Namecall · 1.0.0 (build 20261020)
Everyone in the old photos gets a name.
The Chinese name is 时光点名. In the app, the English name on the screen is Namecall. The pictures below are the English screens.
On the morning of 20 October 2026, the scan engine went through one unorganized backup folder on the author's own computer.
It started at 07:23:08 and finished at 08:01:42 (UTC+8).
The folder holds 165 photos and 115 videos. It is not a separate set of 1,000 photos. All 280 files were scanned. None were skipped.
At the average time per photo measured here, 1,000 photos on this same PC would take about 11 minutes. That figure is an estimate. A separate run of 1,000 photos was not done.
| PC | Lenovo 82SN |
|---|---|
| Processor | AMD Ryzen 7 6800H, 8 cores, 16 threads. Windows reports a clock of 3.2 GHz. |
| Memory | 4 × 4 GB Micron DDR5-6400, 16 GB installed. The graphics chip shares some of it, so Windows sees about 13.7 GB. |
| Graphics | AMD Radeon graphics, driver 31.0.21024.2004. This PC has no separate graphics card. Recognition ran on the processor. |
| System | Windows 11 Home (Chinese), 64-bit, version 10.0.22631. |
| Disk | UMIS NVMe SSD, about 1 TB. The test files are on drive E:, NTFS, about 715 GB. About 24 GB was free when the test started. |
The app is Namecall 1.0.0 (build 20261020). The clock was measured with the scan engine in the current source, Python 3.10.11, ONNX Runtime 1.23.2, CPU only, 6 inference threads. During a scan the program lowers its own priority. That behavior is included in the times below.
| Location | E:\202610视频备份（未整理）, including subfolders. |
|---|---|
| Photos | 165 files (160 JPG, 5 PNG), about 914 MB. |
| Videos | 115 MP4 files, about 51.5 GB. Total length 1 hour 34 minutes 25 seconds. 114 clips are 3840×2160. One clip is 1920×1080. |
Video scan was turned on, in Quick mode: about one look every 2 seconds, and at most 5 people kept from each video. Deep mode was not used.
| All 280 files | 38 minutes 34 seconds |
|---|---|
| 165 photos | 1 minute 50 seconds. Average 0.67 seconds per photo. Half of the photos finished within 0.60 seconds. The slowest photo took 2.6 seconds. About 90 photos per minute. |
| 115 videos | 36 minutes 15 seconds. Half of the videos finished within 7.7 seconds. The slowest video took 2 minutes 55 seconds. |
| Compared with playback | Playing these videos would take 1 hour 34 minutes. Quick scan finished them in 36 minutes, about 2.6 times faster than playback. It does not look at every frame. |
| Faces found | 926 faces. |
| People at the end of the scan | 204 people. |
| After opening the album | The app merges people who look the same. The album then shows 160 people and 6 Often together groups. |
Of the 38 minutes 34 seconds, recognition of the photos and videos took about 38 minutes. About half a minute was loading the model and saving the index.
During the scan, memory use on the whole PC reached 95% (Windows sees about 13.7 GB). That is the whole computer, not this program alone. Near the end, the scan process itself was about 760 MB.
These pictures are the same result, opened afterwards. The faces are from the author's own backup. Nobody has been named yet, so the cards show numbers. Portraits are blurred so nobody can be recognized. The English screenshots use the English interface.
