When the Domestic Beauty SDK Market Enters the Stage of Refined Competition

As the domestic beauty SDK (Software Development Kit) market gradually enters the stage of refined competition, many technical teams have begun to turn their attention to overseas markets. After all, in regions such as Southeast Asia, the Middle East, Europe, and the United States, live-streaming e-commerce and social entertainment apps are experiencing robust growth, and the demand for beauty features is also releasing rapidly.

However, the overseas market is not a "simple replication of the domestic model". Differences in user aesthetics, device environments, and compliance requirements across various regions pose new challenges to beauty technology.

Letmagic Beauty SDK's Overseas Layout

Letmagic Beauty SDK started its overseas market layout in 2023. Initially, it also encountered many difficulties:

For Southeast Asian users: The beauty effects designed suffered from "excessive whitening" due to insufficient skin tone adaptation.

For the European and American markets: The tailored special effects frequently lagged on low-end devices due to device compatibility issues.

It was not until we deeply integrated AI technology into the overseas version of the SDK that we truly found the key to breaking through these challenges.

This article will, from the perspective of practical implementation, discuss how AI helps Letmagic Beauty SDK solve the pain points in the overseas market and share our insights on the path of global innovation.

I. "Special Challenges" in the Overseas Market: More Complex Demand Scenarios Than Domestic

Before entering the overseas market, our team spent 3 months visiting 12 partners in 6 core markets across Southeast Asia, the Middle East, Europe, and the United States. We found that the complexity of overseas beauty demand far exceeded expectations, mainly concentrating on three aspects.

1. Aesthetic Differences: Not "The Whiter, the Prettier", but "Natural Optimization That Fits Local Skin Tones"

Domestic users have a high acceptance of "fair and translucent" skin tones, but aesthetic preferences vary greatly across different overseas regions:

Southeast Asia: Users prefer to "retain their original skin tone while only optimizing blemishes". For example, Indonesian users hope to keep their wheat-colored base after skin smoothing; Malaysian users care about not losing the details of their makeup during the skin-smoothing process.

Middle East: Users pay more attention to "facial three-dimensionality" (e.g., slightly enhancing the shadow of the nose bridge and optimizing the contour around the eyes) rather than simply pursuing whitening.

Initially, we adopted the domestic whitening parameters, but received a lot of negative feedback like "the skin tone looks fake, like wearing powder" and "lost my own characteristics". Later, we realized that the core of overseas beauty is not to "change the skin tone", but to "optimize based on the original skin tone"—which places higher requirements on the accuracy of skin tone recognition.

2. Device Fragmentation: A Vast Adaptation Range from High-end iPhones to Entry-level Android Devices

The mainstream devices in the domestic market are concentrated in the mid-to-high-end segment, but the range of devices in many overseas regions is surprisingly wide:

India: Entry-level Android devices priced at 500 yuan account for more than 40% of the market.

Europe and the United States: Although iPhone users have a high proportion, many users still use old models such as iPhone 8 and iPhone SE.

Traditional beauty SDKs face frequent problems on these devices:

On entry-level Android devices: When beauty functions are enabled, the frame rate drops directly from 30fps to 15fps, with obvious lag.

On old iPhones: When running special effects, excessive memory usage causes the app to crash.

A Southeast Asian social app client reported that due to device adaptation issues, the user churn rate related to the beauty function reached 12%—a situation rarely seen in the domestic market.

3. Compliance and Localization: No "Red Lines" Can Be Crossed in Data Security and Privacy Protection

Compliance requirements vary greatly across different overseas regions. For example:

The EU's GDPR (General Data Protection Regulation) and the personal data protection laws of some Southeast Asian countries have strict regulations on the processing of facial data.

The "cloud-based facial data processing" model commonly used in China simply does not work in many overseas regions: GDPR requires "localized data storage", and some Middle Eastern countries prohibit the transmission of facial data overseas.

In addition, localization needs also include language support and holiday-themed special effects:

Holiday themes: Corresponding themed stickers need to be launched during Islamic Eid al-Fitr and Western Christmas.

