You find a model that turns a hummed melody into notes, separates a recording into stems or generates material for a new track. Someone else may be looking for exactly that task. A good directory suggestion helps them find the model and understand what they can do with it.
Start with the alphabetical model list to check whether it already has a page. Then sign in and suggest a model. You can recommend someone else’s work; give its creators credit and link their sources.
Give the next musician a clear starting point
Include the model’s name, an official source and a short explanation of what goes in and what comes out. “Turns a solo piano recording into MIDI notes” is easier to judge than “a breakthrough music model.” Name the release when several versions share a name.
The publisher’s model card, project page or repository makes a good starting source. Add a paper, code repository, demo or license page when you have one. Keep each link specific to the model or release you are recommending.
If the weights are on Hugging Face, link the model card and include its exact namespace/model identifier in your notes. Name the checkpoint or variant you mean; a demo Space or dataset has a different identity. Add a sourced release date when you can find one. A paper date or a repository upload date may describe something different.
Keep availability separate from the idea
A paper can describe a model without releasing its weights. A demo can let you try a model without providing code. Mention those differences in your notes, and link the terms rather than guessing what they allow.
An archived code repository does not tell us whether the model still runs. If the author points to a maintained fork or replacement, include that link and explain which source says so.
Your suggestion goes to a private review queue. Submitting it does not publish a directory page or establish ownership of the model. The review helps keep names, attribution, sources and available ways to use a model clear for everyone browsing.
