Hi Sivaraam,
Thanks for sharing the announcement mail. I'd seen this project over the weekend, the idea sounded exciting :)
The community is a bit more conservative though whenever anything related to "AI" is built and announced.
I personally don't see any privacy concerns here (those images are already public) and would love to integrate it with the app in case we have any API/SDK available.
Congratulations team for building such an impressive project :)
Best, Ritika
On Tue, 5 May, 2026, 12:28 am Kaartic Sivaraam, kaartic.sivaraam@gmail.com wrote:
Great work team! Good to see work on the semantic search area. Gave it a few attempts, the top results are decent.
When this progresses, I would love to also see this integrated with the Commons Android app!
Kudos again on your work on this!
On 3 May 2026 5:25:36 am IST, Vasanth Gopa via Commons-l < commons-l@lists.wikimedia.org> wrote:
Hello everyone,
We would love to get your feedback on a tool we have been working on: https://wise.wmcloud.org/
This is a search tool for images and videos on wikimedia. Currently is searching only on media of the day (approximately 5000 videos). It
searches
only the visual content of the file (not on metadata, filename, or structured data). It also does face detection and recognition.
Please give it a go. Here's some examples:
*Semantic example queries:* "man at a train station" "एक व्यक्ति रेलवे स्टेशन पर" (Hindi) "horse in an airplane" "విమానంలో గుర్రం" (Telugu) "man with a flower" "pirate with a pistol"
*Face search (select "Faces"):*
Click the green "Image" button and paste an image URL
Example inputs:
https://upload.wikimedia.org/wikipedia/commons/0/0d/Neil_Armstrong_pose.jpg
[warning: some videos are ogv videos and your brownser may not play them, so watch them on Commons, i.e., after clicking the search result, scroll down to the media metadata table, and follow the link to commons]
*Key features include:*
- Semantic search using natural language to find relevant images on
Commons
from the visual content only (not the structured data or description, only the image itself)
- Face search: upload or paste a face image, and the tool will try to
identify and locate that person across images and videos, including timestamps where they appear
- Audio search: search within audio files to find relevant segments
- Multilingual search: supports queries in multiple languages
We are actively improving the tool and would really appreciate any feedback, suggestions, or ideas from the community. Here's some more immediate future plans:
- Search on all wikimedia images (instead of only videos / media of the
day)
- Show similar images when uploading an image to suggest filenames,
categories, and other relevant metadata
More technical details and a place to share feedback at https://commons.wikimedia.org/wiki/Commons:Wise Phabricator: https://phabricator.wikimedia.org/T424092 Announcement on Village pump:
https://commons.wikimedia.org/wiki/Commons:Village_pump#Introducing_WISE%3A_...
Regards,
Gopa Vasanth https://gopavasanth.github.io/ Twitter https://twitter.com/gopavasanth1999 | LinkedIn https://www.linkedin.com/in/gopa-vasanth/ | GitHub https://github.com/gopavasanth | Gerrit https://gerrit.wikimedia.org/r/#/q/gopavasanth
“Yesterday is not ours to recover, but tomorrow is ours to win or lose.”
-- Sivaraam
Sent from my Android device with K-9 Mail. Please excuse my brevity.