Gladia
Multilingual speech-to-text and audio intelligence API for real-time and async transcription
About Gladia
Gladia is a Paris-based speech-to-text API built on its Solaria models, covering 100+ languages with automatic language detection and mid-sentence code-switching. It offers real-time WebSocket streaming with partial transcripts in under 103ms and async batch transcription with speaker diarization, plus audio intelligence add-ons such as translation, NER, summarization, PII redaction, and audio-to-LLM. Pricing is pay-as-you-go per audio hour from a prepaid credit wallet, with EU data residency and GDPR, HIPAA, SOC 2 Type II, and ISO 27001 compliance.
Pros & Cons
Pros
- Fast real-time streaming with sub-103ms partial transcripts and sub-300ms end-to-end latency, suited to voice agents
- Native code-switching across 100+ languages, including many languages other STT APIs don't cover
- Low per-hour pricing that drops to $0.20/hr async on committed Growth plans
- Every feature and language included on pay-as-you-go, with €50 of free credits that don't expire
- EU-headquartered with EU data residency and GDPR, HIPAA, SOC 2 Type II, and ISO 27001 compliance
Cons
- Real-time speaker diarization isn't supported; you need multi-channel audio or an async re-run for speaker labels
- Audio-to-LLM and PII redaction work only on pre-recorded audio
- Solaria-3, the most accurate model on messy audio, is async only
- Prepaid credit wallet means topping up or configuring auto top-up before usage
- Some G2 reviewers report costs adding up at high audio volumes before volume discounts apply
Key Features
Solaria speech models
Solaria-1 handles both real-time and async with broad language coverage; Solaria-3 is an async-only model tuned for messy audio such as call-center and sales recordings.
Real-time streaming STT
WebSocket streaming with partial transcripts in under 103ms, sub-300ms end-to-end latency, and adjustable endpointing to trade speed for accuracy.
Async batch transcription
REST upload of pre-recorded audio and video with speaker diarization, automatic language detection, and word-level timestamps.
100+ languages with code-switching
Detects and transcribes mid-sentence language switches across every supported language without resetting a language parameter, including 42 languages many other STT APIs lack.
Speaker diarization
Identifies and labels speakers in pre-recorded audio; for live sessions, multi-channel streaming separates participants over one WebSocket.
Word-level timestamps
Every utterance returns word-level timing on all plans, in both real-time and async modes.
Custom vocabulary
Keyterm prompting improves recognition of product names, acronyms, and domain-specific terms.
Audio intelligence add-ons
Translation into 100+ languages, named entity recognition, sentiment analysis, summarization, and chapterization on top of transcripts.
Audio-to-LLM and PII redaction
Runs prompts over transcripts with access to 400+ LLMs for summaries and action items, and masks sensitive data; both are async only.
Compliance and data controls
EU data residency, GDPR, HIPAA, SOC 2 Type II, and ISO 27001, with model-training opt-out on paid plans and zero data retention on Enterprise.
Pricing
- Async transcription at $0.61/hr
- Real-time transcription at $0.75/hr
- €50 one-time free credits (roughly 80 async or 60 real-time hours)
- 30 real-time and 25 async concurrent requests
- Language detection and code-switching, diarization, 100+ languages
- Word-level timestamps
- Help center and Discord support
- Everything in Starter
- Async as low as $0.20/hr, real-time as low as $0.25/hr with upfront commitment
- Flexible concurrent requests and custom volume discounts
- 99.9% uptime SLA and priority processing queue
- Automatic model training opt-out
- Everything in Growth
- Annual plan with custom models, fine-tuning, and debundled pricing
- Unlimited concurrent requests and dedicated infrastructure
- Zero data retention and default training opt-out
- Dedicated Slack, account manager, and custom hosting
Best For
Voice agents
Stream caller audio over WebSocket and get partial transcripts fast enough for an agent to respond without dead air.
Meeting assistants and note-takers
Transcribe recorded meetings with diarization, then generate summaries and action items with audio-to-LLM.
Contact centers and CCaaS
Transcribe multilingual call recordings with Solaria-3, redact PII, and extract entities and sentiment for QA.
Live captioning and media
Generate low-latency captions for live events and broadcasts, and translate transcripts into 100+ languages.
Multilingual products
Handle users who switch languages mid-sentence without configuring a language per call.
Similar Tools
Synthflow
No-code AI voice agents for automated phone calls
Vida
AI Agent OS that calls, texts, emails, and chats at enterprise scale
Smartly.AI
Generative AI Powered Chatbots for Customer Service
Groq
Ultra-fast AI inference powered by custom LPU silicon
Agora
Real-time voice, video, and messaging APIs for developers
Cohere
Your next breakthrough, powered by AI