How AI‑Powered LoRA Techniques Supercharged a September Recruitment Campaign Design
This article details how 58.com’s design team leveraged AI tools such as Stable Diffusion, LoRA models, and prompt engineering to create a visually rich September recruitment landing page, outlining the project background, design strategy, technical exploration, and measurable results.
Project Background
To meet the high demand of the September recruitment season, 58.com launched a special hiring event aimed at driving user conversion, increasing user value, and improving next‑day retention.
Design Strategy
Strengthen brand image with dual reinforcement of brand symbols and colors.
Visual concept focuses on a multi‑industry, job‑rich scene for the September hiring peak.
Technical exploration combines multiple LoRA applications and a custom LoRA model to boost design efficiency.
Implementation Details
Brand Reinforcement
Maintain the glass‑like brand texture and a transparent, positive emotional visual language, using an 8:2 color ratio.
Visual Concept
Background layer uses newspaper and message‑box elements to convey abundant job information; the character layer features multiple figures to represent diverse positions; the foreground adds flying gold coins to enhance the event atmosphere.
Technical Exploration
Multiple LoRA models (glass texture, character, newspaper background) are combined via the Additional Networks plugin, allowing weighted control of each model to generate high‑quality images directly from Stable Diffusion and then fine‑tune them.
Sketch & Composition
After clarifying the visual concept, reference images were collected and composited in Photoshop to build the layout.
Prompt Engineering
Reverse‑engineered prompts were crafted to guide the AI, covering aspects such as texture, subject, clothing, emotion, tone, and environment.
Pose Extraction
Four character poses were selected from extensive searches and extracted using OpenPose for accurate body and facial expression control.
Brand Symbol Control
Segmentation masks were used to position characters and apply the glass‑texture symbol precisely.
Multiple LoRA Application
The Additional Networks plugin sets different weight ratios for each LoRA; the character LoRA receives a higher weight, while the background LoRA is fine‑tuned for optimal effect.
Initial Draft Completion
AI‑generated images were reviewed, the best draft selected, and local re‑painting corrected facial and hand details; an upscaling script added final detail and consistency.
Photoshop Refinement
Final touches added gold‑coin elements and the main title to enhance the overall atmosphere.
Results
The September recruitment campaign met all performance expectations; the AI‑driven workflow not only boosted business metrics but also demonstrated promising potential for future operational applications.
Additional Networks Plugin Overview
The plugin enables mixing multiple LoRA models, provides download and installation instructions, and specifies that LoRA files must reside in extensions\sd-webui-additional-networks\models\lora for activation.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
