What 520 AI portraits taught us about running a booth at a campus event.

On 31 October 2026 we ran the X-CAPTURE AI Portrait Booth at TEDxKMCT 2026, on the KMCT campus in Calicut. It was the first deployment of X-CAPTURE, the XSEDES product family for photo and video booths. Across one main booth and two tablet stations, the booth logged 559 futures imagined and generated 520 AI portraits.

A first deployment is a test. This article is what the station exports taught us. The full figures are in the TEDxKMCT 2026 case study and the engagement report.

One style, made for the event.

Every portrait used one style, built in the event’s own poster language. It came in four layouts and carried the guest’s name, the future they typed in and a line to go with it.

One style made the 520 portraits read as one body of work. It also meant that a guest who wanted a different result had one option: take the picture again.

Retakes are a signal, not a fault.

The booth recorded 252 retakes, about one for every two finished portraits. A retake is a guest choosing to stay at the booth rather than leave with the first result.

We read that as demand for choice. Two or three event styles would turn many of those retakes into second sessions, each with its own portrait.

Plan for the breaks, not the average.

Activity rose through the morning and peaked in the hour from 15:00, with 139 pictures and 86 portraits across the three stations. The hours from 13:00, 14:00 and 17:00 each brought close to 100 pictures.

Queues form when sessions in the hall end. An hourly average hides that. Plan capacity for the busiest break in the programme, and place stations where people leave the hall.

The tablet stations carried 42 % of all portraits. In the peak hour Station 2 alone took 58 pictures, more than the main booth’s 51. Spreading the load across stations kept one queue from setting the pace for the day.

Know what happens when generation fails.

Generation needs an internet connection and a cloud image service. On the day, 16 of 536 generation attempts failed, a success rate of about 97 %.

Each failure could be retried on the spot, so it cost the guest seconds, not their portrait. The average time from capture to finished portrait was 15.1 seconds, and the three station averages sat between 14.8 and 15.2 seconds from the first hour to the last.

Decide the retry path before the event. Then decide what the booth does if the connection drops: capture now and deliver later, with a way to match each portrait to the right person.

Deliver to the phone, then measure the download.

Printing was not offered at this edition. Each portrait went home through a QR code on the final screen. The booth showed 425 QR codes, 82 % of the portraits generated.

The system counts codes shown, not downloads or shares. That is the gap to close next. A short download link with its own count, or an opt-in email, would show what happened after the guest walked away.

Report the whole window, and label it.

The programme ran from 9:00 am to 5:00 pm. The booth was active from 08:13 to 18:31, which covers set-up, testing and use after the programme. The hours from 17:00 and 18:00 logged 79 portraits.

We report the whole logged window and say what each part of it was. A report that trims the edges to match the programme would be tidier and less true.

Count from the exports, not from memory.

The report was built from the three station exports, added together without adjustment. The system records sessions and generations. It does not identify guests, so a session is not a person: one guest could run several sessions, and a group portrait is one session.

That is why we count futures imagined and portraits, not people. Every number in the report carries its source and its condition.

What we will change next time.

• Offer two or three styles, so a second session gives a new result rather than a retake.
• Add a station near the hall exit for the breaks.
• Measure downloads as well as QR codes shown.
• Print on site where the brief needs a keepsake. Printer integration has been tested but not yet used at an event.

Bring an AI portrait booth to your event.

Read how the X-CAPTURE AI Portrait Booth works, browse the X-CAPTURE range or see the TEDxKMCT 2026 case study. For the planning questions that come before a booking, read AI photo booths for brand activations. Then tell us about your event.