@seclink: Development of Face Swapping Technology: DeepFaceLab (DFL): Industrial-grade video/SaaS pipelines are basically no longer used. Reasons for elimination: it requires case-by-case handling, each face swap requires manual retraining of the network for hours, resulting in extremely high computational cost and human intervention, making it completely unscalable (Scale). Only a very few film VFX studios…

X AI KOLs Following News
face-swapping deep-learning computer-vision video-processing open-source ai-ethics

Summary

The article summarizes the reasons for the elimination of face swapping methods such as DeepFaceLab, Inswapper_128, GFPGAN/CodeFormer, pointing out that due to issues like cost, resolution, and temporal flickering, they have largely exited industrial-grade applications, with only a few film VFX scenarios still using them.

Development of Face Swapping Technology: DeepFaceLab (DFL): Industrial-grade video/SaaS pipelines are basically no longer used. Reasons for elimination: It requires case-by-case handling, each face swap requires manual retraining of the network for hours, resulting in extremely high computational cost and human intervention, making it completely unscalable (Scale). Only a very few film VFX studios still retain it for high-budget one-off productions. Original Inswapper_128 (FaceFusion/Roop default weights): Prohibited for serious commercial projects. Reasons for elimination: First, InsightFace's non-commercial open source license carries compliance risk; Second, the 128x128 resolution is too low, resulting in blurry facial features and ghosting edges after swapping. GFPGAN / CodeFormer face restoration: Has faded out of video streams. Reasons for elimination: Frame-by-frame face restoration leads to "plastic face" and severe temporal flickering. In videos, faces alternate between bright and dark, sharp and blurry, failing industry review standards.
Original Article
View Cached Full Text

Cached at: 06/28/26, 04:04 PM

Face Swapping Technology Development:

DeepFaceLab (DFL):

Industrial-grade video/SaaS production pipelines have largely abandoned it.

Reason for obsolescence: It requires case-by-case processing — every new face swap demands hours of manual retraining of the network, resulting in extremely high computational cost and human intervention, making it completely unscalable.

Only a handful of high-budget film VFX studios still keep it for one-off productions.

Original Inswapper_128 (FaceFusion/Roop default weights):

Prohibited for serious commercial projects. Reasons for obsolescence:

First, InsightFace’s non-commercial open-source license poses compliance risks (legal exposure);

Second, the 128×128 resolution is too low, leading to blurry facial features and edge ghosting after the swap.

GFPGAN / CodeFormer Face Restoration:

Fading out of video streams. Reasons for obsolescence:

Frame-by-frame face restoration introduces “plastic face” artifacts and severe temporal flickering. In a video, faces alternate between bright and dark, sharp and blurry — unacceptable for industrial review.

Similar Articles

@IndieDevHailey: Stunning! Has face-swapping technology evolved to this level? Transform into any celebrity, any influencer, anyone you want to be in one second! The strongest local AI face-swapping tool in 2026, with 28.4k+ Stars on GitHub, it's none other than FaceFusion! It can make the entire video look as natural as if it were originally shot, from photos to videos,...

X AI KOLs Timeline

This tweet introduces a local AI face-swapping tool called FaceFusion, which has 28.4k+ stars on GitHub. It claims to achieve natural and realistic video face swapping, supporting features such as photo to video, multi-person scenes, expressions, and lip synchronization.

@leeoxiang: I directly gave these skills to my wife to use, and it turned out disastrously: 1. Domestic environment + Agent got stuck at the first step of installing ffmpeg; 2. The frontend code generated by domestic large models simply didn't meet the requirements I set in my skill. It's really hard for users in China who can't use a VPN or access the best models.

X AI KOLs Following

User @leeoxiang tried to give domestic users a video editing skill based on Claude Code and hyperframes, only to find that network restrictions and insufficient capability of domestic large models caused installation and code generation failures, reflecting the obstacles domestic users face in using advanced AI tools.

@seclink: https://x.com/seclink/status/2067968283492712846

X AI KOLs Following

This article, based on the sharing of researcher Victoria Lin, systematically reviews the mainstream technical approaches of native multimodal large models (Chameleon, Transfusion, MOT) and their pros and cons. It points out that multimodal AI is still in the early exploration stage, with open problems such as gaps in scaling laws, inconsistency between image understanding and generation encoding, and connection with the physical world.