TencentCloud Image Face Fusion Pro

Problem

In AI-generated imagery, virtual makeup, and stylized template workflows, a common task is transferring a real face onto a template image while preserving the template's composition, lighting, and style. Building this manually often means handling image encoding, Tencent Cloud authentication, choosing FuseFaceUltra parameters, parsing the synchronous result, and adding failure handling. Letting a model 'generate' the fusion result directly is less reliable and can easily mislead users.

How It Works

The skill wraps Tencent Cloud's Face Fusion (Professional) API into a script-based flow. Typical inputs are --face for the user face image and --model for the template image, both accepted as URLs or local paths. It calls main.py to request the FuseFaceUltra API and supports --swap-model-type for different output modes: 1 general entertainment, 2 cinematic natural, 3 cinematic high-quality natural, 4 high-similarity for ID photos, 5 friendly to closed eyes or occlusion, and 6 extremely high similarity for e-commerce, ID, or cultural tourism use. The result is returned synchronously, and the AI-generated watermark is enabled by default through --logo-add 1. Before use, configure Tencent Cloud API keys in environment variables and enable the service; do not hard-code secrets.

Boundaries

Input sizes are constrained by the API: base64 payloads should be no larger than 10M, and URLs should be no larger than 20M. This skill targets image face fusion, not real-time video, batch moderation, or copyright clearance. If the API call fails, return a clear error instead of inventing a fusion result.