
How to Map Different Faces to People in a Group Photo
Assign a different replacement portrait to each selected person in one group photo, with a reproducible four-face production test, cost and limits.
Inspect studyFaceSwapAI Datasets & Research
Versioned datasets, measured benchmarks and dated vendor-claim snapshots with machine-readable source files, traceable inputs and explicit limitations.
Technical report
Measurement studies publish exact inputs, transformations, detector boundaries and component values. Public-claim snapshots instead normalize what named vendors state on cited pages at a recorded time; those statements are not treated as independently verified outputs. Private customer media, identity embeddings and unsupported provider-accuracy claims are excluded.

Assign a different replacement portrait to each selected person in one group photo, with a reproducible four-face production test, cost and limits.
Inspect study
Find the best AI face swap tool for your photo, group, video or GIF job. Compare seven official claims pages for free access, watermarks, cost and privacy.
Inspect study
Evaluate a face swap API for authentication, pricing, uploads, async tasks, errors, privacy and production readiness with a reproducible checklist.
Inspect study
Evaluate AI face swap quality with 14 source-linked metrics for identity, attributes, image and video distributions, temporal stability and detector limits.
Inspect study
Inspect four paid photo face swaps using one synthetic source and neutral, left-turn, blink and right-turn targets, with exact files and public JSON.
Inspect study
Is AI face swap safe? Check consent, 24-hour task-file deletion, upstream processing, training disclosures, account linkage, watermarks and takedowns.
Inspect study
A reproducible audit of eight published face-swap reference frames with measured resolution, brightness, contrast, edge detail, face count and readiness points.
Inspect study
A controlled face-swap input study measuring blur, low resolution, darkness, overexposure and JPEG compression while face detection still succeeds.
Inspect study
A controlled group-photo experiment measuring how four detected faces change as the same image shrinks from 1200 to 32 pixels wide.
Inspect study
Inspect one controlled 10-second production video face swap with the exact input, output, 240-frame container data, eight matched samples and public JSON.
Inspect studyUse the versioned JSON files for reproduction or citation. Each record identifies its version, method, variables, source boundary and explicit limitations.
Suggested citation: FaceSwapAI Editorial (2026), FaceSwapAI Face Swap Research and Open Data Catalog, version 1.5.0, https://faceswapai.com/research. Cite the individual landing page and version when using one dataset.
This catalog is not a face-recognition training corpus, provider leaderboard, endorsement or proof of generation realism. Measured input studies do not grade output quality, and vendor-claim snapshots record dated statements rather than independently verified product behavior. The controlled photo and video studies record bounded production outcomes; they do not turn detector, box or pixel observations into universal identity, continuity or provider-quality scores.
Apply the published checks to your own permitted images before spending a credit.