German Voice Cloning TTS Model using F5-TTS Architecture

Model Details

• Developed by: Johanna Reiml and team at KI-Servicezentrum, Hasso-Plattner-Institut (HPI)
• Base Model: SWivid/F5-TTS
• Paper: F5-TTS: A Fairytaler that Fakes Fluent and Faithful Speech with Flow Matching

Key Features & Capabilities

• Generates natural-sounding German speech from text
• Clones voices using minimal reference audio (few seconds)
• Suitable for audiobooks, voice assistants, and accessibility applications

Technical Specifications

• Datasets: Common Voice (Mozilla) and Emilia_DE
• Process: Fine-tuned checkpoints of base F5-TTS model
• Trained on Hardware: 8x NVIDIA H100

Contact

• AI Service Center: kisz@hpi.de
• Johanna Reiml: johanna@reiml.dev
• Enes Suermeli: muhammed.suermeli@student.hpi.uni-potsdam.de
• Kajo Kratzenstein: kajo.kratzenstein@student.hpi.de
• Carlos Menke: carlos.menke@rwth-aachen.de

Acknowledgements

Provenance and EU AI Act Art. 53 note

• Upstream model: SWivid/F5-TTS — published by SWivid.
• Upstream licence: cc-by-nc-4.0. This repository redistributes under the same terms; it grants no rights the upstream licence does not.
• What was done here: format conversion and/or quantisation only (MLX). No training, no fine-tuning, no merging, no distillation, no change to architecture, vocabulary or capability. Only the numeric representation of the upstream weights differs.
• Training data: documented — where it is documented at all — by the upstream provider; see the upstream model card. No training data was used, added or selected by this repository.
• Provider status: under Regulation (EU) 2024/1689 the upstream authors remain the provider of this model. Converting the serialisation format does not make this repository the provider of a new general-purpose AI model, and no such claim is made. Questions about training content, copyright policy or model capability belong upstream.