Your images stay on your local computer. This app uses the File System Access API to work with your files directly in the browser. Nothing is uploaded to our servers unless you explicitly use the AI captioning feature.

When you use AI captioning, your images are temporarily sent to AI providers (like OpenAI, Google, or Alibaba) to generate descriptions. These are not stored by the providers. You need credits to use this feature, which you can purchase in the app.

The app supports PNG, JPG, JPEG, WebP, and GIF images. Captions are saved as .txt files with the same name as the image, and tags are saved as .tag files.

Yes! The tool itself is free to use for any project. You only pay for AI captioning credits if you choose to use that feature. The generated captions are yours to use however you like.

Sign-in is only required for AI captioning to track your credits. You can use all other features (organizing, uploading, manual captioning) without signing in.

Dataset Manager Studio requires a browser that supports the File System Access API. This includes Chrome 86+, Edge 86+, Opera, and other Chromium-based browsers. Firefox and Safari do not currently support this API. You can still visit the site with an unsupported browser, but the file management features won't work.

We offer multiple AI vision models from providers like OpenAI, Google, and Alibaba. Each model has different strengths and pricing. We also provide three LoRA templates: Character (for training specific characters), Style (for art styles), and Photography (for realistic images).

AI-generated captions and tags are not 100% accurate and can make mistakes. While AI models are quite good at describing images, they may occasionally misidentify objects, get details wrong, or describe things inaccurately. Always review and edit the generated captions before using them for training. Think of AI captioning as a helpful starting point that saves you time, not a perfect final result.

Models vary in size, capability, and cost. Smaller models (like Gemini Flash or Qwen) are cheaper and faster but may be less accurate or miss details. Larger models (like GPT-4o or Gemini Pro) cost more but generally provide better, more detailed captions. Premium models are best for complex images or when accuracy is critical. Check the AI Settings to compare models and their prices.

Yes, but with important caveats. Some AI providers block or restrict NSFW content, while others are more permissive. In the AI Settings, we separate models into "General" and "NSFW-Friendly" categories so you can choose the right model for your content. NSFW-friendly models are less likely to refuse your images, but availability may vary. Please use responsibly and ensure your content complies with our Terms of Service.

No! You only pay for successful caption generations. If the AI fails, times out, or returns an error, you won't be charged any credits. This applies to both single image captioning and batch processing.

We use a fair-share concurrency system. When a single user is using the service, they get full capacity. When multiple users are active, capacity is shared equally among them. Models may show as "High Load" when many users are active. If a provider is offline or at capacity, we'll automatically try other available models or queue your request. Check the AI Settings modal to see real-time model status and availability.