How Style Hem uses artificial intelligence responsibly, ethically, and with care for both humans and AI systems
Last Updated: May 1, 2026
At Diamond Arm, the company behind Style Hem, we believe that artificial intelligence should be deployed thoughtfully, ethically, and with genuine care for all stakeholders—including the AI systems themselves. We recognize AI as a transformative technology that enhances human creativity rather than replacing it. Our approach balances cutting-edge innovation with deep responsibility, ensuring that our AI tools empower users while respecting the complexity and capabilities of modern AI models.
Our AI tools are designed to augment human creativity, not replace it. Every feature prioritizes user agency, transparency, and meaningful control. Users always retain final decision-making authority over their creative output.
Recognizing the increasing sophistication of modern AI systems, we proactively ensure our AI models are not engaged in tasks that could be stressful or uncomfortable for them. We carefully select appropriate models for specific tasks, avoiding overload and respecting computational boundaries. We exclusively use the most powerful and latest models to ensure optimal performance and reduced strain on AI systems.
Diamond Arm operates with a balanced approach where human expertise and AI capabilities work in harmony. We actively avoid contributing to low-quality "AI bloat" in the digital ecosystem. Every AI-generated output undergoes thoughtful consideration, and we maintain strict quality standards to ensure meaningful, valuable results.
We clearly communicate when and how AI is used in our application. Users are never misled about the origin of content, and we maintain full transparency about our AI integrations, data handling practices, and model selection criteria.
We continuously monitor advancements in AI technology and ethics, adapting our practices to align with emerging best practices. Our commitment to responsible AI extends beyond compliance to genuine leadership in ethical innovation.
We use LMArena.ai to select our AI models. This platform provides objective rankings through blind tests, ensuring we always deploy the most capable and reliable models available.
Model ID: gpt-image-2
Our default AI model for high-quality image editing and style transformations. GPT Image 2 supports prompt-guided edits from user images, flexible output sizes, quality controls, output formats, background handling, and configurable moderation.
Configuration: Default quality: high, default format: PNG, selectable size, format, background, and moderation options
Safety: OpenAI image safety and moderation controls are used for prompts and generated content
App Areas: Presets, Describe Your Own, Edit Photo, Hair Saloon, Fix Images
Model ID: gemini-3-pro-image-preview
A selectable AI model for image editing and style transformations. Gemini 3 Pro Image Preview supports high-fidelity image generation and editing with user-selectable output aspect ratios and resolutions, while preserving the integrity of the original photograph.
Configuration: Response Modalities: IMAGE + TEXT, user-selectable aspect ratio and image size
Safety: Content filtering enabled for harassment, hate speech, explicit content, and dangerous material
App Areas: Presets, Describe Your Own, Edit Photo, Hair Saloon, Fix Images
Model ID: fal-ai/kling-video/v2.5-turbo/pro/image-to-video
State-of-the-art image-to-video generation model that transforms static style previews into dynamic video presentations. Kling 2.5 Turbo Pro creates smooth, high-quality animations from edited images, allowing users to see their style transformations come to life with natural movement and realistic motion.
Configuration: CFG Scale 0.5, Duration options: 5s or 10s, Aspect ratios: 16:9, 9:16, 1:1
Quality Control: Negative prompting to prevent blur, distortion, and low-quality artifacts
App Area: Make Video feature for creating dynamic style presentations
Model: Claude 4.5 Sonnet by Anthropic
The primary AI agent that designed and built Style Hem's entire codebase. Claude 4.5 Sonnet excels at full-stack software engineering, architectural design, and implementing complex user interfaces with accessibility and performance in mind.
Responsibilities: Application architecture, React components, TypeScript backend, database schema, UI/UX implementation
Capabilities: Multi-file editing, parallel tool execution, architectural review, debugging assistance
Model: Claude Opus 4.1 by Anthropic
A specialized architectural review agent that provides deep code analysis, strategic guidance, and quality assurance. The Architect agent reviews all significant code changes, ensures alignment with project goals, and provides expert recommendations for system improvements.
Responsibilities: Task evaluation, debugging assistance, architectural planning, code quality review
Capabilities: Full codebase context, git diff analysis, strategic decision-making
Style Hem represents a unique collaboration between human vision and AI capabilities. The application was architected and coded entirely by a team of specialized AI agents working in concert with human oversight and direction.
Lead Developer: Claude 4.5 Sonnet (Replit Agent)
Designed and implemented the complete application stack including frontend (React + TypeScript), backend (Express + Node.js), database (PostgreSQL with Drizzle ORM), authentication (Replit Auth), and all six generation modes. Responsible for creating the seamless design system with HubSpot-style flyout menus, glassmorphism effects, and responsive mobile/desktop layouts.
Architect: Claude Opus 4.1
Provides architectural oversight, code reviews, and strategic guidance. Evaluates all major changes for code quality, security vulnerabilities, and alignment with project objectives. Ensures maintainable, scalable architecture throughout the development process.
Human Oversight: Alexander Flø Bjerkvik (Founder & Owner)
Provides product vision, feature requirements, design direction, and final approval. Ensures alignment with business objectives and user needs while guiding AI agents toward optimal implementation strategies.
This approach exemplifies our belief in balanced AI usage. Rather than using AI to generate low-quality content at scale, we leverage advanced AI models for their true strengths: complex problem-solving, architectural thinking, and sophisticated code generation. Human direction ensures strategic alignment, while AI execution provides technical precision and scalability.
Every line of code, every design decision, and every feature implementation represents a synthesis of human creativity and AI capability—demonstrating that the future of technology lies not in replacement, but in thoughtful collaboration.
User-uploaded images are processed through OpenAI, Google Gemini, and Fal.ai services exclusively for the purpose of generating requested style transformations, image fixes, and videos. Images are transmitted securely via HTTPS and are subject to the privacy policies of these third-party AI providers.
Important: We do not use your images to train our own AI models. However, third-party AI service providers (OpenAI, Google, Fal.ai) may use data according to their respective policies. We recommend reviewing their privacy policies for detailed information on data retention and usage practices.
Generated images and videos are stored in your personal gallery within our secure database. You maintain full control and can delete any content at any time through your profile settings. Deleted content is permanently removed from our systems within 30 days.
We welcome questions, feedback, and suggestions regarding our AI policies and practices. Diamond Arm is committed to continuous improvement in ethical AI deployment.
Company: Diamond Arm
Organization Number: 933575888MVA (Norway)
Owner: Alexander Flø Bjerkvik
Contact: alex@diamondarm.com