Instructions to use gradium/phonon-40m-de with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use gradium/phonon-40m-de with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-to-speech", model="gradium/phonon-40m-de")# pip install -U transformers accelerate # Load model directly from transformers import AutoProcessor, AutoModelForTextToWaveform processor = AutoProcessor.from_pretrained("gradium/phonon-40m-de") model = AutoModelForTextToWaveform.from_pretrained("gradium/phonon-40m-de", device_map="auto") - Notebooks
- Google Colab
- Kaggle
Would you consider Apache 2.0 for phonon-40m-de, or a fair-use license for public-sector use?
Hi Gradium team,
Congratulations on the Phonon release β the on-device footprint and streaming quality are impressive, and we'd love to build on it.
Who we are: We're the MUCGPT team at it@M (Munich's municipal IT services provider). MUCGPT is the City of Munich's free, open-source AI assistant for public sector employees.
What we'd like to do: Integrate Phonon 40m (German) as an optional, fully on-device text-to-speech feature so that AI answers in MUCGPT can be read aloud β improving accessibility, especially for visually impaired users, without any cloud round-trip.
Why German, why Apache: Our whole stack is MIT-licensed, and our speech-input pipeline already uses local, permissively licensed models. To keep MUCGPT fully open-source and redistributable by other municipalities, we can only bundle Apache/MIT-licensed components β so phonon-31m-en works license-wise, but German is the language our users actually need, and 31m doesn't cover it.
Our questions:
- Is the CC BY-NC licensing of the phonon-40m models driven by training-data constraints, or is it a deliberate commercial-tier split?
- Would you consider granting a fair-use exception for public-sector use β for example, a license variant that allows municipal administrations and other public institutions to use and redistribute phonon-40m-de with open-source projects like MUCGPT?
An important detail in our case: as a public institution we are not legally non-profit β we're a commercial entity in the legal sense, which means CC BY-NC 4.0 excludes us entirely, despite our use being strictly in the public interest, free of charge for employees, and with no monetization. That's the gap we're stuck in: technically ideal, publicly beneficial, but license-blocked. Happy to be a showcase use case for you.
Hi @freinold
Thanks for reaching out, we're certainly open to discussing it alternative licensing options. Would you mind reaching out by email to contact@gradium.ai ? If you can just copy the content of your initial post, we will be happy to follow up on this.
Best,