Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.slng.ai/llms.txt

Use this file to discover all available pages before exploring further.

Every model on the SLNG platform works through the Unified API. Some run on SLNG infrastructure (SLNG-hosted), others are proxied to the provider’s own API.
SLNG-hosted models use the slng/ prefix in the model identifier. They are optimized for low latency in specific regions and billed through your SLNG account with no separate provider account needed.

Text-to-Speech

Provider: Rime | Hosting: SLNG | Languages: English, HindiNatural prosody with improved voice quality over v1. French and Spanish support planned.
IdentifierDescription
slng/rime/arcana:3-enEnglish
slng/rime/arcana:3-hiHindi
Best for: conversational agents, high-quality voice output.
Provider: Rime | Hosting: SLNG | Languages: Arabic, English, French, German, Spanish150+ voices across 5 languages. Uses unversioned identifiers (arcana: instead of arcana:3-).
IdentifierDescription
slng/rime/arcana:enEnglish
slng/rime/arcana:esSpanish
slng/rime/arcana:frFrench
slng/rime/arcana:deGerman
slng/rime/arcana:arArabic
Best for: broad language coverage, high-volume production.
Provider: Deepgram | Hosting: SLNG + Deepgram | Languages: English, SpanishAvailable both SLNG-hosted (low latency) and proxied to Deepgram (more languages). The model field is required to pick a voice; there is no default.
IdentifierDescription
slng/deepgram/aura:2-enSLNG-hosted (English)
slng/deepgram/aura:2-esSLNG-hosted (Spanish)
deepgram/aura:2Proxied to Deepgram (7 languages)
Best for: low-latency voice apps, multilingual IVR.
Provider: Cartesia | Hosting: Cartesia | Languages: 40+Streaming TTS with controls for speed, volume, and emotion.
IdentifierDescription
cartesia/sonic:3Proxied to Cartesia
Best for: real-time voice agents, expressive multilingual output.
Provider: Murf | Hosting: Murf | Languages: 16 localesReal-time multilingual TTS over WebSocket with multiple encodings and sample rates.
IdentifierDescription
murf/murftts:falconProxied to Murf
Best for: low-cost voice agents, IVR, assistants.
Provider: KugelAudio | Hosting: KugelAudio | Languages: 2687 voices with expressiveness control across 26 languages. Three variants trade off latency and quality.
IdentifierDescription
kugelaudio/kugel:1-turboLowest latency
kugelaudio/kugel:1Higher quality, higher latency
kugelaudio/kugel:2Latest version, EU region
Best for: multilingual voice agents, content creation.
Provider: Soniox | Hosting: Soniox | Languages: 50+Streaming and one-shot synthesis across 50+ languages.
IdentifierDescription
soniox/tts-rt:v1Proxied to Soniox
Best for: multilingual streaming, long-form synthesis.
Provider: Sarvam AI | Hosting: Sarvam AI | Languages: 11 Indian languagesPurpose-built for Indian languages with 30+ speaker voices.
IdentifierDescription
sarvam/bulbul:v3Proxied to Sarvam AI
Best for: India-focused voice apps, regional language support.

Speech-to-Text

Provider: Deepgram | Hosting: SLNG + Deepgram | Languages: English, Spanish, HindiAvailable both SLNG-hosted and proxied.
IdentifierDescription
slng/deepgram/nova:3-enSLNG-hosted (English)
slng/deepgram/nova:3-esSLNG-hosted (Spanish)
slng/deepgram/nova:3-hiSLNG-hosted (Hindi)
slng/deepgram/nova:3-multiSLNG-hosted (auto-detect)
deepgram/nova:3Proxied to Deepgram (17 languages)
deepgram/nova:3-medicalProxied to Deepgram (medical English)
Best for: low-latency streaming, voice agents.
Provider: Deepgram | Hosting: Deepgram | Languages: 36Proven model with VAD, diarization, and smart formatting.
IdentifierDescription
deepgram/nova:2Proxied to Deepgram
Best for: production transcription, multi-speaker audio.
Provider: Soniox | Hosting: Soniox | Languages: 60+Real-time STT with speaker diarization, automatic language identification, and configurable endpoint detection. Returns native Soniox token frames.
IdentifierDescription
soniox/speech-ai:rt-v4Proxied to Soniox
Best for: multilingual streaming, multi-speaker conversations.
Provider: Reson8 | Hosting: Reson8 | Languages: 9 EuropeanReal-time STT with word-level timestamps, confidence scores, and partial results. Dutch, English, French, German, Italian, Polish, Portuguese, Spanish, Swedish.
IdentifierDescription
reson8/reson8stt:v1Proxied to Reson8
Best for: European-language voice agents, low-latency transcription.
Provider: Sarvam AI | Hosting: Sarvam AI | Languages: 24Speech recognition for 24 Indian and European languages with code-mixing support. HTTP-only; not available over WebSocket.
IdentifierDescription
sarvam/saaras:v3Proxied to Sarvam AI
Best for: India-focused batch transcription, code-switched audio.

Choosing a model

Use caseRecommended STTRecommended TTS
Voice agents (low latency)slng/deepgram/nova:3-enslng/rime/arcana:3-en
Multilingual transcriptiondeepgram/nova:3cartesia/sonic:3
Indian languagesslng/deepgram/nova:3-hisarvam/bulbul:v3
Batch transcriptiondeepgram/nova:3
Medical transcriptiondeepgram/nova:3-medical
For voices, regions, and detailed model specs, see the Model Catalog.