Tools & Platforms
Loresight
I kept being the one at the table saying "hang on, let me look it up", or I would remember a detail wrong. I built Loresight to help me out.
It listens while you play. Say a spell name and the spell comes up. Say a monster's name and the stat block is on screen before you've finished the sentence. You tell it which categories to pay attention to (e.g. spells, items, monsters), so it isn't throwing item cards at you every time somebody mentions a rope. The microphone only runs while you've got listening switched on. It also works as a normal reference app so you can use quick search too.
Loresight is free and will never be for sale. If anyone asks you to pay for it, it is not Loresight. The only official downloads are this Google Play listing and the GitHub releases page.
Four systems come built in: D&D 5e (SRD 5.1), the 2024 5e SRD, Pathfinder 2e Remaster under the ORC licence, and Daggerheart SRD 1.0.
You can add more content and systems. A system is just a JSON content pack, and once it's loaded the app treats it like any of the built-in ones. You can browse it, search it, and the assistant will spot its names while you talk. There's nothing D&D-specific buried in the code. Homebrew sits next to the official content rather than replacing it.
Browsing works the ordinary way when you'd rather do it yourself. Search by name or work down through the types. Entries link to each other, so you get from a monster to the spell it casts in one tap. Search only ever covers the system you're playing, so Pathfinder results never turn up in a D&D game.
There's a notetaker as well. It transcribes anything it hears directly to text. It won't identify people, but where possible it will add "new speaker" with a line break so sentences don't run into each other.
All of this runs on the device. Speech recognition is local, audio is processed and thrown away immediately and is never saved, and there's no account, server or analytics anywhere in it. Turn the wifi off and nothing changes. The install is large for the same reason: the speech models are inside the app.
Android is the primary platform and the best-tested build. The Windows, Mac and Linux apps are the same thing, but they aren't code-signed because there's no budget for the certificates those platforms charge for, so expect a one-time "unidentified
Fair warning: the listening is best effort and won't always be 100% accurattrade-off in on-device speech models. I'm also limited to allowed contentonly for this release, since that's what can be handed out, with the attributions in the app. This is why I made it extensible with an import/export system.
There is no affiliation with or sponsored support from any of the systems included.