AI audio infrastructure for voice products that transcribes and enriches conversations through a single API
Foundation model for mobile, wearables, robots, smart home, automotive and microcontrollers
Stage
Not given
Startup
Founded
Not given
Not given
Based in
Not given
Not given
Pricing model
Freemium
Open source
Pricing
Self-serve async: starts at 0.16€/hour (Solaria-3) and 0.27€/hour (Solaria-1) with automatic volume discounts down to 0.12€/hour at 10,000+ monthly hours. Real-time: 0.68€/hour. Add-ons: Audio summarization 0.03€/hour, translation 0.05€/hour, named entity recognition 0.07€/hour, PII redaction 0.07€/
Not given
Key features
Real-time and async speech-to-text transcription in 100+ languages
Speaker diarization with 3x lower error rate than competitors
Native code-switching for mid-sentence language detection
Named entity recognition for extracting names, dates, and organizations
PII redaction to mask sensitive data automatically
Sentiment analysis tied to speaker and timestamp
Audio-to-LLM pipeline for generating summaries and action items
Custom vocabulary and keyterm prompting for domain-specific accuracy
Speech-to-text translation in 100+ languages
Enterprise-grade infrastructure with 99.9% uptime SLA and EU data residency
Tool calling with function selection and argument filling
Structured extraction with guaranteed JSON parsing
Text embeddings for local semantic search and matching
Confidence scoring for routing and gating calls
Laddered subnetwork architecture from 2 to 20 layers
LoRA fine-tuning support for customization
400-4k tokens per second inference speed on Raspberry Pi 5
Python package with pip install and Hugging Face distribution
What makes it different
29% higher accuracy than alternatives on conversational speech
100+ languages with native code-switching from day one
Built-in audio intelligence features included in core API, not billed separately
100% EU data residency with GDPR, HIPAA, SOC 2 Type II, and ISO 27001 compliance
Sub-300ms latency for natural real-time voice experiences
Beats models 10x its size on mobile tool calling benchmarks
Matches models 2-3x larger on structured extraction tasks
Entire model in a single 8-29 MB binary for tiny devices
Simple Attention Network architecture uses 2x fewer MFLOPs than transformers
Fine-tuned subnetworks outperform larger models like DeepSeek V4 Flash
Gladia