What changed, and when.

1.0.0 — first release

17 August 2026

• Right-click, hover badge, screen region, drop, paste and whole-page batch as ways to read an image.
• Structured ten-dimension reading with a negative list, in fifteen languages.
• Nine output formats: plain English, Midjourney, Stable Diffusion, Flux, DALL·E, Ideogram, Gemini, video and JSON.
• Colour palette and aspect ratio measured in the browser rather than asked of the model.
• Eight one-tap variations and a loose-to-verbatim faithfulness dial, both rendered locally with no second API call.
• Compare two readings: blend subject with style, and diff them field by field.
• Providers: OpenAI, Anthropic, Google, OpenRouter, Ollama, LM Studio, and any OpenAI-compatible endpoint.
• EXIF stripped from every image before it is sent, proven by a test against a real photograph.
• Per-origin site permissions, requested only when needed, with a screenshot fallback when declined.
• Searchable local library with favourites, and JSON, CSV and Markdown export.
• Free tier: 10 images a day. Pro: $15 once, 3 browsers, lifetime updates.