I use musescore for that purpose, works pretty reliably.
https://musescore.com/upload?format=PDF
Just need to be logged in
You need to be paying them a subscription or a one-off fee, by the looks of it. It says “Upload your first for free”, but it still wants £3.99.
Hi all,
There is a very promising open-source OMR project called HOMR that converts sheet music PDFs directly into MusicXML, and it works quite well.
I think the logical next step for workflow automation is tying Dorico into an AI agent loop (e.g., via dorico-mcp-server). That way, an agent could pass MusicXML into Dorico, visually compare Dorico’s rendered PDF back to the original scan, and use that visual feedback loop to fix formatting or notation errors automatically.
Has anyone experimented with MCP servers or AI agent workflows alongside Dorico yet?
Hello Copello,
welcome to the forum.
I saw some MCP related threads recently.
https://forums.steinberg.net/search?q=%23dorico%20mcp%20order%3Alatest
The first problem I can see with this is the Dorico MusicXML import options. You have choices about how much of the layout of the imported XML to preserve and how much you want to respect your own options set within Dorico.
I can imagine a perpetual battle being set up because the AI wouldn’t really know where or why a difference originated.
Exporting to MusicXML first feels like an unnecessary extra step. A native Dorico CLI interface (e.g., ./dorico input.dorico --export output.pdf) integrated directly into an MCP loop would be a far better approach.
I tested something similar on Linux using ./musescore—it works reasonably well. Oddly enough, taking screenshots in Dorico, uploading them, and prompting against the visual output yields pretty impressive results.
It makes you wonder if Steinberg might eventually integrate an AI assistant or LLM rag agent directly into their web documentation.
Thanks for link @TonH. Looks like there’s a growing community interest into this topic.