KugelAudio
EU-hosted voice AI platform offering text-to-speech, voice cloning, speech-to-speech for phone calls, and audio noise removal via API.
At a Glance
Engagement
Available On
Alternatives
Listed Nov 2026
About KugelAudio
KugelAudio is a Berlin-based voice AI platform that provides text-to-speech models, voice cloning, a speech-to-speech model for phone calls, and an audio cleanup model. It is developed and hosted in Europe, and the company states it is GDPR compliant. Developers access it through a REST and WebSocket streaming API with Python and TypeScript SDKs.
What It Is
KugelAudio offers voice models for live calls and voice agents. The text-to-speech line (kugel-3 and the lower-cost kugel-2-turbo) generates speech in 24 languages, with native pronunciation, accents and dialects. The site says the models are trained on edge cases such as IBANs, phone numbers, email addresses, postal codes and street names. Voices can be streamed from a catalogue or cloned from a recording. Clarity (clarity-1) removes background noise and other voices from calls and recordings.
Fluid 1 speech to speech
Fluid 1 is a speech-to-speech model for phone calls. According to KugelAudio, it listens while the caller talks, stops when interrupted, and lets the customer's chosen model handle the reasoning; it is described as compatible with OpenAI Live. KugelAudio reports that Fluid 1 solves 84% of customer requests in τ²-bench, noting this is its own measurement rather than a leaderboard submission.
Developer workflow
A client is created with an API key, the connection can be prewarmed, and audio is generated by specifying text, model, voice ID and language, then saved as a file. The site describes drop-in use with Pipecat and LiveKit pipelines. Features include word-level timestamps, an integrated normalizer, pronunciation dictionaries, speed control, break tokens and IPA notation on kugel-3. KugelAudio states a median first audio of 91 ms for Kugel 3 over WebSocket streaming, and presents this as a comparison against other vendors' models.
Data sovereignty and deployment
KugelAudio says it builds, trains and hosts in the EU and does not train on customer content. An EU Sovereign Cloud option for EU/EEA customers uses EU-company hosting, and an on-premise option runs the models in the customer's own data centre with the same SDK and API. Enterprise arrangements cover SLAs, a custom DPA and dedicated support.
Community Discussions
Be the first to start a conversation about KugelAudio
Share your experience with KugelAudio, ask questions, or help others learn from your insights.
Pricing
Text to Speech (kugel-3.5)
Default TTS model, billed per minute of generated audio, excl. VAT.
- €0.0525 per minute of generated audio (excl. VAT; €0.062475 incl. 19% VAT)
- IPA notations
- Precise on numbers, emails and phone numbers
- Guaranteed EU hosting
- EU Sovereign Cloud (EU/EEA)
- Voice cloning
- REST and WebSocket streaming
- Word-level timestamps
- Integrated normalizer
- Dictionaries
- Speed control
- Break tokens
Text to Speech Turbo (kugel-2-turbo)
Lower-cost TTS model for high volumes, billed per minute of generated audio, excl. VAT.
- €0.035 per minute of generated audio (excl. VAT; €0.04165 incl. 19% VAT)
- IPA notations: Not available
- Precise on numbers, emails and phone numbers: Not available
- Guaranteed EU hosting
- EU Sovereign Cloud (EU/EEA)
- Voice cloning
- REST and WebSocket streaming
- Word-level timestamps
Speech to Text (kugel-ear-1)
Speech to text billed per minute of input audio, excl. VAT.
- €0.008 per minute of input audio (excl. VAT; €0.00952 incl. 19% VAT)
- Flags the words it is unsure of
Clarity (clarity-1)
Background noise and voice removal billed per minute of input audio, excl. VAT.
- €0.0015 per minute of input audio (excl. VAT; €0.001785 incl. 19% VAT)
- Billed per second, rounded up, one second minimum per request
Enterprise
On-premise hosting, SLAs and committed rates agreed in an Enterprise contract.
- On-premise hosting
- Service level agreements (SLAs)
- Custom data processing agreement (DPA)
- Terms for professional secrecy holders (§ 203 StGB)
- Dedicated account manager
- 24/7 support
- Higher custom concurrency limits
- Lower rates with committed usage
Capabilities
Key Features
- Text-to-speech in 24 languages
- Voice cloning
- Real-time REST and WebSocket streaming
- Fluid 1 speech-to-speech for phone calls
- Clarity background noise and voice removal
- Word-level timestamps
- IPA notation support
- Integrated text normalizer
- Pronunciation dictionaries
- Speed control and break tokens
- EU Sovereign Cloud hosting
- On-premise deployment
- Python and TypeScript SDKs
