Kugel 3.5

Kugel 3.5 is a text-to-speech model that accurately handles short replies, IPA pronunciation, timed pauses, and spelled-out codes.

Kugel 3.5 is a text-to-speech model designed to get the fine details of phone conversations right. It improves on Kugel 3 by correctly generating short replies, IPA pronunciations, timed pauses, and character-by-character spelled-out codes. Users can provide IPA transcriptions in a pronunciation dictionary, set precise pause lengths using break tags, and spell out order numbers or licence plates. It maintains the same voices, languages, and API as Kugel 3. The model costs €0.0525 per minute of audio excluding VAT, billed per second of generated audio. It is available via the API by setting the model to kugel-3.5 or in the dashboard model list.

Kugel 3.5 at a glance

Best for: Software engineer
Runs on: API
Pricing: Kugel 3.5 €0.0525/min; Kugel 3 €0.07/min; kugel-2-turbo €0.035/min; enterprise plans priced separately
Released: October 2026
Category: AI audio tool
Website: kugelaudio.com

Key features

Generates clean short German replies
Uses IPA transcriptions from pronunciation dictionaries
Executes timed pauses via break tags
Reads codes character by character via spell tags
Maintains voices and languages from Kugel 3
Billed per second of generated audio

Use it for

Reading out order numbers and licence plates
Generating natural short conversational replies
Pronouncing brand names using IPA transcriptions
Inserting precise pauses into synthesized speech

Alternatives to Kugel 3.5

Kugel Ear — It tells your LLM which words it is unsure of so your voice bot asks once instead of guessing.
Speaches — It acts as an Ollama equivalent for running local TTS and STT models.
AuK — It handles zero-shot TTS, editing, and enhancement via natural language.
Vosk — It enables offline, low-latency speech recognition on small devices.
Kokoro-FastAPI — Generate hours of high quality speech with an OpenAI-compatible endpoint.
AssemblyAI — Industry-leading speech models deliver high accuracy and low latency for voice applications.
Piper — Runs locally and offers fast neural text-to-speech synthesis.
audio.cpp — Runs modern audio models locally with high-performance, native C++ execution.
kokoro-onnx — Runs local TTS inference with efficient ONNX runtime performance.
FluidAudio — Runs local audio AI models on Apple Neural Engine.
PaddleSpeech — It offers state-of-the-art models and streaming systems for speech tasks.
Moonshine — Runs entirely in your browser with no audio leaving your device.

All Kugel 3.5 alternatives