Streaming via /v1/tts/stream
Explore →

Speech-to-text

Lugemi Echo Listen

ASR for accents, dialects, and code-switching with word-level timestamps.

  • Multipart transcription
  • Dialect-aware recognition
  • POST /v1/audio/transcriptions
  • Live sessions via /v1/live/*
Explore →

Translation

Lugemi Baobab

Africa-first dialect-aware MT with glossary, TM, formats, and SSE stream.

  • 213-language registry
  • Default en→Twi (ak) pairs
  • POST /v1/translate · /stream · /formats
  • Document jobs /v1/documents/translate
Explore →

Reasoning chat

Lugemi Atlas

Multilingual reasoning for dialect, culture, and vertical speaking-agent tasks.

  • Chat completions with optional translateReplyTo
  • Serving residency (e.g. Accra · af-west-1)
  • POST /v1/chat/completions
  • MCP tool surface for IDEs
Explore →

Embeddings

Lugemi Vector

Cross-lingual embeddings for RAG, retrieval, and knowledge grounding.

  • POST /v1/embeddings
  • Knowledge query /v1/knowledge/query
  • Dataset + vector cloud surfaces
Explore →

View live model matrix →

Browse by capability

Everything we support end-to-end — speech, translate, agents, studio, connectors, and enterprise governance.

Text to speech

Convert text into lifelike Echo speech with own:* voices.

/v1/tts/synthesize

Speech to text

Transcribe spoken audio with timestamps and dialect cues.

/v1/audio/transcriptions

Translate

Dialect-aware Baobab MT across the language registry.

/v1/translate

Language detect

Identify language, script, and confidence before MT.

/v1/detect

Chat completions

Atlas reasoning with optional reply translation.

/v1/chat/completions

Voice chat agent

Spoken STT → agent → TTS turns with auto language detect.

/v1/voice/agent/turn

Voices & accents

Catalog, Accent Identity packs, and cultural speech varieties.

/v1/tts/voices

Multi-speaker dialogue

Turn-based scenes with distinct Echo voices per speaker — podcasts, ads, NPCs.

/v1/dialogue/synthesize

Forced alignment

Word timings with SRT/VTT for captions, dubbing QA, and LearnBack.

/v1/align

Voice Design

Prompt a creative brief; receive ranked live Echo voice matches.

/v1/voice-design

Speech Player embeds

Drop narrated players on articles, docs, and help centers.

/v1/speech-embed

Voice cloning

Consent-gated clones with watermarked synthesis.

/v1/voice-clones

Dubbing & interpret

STT → MT → TTS cascades for live and offline dubbing.

/v1/interpret

Documents

Async PDF/DOCX translation jobs with layout awareness.

/v1/documents/translate

Localize

i18n JSON/YAML trees with ICU and glossary fidelity.

/v1/localize

Streaming

SSE translate and TTS streams for live captions and agents.

/v1/translate/stream

Live sessions

Realtime audio sessions with events and repair loops.

/v1/live/sessions

Voice Studio

Workspace generations, assemblies, and review takes.

/v1/voice-studio

Audio intelligence

Enhancement, isolation, emotion, and speaker analytics.

/v1/audio-intelligence

Pronunciation

Pronunciation intelligence and scoring for learning agents.

/v1/pronunciation

Voice biometrics

Speaker verification and identity-aware voice checks.

/v1/voice-biometrics

Knowledge & RAG

Grounded queries with citations over workspace corpora.

/v1/knowledge/query

Embeddings

Cross-lingual vectors for retrieval and matching.

/v1/embeddings

Languages & coverage

Full registry, locales, African strategic set, coverage matrix.

/v1/languages

Glossary & TM

Terminology and translation memory applied on every turn.

/v1/glossary/terms

Language Passport

Private approved meaning memory synced to glossary, ASR, Accent Identity, and OWN TTS.

/v1/passport/me

MCP

Model Context Protocol bridge for Cursor and agent IDEs.

/v1/mcp

Connectors

Slack, office, storage, and platform connector catalog.

/v1/connectors/platform

MeetingBridge

Live meeting captions, sync, and bilingual facilitation.

DealBridge

Negotiation and trade sessions with dialect-aware turns.

AccessLine

Telephony dialer and inbound voice access lines.

Integrity & provenance

Watermarks, consent attestations, and audit verification.

/v1/language-integrity/verify

Usage & billing

Character, token, and STT minute metering against plans.

/v1/usage/summary

Marketplace

Listings, installs, and Connect onboarding for extensions.

/v1/marketplace/listings

Datasets & training

Dataset versions, fine-tunes, and training job launchers.

/v1/datasets

OCR

Document and image text extraction into translate pipelines.

Authentication

Bearer keys (Lugemi machine auth): Authorization: Bearer lg_live_… for production or lg_test_… for soft sandbox. These keys are first-party Lugemi auth — Playground, Studio, and API clients never depend on console session cookies. Console sign-in uses Lugemi account auth for humans; OpenAPI and Guides load via same-origin /api/openapi without a session.
Create API key
POST/v1/translate

Baobab dialect-aware MT. Use source: "auto" or explicit registry codes such as ak, sw, yo.

Translate request
{
  "text": "Welcome to Lugemi",
  "source": "en",
  "target": "ak"
}
POST/v1/tts/synthesize

Echo neural speech. Prefer authenticated Studio calls so Kokoro/Piper weights serve when Echo Voice is live.

Speech request
{
  "text": "Mema wo akye. Akwaaba ba Lugemi.",
  "voice": "own:en-gh-female",
  "language": "en-GH",
  "format": "mp3"
}

Voice & speech

Echo Voice TTS, Echo Listen ASR, accents, clones, and live audio.

