Turn a placard into a conversation.

Humy is a web-based platform that lets museums and cultural institutions build immersive, conversational experiences with historical figures. A figure can be given an archive of primary sources your curators approve — accession and archive records, photographs, paintings, audio — which it searches during the conversation and quotes with the reference attached, so the experience stays anchored to your curatorial narrative.

Three ways: a QR code on an object label or wall panel, a direct link shared in marketing and the members' newsletter, or an iframe embedded in your exhibition microsite or a gallery kiosk. No app download. Each surface can be configured for anonymous access or to require a visitor's name — your choice per surface.

Both. Visitors can type a question and receive richly formatted text with images, references, and tables. They can also start a real-time voice call with the figure, powered by Gemini 3.1 Flash Live, and have a natural spoken conversation of up to 10 minutes — the same Humy, two modes. Images and rich formatting are not available during a voice call.

No. Humy is built to augment curatorial expertise, not replace it. Your curators provide the sources and the interpretive direction. Humy extends that expertise to every visitor in their language, around the clock — freeing front-of-house staff to focus on high-value, in-person interpretation.

An audio guide plays a fixed script and has to be re-recorded for every language and every reading level. A Humy figure answers the question the visitor actually asked, in 50+ languages, at the reading level you set — and it can go deeper for the visitor who wants more without holding up the gallery. There is no hardware to buy, clean, charge, or replace: visitors use the phone already in their pocket.

No. Humy runs in the browser. Print a QR code for an object label, embed an iframe in your exhibition microsite, run it on a kiosk you already own, or share a link in a marketing email. There is nothing for visitors to download and nothing new for your team to maintain on the gallery floor.

Yes. Because the experience is a link, a QR code, or an iframe rather than installed hardware, a figure built for a temporary show travels with it to the next venue and can stay open on the microsite after the show closes. Pricing is scoped the same way — a single exhibition run, a permanent gallery, or an institution-wide deployment.

You edit the sources, and the figure changes with them. New scholarship, a reattributed object, a correction a visitor points out, a loan that arrives late: each of those is a change in the curation workspace rather than a reprint, a re-record, or a call to the fabricator. Interpretation that would otherwise be fixed for the life of an installation stays current between capital projects, and a figure built for a temporary show can be revised and reused when it tours.

Yes. Publish a separate edition of the same figure for school visits — its own reading level, instructions, and language — and turn on named entry for that QR code or link so the group's conversations are identifiable. Teachers can also use the same figures for pre-visit and post-visit work from the classroom.

It depends on how much of your collection you want the figure to draw on, but a focused exhibition figure is typically ready in a few weeks: we scope it with your curators, curate and transcribe the sources together, then your team tests it before the QR codes go up. Write to hello@humy.ai and we will scope the timeline against your opening date.

A figure answers out of the archive your curators select. When it quotes a passage, the model emits only an anchor pointing at part of a document and our backend resolves it against the original — the model never writes the quotation itself, and every quote links back to the record it came from. Ordinary conversation runs on a light, fast tier, and archive-grounded answers move up to a stronger one for reliable multi-turn source search. As with any AI system the surrounding prose is model-generated, so we make the underlying document one click away rather than asking visitors to take it on trust.

Up to 50 sources per figure: scanned books and articles, handwritten letters, catalogue and archive records, paintings, photographs, audio recordings of speeches, and 3D-scanned artifacts. Scans without a text layer are transcribed page by page with Gemini 3.1 Pro, including historical handwriting, up to 1,000 pages per source. Transcription uses credits and you approve an estimated cost before any page is processed.

Your curators do, source by source. Each one can stay private to your institution or be shared across it, and figures can additionally draw on sources curated by the Humy team if you want them to. Everything is managed in one curation workspace where your team uploads, organizes, transcribes, and reviews the material.

Yes. Our team builds tailor-made Humies alongside your curators using your biographies, transcripts, images, and video. You can also publish multiple editions of the same Humy — a family edition, a scholar's edition, a tour-group edition — each with its own reading level, instructions, and language.

Every conversation is logged in the dashboard. You can export the full history as PDF or CSV. Our data team can also prepare custom reports — most-asked questions, language distribution, visitor journeys — to feed summative evaluation, label and panel revisions, docent briefings, funder reports, and future programming.

Pricing is flexible — from a single exhibition run to a permanent gallery or an institution-wide deployment. Contact us for a quote scoped to your needs.