FaceMagic

Historical 2021 policy: FaceMagic said uploaded photos/GIFs/videos were processed for face swapping, facial-feature data was used for the function, generated content was stored for 24 hours, and uploads were stored on remote servers in Hong Kong during use. It also stated it did not use those photos/facial features for face recognition or unique biometric identification. These are historical company statements, not an independent audit.

Current Google Play data-safety disclosure: the developer states no data is shared with third parties; the app may collect app activity, app info/performance and device/other IDs; data is encrypted in transit; users can request deletion.

Deepfake risk: FaceMagic can create convincing identity transformations. Historical terms prohibit unlawful, pornographic, privacy-invasive, child-exploitative and rights-infringing content and place responsibility on users. Independent reporting in 2022 raised concerns about non-consensual sexual deepfake use and advertising. The existence of restrictions does not by itself establish how effectively they are enforced.

[COMPANY-REPORTED] FaceMagic historically described “advanced face recognition technology” and processing through a “Deep Neural Network.” Current materials show model-branded generation systems, but do not publish technical architecture.

Not publicly disclosed: training datasets, face/video dataset provenance, model architecture, facial embeddings, landmark stack, identity-preservation method, GPU/cloud vendor, inference pipeline, temporal-consistency method, rendering stack, encryption implementation, latency benchmarks, open-source dependencies or third-party model/API providers.

General descriptions of GANs, autoencoders or DeepFaceLab on the FaceMagic blog should not be treated as proof that FaceMagic itself uses those exact architectures.

Direct overlap: Reface, DeepSwap and FacePlay compete most closely in consumer face swapping, templates and social-ready content. Broader substitutes: Runway, Kling, Pika, CapCut and other generative-video editors compete for AI video creation time and spend but are not exact face-swap equivalents.

[INFERENCE] FaceMagic’s differentiation is its combination of legacy face-swap workflows with newer text/image/reference video generation and credit-based monetization in one consumer-oriented interface. The moat is unclear because model provenance, proprietary research and defensible technical IP are not publicly documented.

Historical provider vs. current owner: 2021 FaceMagic legal pages name DeepArt Limited, while Google Play and trademark records now name Deep Creation Limited. The trademark assignment gives strong evidence of a rights transfer, but no public document reviewed explains the full transaction or whether Deep Creation acquired the whole FaceMagic business.

DeepArt / Insight Technology: 2022 reporting described Insight Technology LTD as a renamed/related operator identity. Current 2025 listing documents again identify DeepArt Limited inside the Xinguodu group. Because naming/operating roles changed over time, historical app-store names should not be used as proof of current ownership.

Founder identity: Product Hunt identifies “Ange” as a maker, but no reliable evidence reviewed establishes the account as the founder, CEO, inventor or legal owner.