POST/v1/tts/synthesize

Neural text-to-speech

Synthesize with own:* Echo voices or clone:{id}. Preferred Studio path when authenticated.

View in API reference →
GET/v1/tts/voices

Voice catalog

List Echo voices with locale, gender, verification, and synth engine metadata.

View in API reference →
POST/v1/audio/speech

Legacy speech synthesize

Compatible speech endpoint used by SDKs and audio console (same Echo gateway).

View in API reference →
POST/v1/audio/transcriptions

Speech-to-text

Multipart transcription with timestamps, dialect cues, and metered minutes.

View in API reference →
GET/v1/accents/identity

Accent Identity packs

Country-complete cultural identity packs with sample phrases and Echo voice links.

View in API reference →
GET / POST/v1/voice-clones

Voice clone management

Consent-gated clones with watermark headers on synthesis.

View in API reference →
POST/v1/interpret

Live interpreter cascade

Audio → STT → Baobab MT → Echo TTS for realtime interpretation.

View in API reference →
POST/v1/voice/simulate

Voice FAQ simulator

Multi-turn speaking-agent simulation with sentiment and confidence.

View in API reference →
POST/v1/demo/speech

Public Play speech

Rate-limited Play via Native Speech Engine; remaps formant-only packs to cultural English/French.

View in API reference →
GET/v1/speech/capabilities

Native Speech Engine registry

Task/variety manifests, verification status, and honesty rules (no nationality inference).

View in API reference →
POST/v1/speech/syntheses

Native Speech Engine synthesis

Verified TTS with diagnosis, pronunciation aliases, and provenance headers.

View in API reference →
POST/v1/speech/diagnose

Speech delivery diagnosis

Layered beep/echo/unintelligible checks before playback or retraining.

View in API reference →

Translation & localization

Baobab MT, documents, formats, glossary, and SSE streams.

POST/v1/translate

Dialect-aware translation

Africa-first pairs (default en→ak) with glossary/TM and honest local packs.

View in API reference →
POST/v1/translate/formats

Structured formats

Translate HTML, Markdown, SRT/VTT, and structured payloads.

View in API reference →
POST/v1/translate/chat

Chat-oriented translate

Turn-aware translation helpers for conversational agents.

View in API reference →
GET/v1/translate/engine

Translate engine catalog

Baobab engine capabilities, residency, and feature matrix.

View in API reference →
GET / POST/v1/glossary/terms

Terminology glossaries

Workspace brand lexicons applied deterministically in MT.

View in API reference →
POST/v1/pragmatics/translate

Pragmatic translation

Tone and speech-act aware localization for CX and sales.

View in API reference →

Intelligence & models

Atlas chat, RAG, embeddings, model matrix, and MCP.

POST/v1/chat/completions

Atlas chat completions

Regional reasoning with optional translateReplyTo for speaking agents.

View in API reference →
POST/v1/voice/agent/turn

Voice chat agent turn

Multipart audio → STT → chat → TTS with auto language detect.

View in API reference →
GET/v1/languages

Language registry

Complete language list with count/total/languageCount and African strategic set.

View in API reference →

Realtime & agents

Live sessions, streaming runtime, MeetingBridge, DealBridge, AccessLine.

POST/v1/mix/transcribe-translate

Transcribe + translate mix

Combined STT and Baobab MT for mixed media pipelines.

View in API reference →
GET/v1/connectors/platform

Platform connectors

Voice, office, storage, and chat connector catalog with demo hooks.

View in API reference →
POST/v1/grounded/interpret

Grounded interpret

Interpretation with knowledge grounding for enterprise agents.

View in API reference →

Studio & media intelligence

Voice Studio, audio intelligence, pronunciation, biometrics, OCR.

GET / POST/v1/voice-studio

Voice Studio workspace

Generations, assemblies, and review workflows for production voice.

View in API reference →
POST/v1/dialogue/synthesize

Multi-speaker dialogue

Concatenate per-turn Echo synthesis into a dialogue WAV mix.

View in API reference →
GET/v1/audio-intelligence

Audio intelligence

Enhancement, isolation, emotion, and speaker analytics APIs.

View in API reference →
GET/v1/pronunciation

Pronunciation intelligence

Scoring and feedback for learning and speaking agents.

View in API reference →

Integrity, usage & enterprise

Provenance, audit, billing, orgs, keys, and marketplace.

POST/v1/language-integrity/verify

Provenance verification

Cryptographic watermarks and consent attestations.

View in API reference →
GET/v1/language-integrity/protocol

Adoption protocol

Formal bilingual filing and synthetic audio compliance protocol.

View in API reference →

TypeScript SDK quickstart

Ship with @lugemi/sdk — translate, speech, and chat against the same first-party endpoints.

TypeScript example
import { Lugemi } from '@lugemi/sdk';

const client = new Lugemi({
  apiKey: process.env.LUGEMI_API_KEY!,
  baseUrl: 'https://api.lugemi.com',
});

const mt = await client.translate({
  text: 'Welcome to Lugemi Language Intelligence',
  source: 'en',
  target: 'ak',
});

const speech = await client.speech.synthesize({
  text: mt.text,
  voice: 'own:en-gh-female',
  language: 'ak',
});

Configuration

.env
# .env
LUGEMI_API_KEY=lg_live_...
LUGEMI_BASE_URL=https://api.lugemi.com

Verifying OpenAPI…

Every tagged operation also appears in the interactive API reference (searchable by path, method, and OpenAPI tag) and in Swagger at https://api.lugemi.com/docs.

Documentation · Lugemi