KugelAudio GmbH
KugelAudio builds and hosts production-ready voice AI in Europe, including its own text-to-speech models, real-time voice agents and related speech APIs. It focuses on natural multilingual speech, accurate handling of numbers and names, EU data residency/GDPR compliance, and on-premise deployment for sensitive workloads.
At a Glance
- European enterprises
- Regulated industries and organizations with sensitive data
- Healthcare
- Finance and insurance
- +3 more
AI Tools by KugelAudio GmbH
(1)KugelAudio
European Voice AI API for Developers
Discussions
No discussions yet
Be the first to start a discussion about KugelAudio GmbH
Latest News
Fluid 1 expanded to eight languages and a US endpoint was documented alongside the EU endpoint.
Kugel Fluid 1 launched publicly as a speech-to-speech voice-agent model over one WebSocket, compatible with the official OpenAI Python SDK.
KugelAudio added speech-to-text to its LiveKit and Pipecat agent integrations.
Kugel 3.5 became the default TTS model; speaker consent became required for voice cloning.
Products & Services
Default production text-to-speech model, optimized for short replies, pauses, spelled-out codes, numbers, emails and phone numbers; supports REST/WebSocket streaming, voice cloning and word-level timestamps.
Text-to-speech model available through the API, with multilingual speech, streaming and voice features.
Lower-cost, higher-volume text-to-speech model; it omits IPA notation and is less precise on numbers and codes.
Speech-to-text model that transcribes callers and flags words it is unsure about; supports nine European languages.
Market Position
KugelAudio positions itself as a German/EU-built alternative to US-centered voice APIs such as ElevenLabs, Cartesia, Deepgram, Murf AI and cloud-platform TTS from OpenAI, Azure, Google and Amazon. Its differentiators are EU Sovereign Cloud included on every plan, German/EU company and hosting, on-premise Kubernetes deployment, per-minute billing, multilingual European pronunciation, and an ElevenLabs-compatible endpoint; it competes on low latency and handling of numbers, codes and edge-case pronunciation.
Leadership
Founders
Kajo Kratzenstein
Co-founder and CTO. Researched text-to-speech for two years at the Hasso-Plattner-Institut and built the model that became Kugel. Previously built a production aneurysm-detection model at Floy GmbH and was CTO at Sagemode, where he worked on an in-game AI coach; he holds computer-science degrees from the Hasso-Plattner-Institut.
Viktor Presber
Co-founder responsible for go-to-market and product. Previously co-founded full-house.io and took it to $300K ARR in six months with enterprise clients including Siemens Energy, Hitachi Rail and Barilla; previously co-founded Sagemode and ran a film-production company.
Executive Team
Kajo Kratzenstein
Co-founder and Geschäftsführer (managing director); CTO
TTS researcher at the Hasso-Plattner-Institut; previously CTO at Sagemode and builder of a production aneurysm-detection model at Floy GmbH.
Viktor Presber
Co-founder and Geschäftsführer (managing director); go-to-market and product
Previously co-founded full-house.io and Sagemode and ran a film-production company; built enterprise-sales experience with Siemens Energy, Hitachi Rail and Barilla through full-house.io.
Founding Story
The founders say enterprise voice buyers were frustrated by providers mispronouncing product names and by legal teams refusing to sign data-processing agreements. Kajo's Hasso-Plattner-Institut TTS research supplied the technical foundation, while Viktor brought enterprise go-to-market experience; together they built a European voice model that could run in customers' own infrastructure so sensitive data would not go to a third party.
Business Model
Revenue Model
Usage-based API and hosted voice-agent billing, priced primarily per audio minute and billed by the second; enterprise contracts add on-premise hosting, SLAs, custom DPAs and committed-usage rates.
Pricing Tiers
Default model; includes EU hosting, voice cloning, REST/WebSocket streaming, timestamps, normalizer, dictionaries, speed control and break tokens.
Lower-cost model for high volumes; no IPA notation and less precision on numbers/codes.
Text-to-speech model with IPA and precision features.
Billed by input audio duration.
Speech enhancement, billed by input audio duration.
Conversation time, billed by the second; back-office model costs are separate.
On-premise hosting, SLAs, custom DPA, dedicated account manager, 24/7 support, higher concurrency and committed-usage rates.
Target Markets
- European enterprises
- Regulated industries and organizations with sensitive data
- Healthcare
- Finance and insurance
- Public sector
- Customer-support and contact-center operators
- Customer-service and telephone voice agents
- Real-time voice applications and interactive applications
- Healthcare, finance/insurance and public-sector voice systems handling sensitive data
- Content creation and custom voice production
- Telephony, browser and app voice interfaces
- On-premise voice AI for organizations requiring data to remain in Europe or inside their own infrastructure