Language: The interface copy not only needs to be accurately translated but also conform to local cultural habits.

All these require close collaboration between the technical and operational teams.

II. How Does AI Break Through the Dilemma? Three Core Technologies Adapt to Overseas Demands

In response to the special challenges in the overseas market, we did not simply modify parameters. Instead, we reconstructed the core capabilities of the overseas version of the beauty SDK using AI technology, focusing on solving the three major issues of aesthetic adaptation, device compatibility, and compliance security.

1. AI Skin Tone Semantic Segmentation: "Personalized Beauty" That Fits Local Aesthetics

To address regional aesthetic differences, we developed the "AI Skin Tone Semantic Segmentation" technology, with two core steps:

Skin tone identification: An AI model first identifies the user's "original skin tone type" (e.g., wheat-colored skin in Southeast Asia, light skin in Europe and the United States, dark skin in the Middle East).

Exclusive parameter generation: The model generates exclusive beauty parameters based on aesthetic data from different regions.

Regional Application Examples:

Southeast Asia: AI retains the hue of the original skin tone, only accurately removing skin blemishes (acne marks, spots) while enhancing skin translucency to avoid "excessive whitening".

Middle East: AI identifies key points of the facial contour, slightly enhancing the shadow of the nose bridge and jawline to improve facial three-dimensionality (without changing the base skin tone).

We also collected aesthetic preference data from over 2,000 users in 6 major overseas regions to train "region-specific beauty models":

The Indonesian version focuses on "makeup retention" (suitable for local users’ preference for heavy makeup).

The European and American version emphasizes "natural skin smoothing" (to avoid over-retouching).

Test data shows that after enabling AI skin tone adaptation, the satisfaction of overseas users with the beauty effect increased by 45%.

2. AI Dynamic Performance Adjustment: Enabling Smooth Operation Even on Low-end Devices

To solve device fragmentation, we implemented "AI Dynamic Performance Adjustment", which works as follows:

Hardware detection: When the SDK is launched, AI first detects the device's hardware configuration (CPU model, GPU performance, memory size).

Performance level adjustment: The AI automatically adjusts the "performance level" of the beauty effect based on the detection result:

High-end devices: Enable full functions (4K resolution, 3D stickers).

Mid-range devices: Retain core functions (skin smoothing, facial shaping, 2D stickers).

Low-end devices: Activate "lightweight mode" (only basic skin smoothing and whitening; complex special effects turned off).

Device-specific Optimization Examples:

Entry-level Android devices (India): AI reduces the complexity of the skin-smoothing algorithm (switching from "partitioned skin smoothing" to "local blemish removal") and turns off real-time shadow calculation. This reduces CPU usage from 60% to below 30% and stabilizes the frame rate above 25fps.

Old iPhones: AI optimizes memory management, reducing the memory usage of the beauty function by 20% to avoid crashes.

More importantly, this adjustment is "dynamic and real-time": If the device temperature rises during a user's live stream, AI will automatically reduce the complexity of special effects to avoid lag caused by overheating.

A Middle Eastern live-streaming app client reported that after enabling this function, the crash rate of the beauty function on low-end devices dropped from 8% to less than 1%.

3. AI On-device Inference: A "Data Security Solution" Balancing Compliance and Privacy

To meet overseas compliance requirements, we completed the inference process of the AI model entirely on the "device side" (local user device), with two core designs:

Local processing: The AI model of the overseas version of the SDK is pre-installed on the user's device. The processing of facial images and the generation of beauty parameters are all completed in real-time locally—fundamentally avoiding the compliance risks of "cross-border data transmission".

Data desensitization: We developed "AI Data Desensitization" technology: The processed facial data automatically removes information that can identify personal identities (e.g., iris and fingerprint features). Even if the data is stored locally, it will not involve privacy leakage.

Currently, the overseas version of Lanji Beauty SDK has passed the compliance certification of GDPR and the personal data protection laws of some Southeast Asian countries, becoming the preferred solution for many overseas apps.

