Facial Recognition Guide
Facial recognition workflows can follow two approaches using: community-tier services and commercial-tier services. Both benefit from using quality images and careful review of results.
Use Quality Images
Select clear, highest available resolution images of your subject. Include at least one headshot-style image. Community-tier services allow only one image per search, but commercial-tier services allow for multiple (3) images per search. When providing multiple images, include different angles and/or appearances of the subject from different periods of their career to broaden the scope of relevant matches.
Community-Tier Services
Community-tier facial recognition services are free and maintained by Stash community user(s). They are simple to use, but have limited accuracy, higher rates of false positives, and unreliable confidence scores. Your visual verification (aka the “eye test”) is essential.
- Stashface - Trained on StashDB images
- Megaface - Trained on a dataset which includes IAFD, indexxx, etc.
Consider this method a simple first round that can often identify mainstream performers without hassle.
Commercial-Tier Services
As of early 2026, changes to PimEyes and Yandex have reduced the efficacy of the commercial-tier workflow.
Commercial-tier services produce more reliable facial recognition matches. Freemium limitations restrict access to their source URLs, making identification difficult. A secondary reverse-image search step (i.e. Yandex) is required to trace the result images and gather additional context on the subject’s identity.
Step 1 - Facial recognition services
- FaceCheck - Best overall choice
- PimEyes - Broadest scope but results becoming more restrictive
- Lensio
Step 2 - Tracing result origins
- Yandex Visual Search - Copy-and-paste results from step 1 into Yandex.
Review Results & Beware of False Positives
Doppleganger effect
It’s not uncommon that facial recognition matches may include unrelated look-alikes.
Arbitrary confidence scores
Confidence scores are not standardized across services. Even high scores can be incorrect. For example, incorrect matches can occur with an 82% confidence match on FaceCheck, or a 100% confidence match on community-tier services. Always use the eye test in conjunction with results.
Non adult-sources
Matches from non-adult sites may be irrelevant and inaccurate. Use caution when assessing these.
Use Facial Similarity Tools
When identification remains uncertain, facial similarity tools can help estimate resemblance similarity between two images. These are not forensic-grade tools. They provide approximations that will vary based on image image quality, angle, lighting, age differences, and other factors.
Tips
For hard-to-find unknown performers, when facial recognition tools fail, try these methods to identify an unknown performer:
-
Follow the trail on Yandex Visual Search. Submit your source image or an uncropped screenshot of the video frame containing the unknown performer. Follow relevant matches to galleries, forum posts, or tube site uploads that may contain more details about the source content.
-
Triangulate with scene partner or studio name. If the scene partner of the subject is known, reference the scene partner’s IAFD page and Check Scene Pairings section. You will need to visually scan the section for the unknown subject’s face. If the scene partner is also unknown, use facial recognition tools to identify them.
-
Use third-party retailer sites. Third-party retailer sites, such as Adult DVD Empire, can be used to filter performers (scene partners) or studios by Clips, providing a thumbnail view of their scenes which can be manually reviewed to match your own subject’s content.
Resources & Tools
Commercial Facial Recognition Services
These premium services provide reliable results but do not disclose the source URL of the matches. Extra steps must be taken to trace matched images.
- FaceCheck - Reliable, often sourcing adult database sites, easiest to use with Yandex.
- PimEyes - Reliable, uses broader scope of sources, but matches are becoming more restrictive
- Lensio
Community Facial Recognition Services
Community projects provided by @CC1234:
- StashFace - Trained on StashDB images. Prone to more false positives than commercial options. Useful if the performer already exists on StashDB. Confidence rate of matches will be misleading. Use the “eye test” to verify matches.
- MegaFace - Trained on @olddude’s meta4allphotos metadata. Prone to more false positives than commercial options, but covers different sources than StashFace (IAFD, indexxx, PornPics, Bang, etc). Confidence rate of matches will be misleading. Use the “eye test” to verify matches.
Other Facial Recognition Services
- CamGirlFinder - Facial recognition reverse image search specializing in “cam girls”. Sources typical “cam girl” platforms.