Show HN: TONE3000 – Community for Neural Amp Modeling with 90k+ models trained
tone3000.com
Since launching in March, we've hit 90K+ models trained by our community that have been downloaded 1.3 million times. NAM is an open-source project using CNNs to model the sound of any guitar amp, pedal, outboard gear and more. These tone models can be played real-time with low latency. TONE3000 started as a fork that enabled training with any dry/wet audio pair via our datacenter GPUs (https://www.tone3000.com/capture?type=dry-wet) and has since grown into a full featured platform with an incredible community of artists training, sharing and discovering tone models.
What I'm releasing today:
-- v1 API (https://www.tone3000.com/api): Lets plugin developers and hardware manufacturers integrate NAM model loading directly into their products. Training endpoints coming soon!
-- Open-sourced our web player (https://github.com/tone-3000/neural-amp-modeler-wasm): Runs NAM models directly in browsers using WebAssembly.
-- Example web app (https://github.com/tone-3000/api): Shows how to integrate the API for TONE3000-powered applications.
The vision:
We hope to push NAM to become an open standard in audio modeling. Our extensive library of tones, API and open-source tools make it accessible to any developer or manufacturer. Ultimately we want to democratize high-quality modeling for all musicians.
Real-world adoption:
Hardware manufacturers are building NAM support into their products. Darkglass Electronics launched ANAGRAM (https://www.tone3000.com/blog/unveiling-anagram-darkglass-el...) - the first bass pedal with built-in NAM support, and Dimehead released a dedicated NAM player (https://www.dimehead.de/). Both load models from our platform. And I’m seeing more and more DAW plugin support (https://wavemind.net/software)
The platform will remain free while I continue to work on monetization tools for our awesome creators. Whether you're a bedroom producer or touring musician, you can capture your favorite gear and share it with the world.
Thoughts? Working on DSP, audio ML? Ideas for where this could go?
Check it out: https://www.tone3000.com