Localization Bonus: Automatic Themed Special Effects

AI can also realize "automatic generation of holiday-themed special effects":

Christmas: AI recommends themed stickers (Christmas trees, snowflakes) based on local users’ usage habits.

Eid al-Fitr: AI generates special effects with Arabic patterns.

This eliminates the need for developers to manually update materials, greatly reducing localization costs.

III. Overseas Implementation Cases: "Value Conversion" from Technology to Business

Technology ultimately serves business. The AI-enabled solution of Letmagic Beauty SDK has been implemented in the businesses of many overseas clients, bringing practical value improvements.

Case 1: Southeast Asian Social App — Increased Beauty Satisfaction and User Retention

Before integration: The app had low user retention due to two issues:

Beauty effects did not conform to local aesthetics (excessive whitening).

Poor device adaptation (lag on low-end devices).

After integrating Letmagic’s overseas SDK:

AI skin tone adaptation solved the "excessive whitening" problem.

AI dynamic performance adjustment stabilized the frame rate of the beauty function on low-end devices above 25fps.

Business Results (3 months after launch):

Usage rate of the beauty function: Increased from 35% to 62%.

Next-day user retention rate: Rose by 9%.

Monthly active users (MAU): Increased by 15%.

The client reported that the beauty function has become one of their core selling points for attracting new users.

Case 2: Middle Eastern Live-streaming E-commerce Platform — Compliance Implementation and Local Market Penetration

Before integration: When expanding into the Saudi Arabian market, the platform was repeatedly required to rectify by local regulatory authorities due to data compliance issues (cross-border transmission of facial data).

After integrating Letmagic’s overseas SDK:

Adopted the AI on-device inference solution: All facial data is processed locally (no cross-border transmission), and the platform successfully passed local compliance certification.

Optimized for local aesthetics: The AI model enhanced facial three-dimensionality, making the display effect of products (cosmetics, clothing) more prominent for anchors.

Business Results:

Anchor live-streaming rate: Increased by 20%.

User order conversion rate: Rose by 11%.

The platform successfully penetrated the Saudi Arabian market.

IV. Future Plans: Continuously Deepening Overseas Market Presence to Make AI Beauty "Understand Global Users Better"

There is no end to the path of innovation in the overseas market. Next, Letmagic Beauty SDK will continue to make efforts in three directions:

1. Deepening Region-specific Models: More Accurate Localization Adaptation

Broaden data coverage: Collect more user data from undercovered regions (e.g., African and Latin American markets) and train corresponding region-specific AI models.

Segment user groups: Develop "segmented user group beauty models" to address differences among users of different ages and genders (e.g., "natural light beauty" for overseas teenagers, "anti-aging optimization" for mature users).

2. Expanding AI Multimodal Interaction: Enhancing User Experience

In the future, we will expand AI from "image processing" to "multimodal interaction":

Speech recognition integration: Recommend corresponding filters based on the user's voice emotion (cheerful, calm).

Gesture recognition integration: Automatically trigger dynamic stickers when the user makes a specific gesture (to enhance social interaction).

This multimodal experience has great potential in overseas social and live-streaming scenarios.

3. Improving the Compliance Ecosystem: Reducing the Threshold for Clients to Go Overseas

In addition to ensuring the compliance of our own beauty SDK, we plan to launch an "Overseas Compliance Solution" to provide clients with services such as:

Data security consulting.

Localization filing guidance.

This helps clients quickly respond to compliance requirements in different regions and reduce the technical and operational costs of going overseas.

Conclusion

In the overseas market, "technical adaptation" is only the foundation, and "understanding local users and respecting local rules" is the long-term way to go. The AI empowerment of Lanji Beauty SDK essentially uses technology to break down "geographical barriers", allowing users in different regions to enjoy a beauty experience that meets their own needs.

In the future, we will continue to focus on "technological innovation" and move forward steadily on the path of the global overseas market. We also look forward to working with more overseas enterprises to jointly explore new possibilities in beauty technology.