diff --git a/content/docs/en/copy-to-word.md b/content/docs/en/copy-to-word.md new file mode 100644 index 0000000..2dab050 --- /dev/null +++ b/content/docs/en/copy-to-word.md @@ -0,0 +1,66 @@ +--- +title: Copy to Word +description: Export recognized formulas directly into Microsoft Word as editable equations +slug: copy-to-word +date: 2026-03-25 +tags: [export, Word, DOCX] +order: 4 +--- + +# Copy to Word + +TexPixel can export your recognized formulas directly into Microsoft Word as native, editable equations — not images. This means you can continue editing the formula inside Word after export. + +## How to Export to Word + +1. Upload your formula image and wait for recognition to complete. +2. Click the **Export** button in the result panel. +3. Select **DOCX** from the file export options. +4. Download the file and open it in Microsoft Word. + +The downloaded `.docx` file contains your formula as a native Word equation (OMML format), which Word renders using its built-in equation editor. + +## Why Use DOCX Export? + +| Method | Editable in Word | Renders Correctly | Copy-Paste | +|---|---|---|---| +| Screenshot / image | No | Yes | No | +| LaTeX string | No (without plugin) | No | Yes | +| DOCX export | **Yes** | **Yes** | N/A | + +The DOCX format is ideal when you need to: +- Submit homework or reports as Word documents +- Share formulas with colleagues who don't use LaTeX +- Continue editing the formula after export + +## Inserting into an Existing Document + +If you want to insert a formula into an existing Word document rather than starting fresh: + +1. Open the downloaded `.docx` file in Word. +2. Select the equation and copy it (`Ctrl+C` / `Cmd+C`). +3. Paste it into your target document (`Ctrl+V` / `Cmd+V`). + +Word preserves the equation formatting during paste. + +## Mixed Content (Text + Formulas) + +If your upload contains a mix of regular text and formulas (e.g., a textbook page), use DOCX export — it's the only format that handles mixed content correctly. LaTeX and MathML export are only available for pure-formula results. + +> **Note:** For mixed-content results, LaTeX/MathML export is disabled. Use DOCX to get a properly formatted document with both text and equations. + +## Compatibility + +DOCX export is compatible with: +- Microsoft Word 2016 and later (Windows and Mac) +- Google Docs (equations render as images when imported) +- LibreOffice Writer (partial support) + +## Tips + +- After pasting into Word, double-click the equation to open the equation editor and make changes. +- If the formula looks different from expected, try re-uploading a higher-resolution image for a more accurate recognition result. + +--- + +[Try exporting a formula to Word →](/app) diff --git a/content/docs/en/image-to-latex.md b/content/docs/en/image-to-latex.md new file mode 100644 index 0000000..182c247 --- /dev/null +++ b/content/docs/en/image-to-latex.md @@ -0,0 +1,80 @@ +--- +title: Image to LaTeX +description: How to convert any formula image into clean LaTeX code with TexPixel +slug: image-to-latex +date: 2026-03-25 +tags: [LaTeX, tutorial] +order: 2 +--- + +# Image to LaTeX + +TexPixel's core feature is converting formula images — from photos, scans, or screenshots — directly into LaTeX code you can paste anywhere. + +## How It Works + +1. **Upload your image** — Drag and drop a JPG or PNG into the upload zone, or click to browse. You can also paste from your clipboard. +2. **AI processes it** — Our model detects the formula region, runs OCR, and generates structured LaTeX in under a second. +3. **Copy the result** — Click the copy button next to the LaTeX output. Paste directly into Overleaf, VS Code, Word, or any LaTeX editor. + +## Input Requirements + +| Requirement | Details | +|---|---| +| File formats | JPG, PNG | +| Max file size | 10 MB | +| Recommended DPI | 150 DPI or higher | +| Background | White or light backgrounds work best | + +## What Gets Recognized + +TexPixel handles a wide range of mathematical content: + +- **Algebra** — equations, inequalities, polynomials +- **Calculus** — derivatives, integrals, limits +- **Matrices** — 2×2 up to large arrays +- **Greek letters** — α, β, γ, Σ, Π, and more +- **Subscripts and superscripts** — `x_i^2`, `a_{n+1}` +- **Fractions** — `\frac{a}{b}`, nested fractions +- **Square roots and radicals** — `\sqrt{x}`, `\sqrt[n]{x}` + +## Example + +Uploading an image of the quadratic formula gives you: + +```latex +x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a} +``` + +An image of an integral: + +```latex +\int_0^\infty e^{-x^2}\, dx = \frac{\sqrt{\pi}}{2} +``` + +## Tips for Best Results + +- **Use clear images** — avoid blur, shadows, or low contrast +- **Crop tightly** — the less background, the better the focus +- **Dark ink on white paper** — ideal for handwritten formulas +- **Avoid rotated images** — keep the formula horizontal +- **One formula per image** — for complex multi-part work, crop each formula separately + +## Limitations + +- Extremely faint or pencil-written formulas may have lower accuracy +- Hand-drawn arrows or annotation marks outside the formula may be ignored +- Very large matrices (10×10+) may have reduced accuracy + +## Copy Options + +After recognition, you can copy output in multiple formats: + +- **LaTeX** — raw LaTeX string +- **MathML** — for web embedding +- **Markdown** — `$...$` inline or `$$...$$` block +- **Plain text** — Unicode approximation + +--- + +Ready to try it? [Upload a formula image now →](/app) diff --git a/content/docs/en/ocr-accuracy.md b/content/docs/en/ocr-accuracy.md new file mode 100644 index 0000000..19bcdc5 --- /dev/null +++ b/content/docs/en/ocr-accuracy.md @@ -0,0 +1,79 @@ +--- +title: OCR Accuracy +description: Understanding TexPixel recognition accuracy and how to get the best results +slug: ocr-accuracy +date: 2026-03-25 +tags: [accuracy, tips] +order: 5 +--- + +# OCR Accuracy + +TexPixel achieves industry-leading accuracy on mathematical formula recognition — but accuracy isn't uniform across all input types. This guide explains what affects accuracy and how to maximize it. + +## Accuracy by Formula Type + +| Formula Type | Typical Accuracy | +|---|---| +| Printed formulas (textbooks, papers) | 95–99% | +| Clean handwritten formulas | 88–95% | +| Scanned documents (300 DPI+) | 93–98% | +| Photos of whiteboards | 82–92% | +| Low-resolution images (< 72 DPI) | 60–80% | + +These are approximate ranges. Individual results depend heavily on image quality. + +## Factors That Affect Accuracy + +### Image Quality + +The single biggest factor. A blurry, low-resolution, or poorly lit image will always produce worse results than a clean scan. + +- **Resolution** — 150 DPI or higher is recommended. 300 DPI is ideal for documents. +- **Contrast** — dark ink on a white background gives the clearest signal to the model. +- **Sharpness** — avoid motion blur or out-of-focus shots. + +### Formula Complexity + +Simple single-line equations are recognized with near-perfect accuracy. More complex structures may have occasional errors: + +- Multi-line equation systems +- Large matrices (6×6 or larger) +- Heavily nested fractions (3+ levels deep) +- Non-standard notation or custom symbols + +### Handwriting Style + +Printed (typed) formulas outperform handwritten ones, but TexPixel handles handwriting well when: + +- Letters are clearly formed and not connected (print style, not cursive) +- Variables are written in distinct sizes (clearly different x and × for example) +- Spacing between symbols is consistent + +### What Reduces Accuracy + +- **Rotated images** — formulas at an angle are harder to parse +- **Overlapping elements** — crossed-out work, annotations, or arrows near symbols +- **Pencil on paper** — low contrast; try increasing image brightness/contrast before uploading +- **Multiple formulas in one image** — crop to the specific formula you need +- **Decorative fonts** — calligraphic or stylized mathematical writing + +## Improving Results + +If you're getting errors, try these steps in order: + +1. **Increase image resolution** — scan at 300 DPI instead of 150 DPI +2. **Improve contrast** — use a photo editor to increase brightness and contrast +3. **Crop tightly** — remove surrounding text and whitespace +4. **Straighten the image** — correct rotation before uploading +5. **Re-photograph** — better lighting, closer distance, sharper focus + +## Reporting Errors + +Found a formula type that TexPixel consistently gets wrong? Let us know — accuracy feedback directly improves the model over time. + +Contact us at: [support@texpixel.com](mailto:support@texpixel.com) + +--- + +[Upload a formula and test accuracy →](/app) diff --git a/content/docs/en/pdf-extraction.md b/content/docs/en/pdf-extraction.md new file mode 100644 index 0000000..bfb02e8 --- /dev/null +++ b/content/docs/en/pdf-extraction.md @@ -0,0 +1,75 @@ +--- +title: PDF Extraction +description: Extract and convert formulas from PDF documents automatically with TexPixel +slug: pdf-extraction +date: 2026-03-25 +tags: [PDF, extraction] +order: 6 +--- + +# PDF Extraction + +TexPixel can process entire PDF documents and extract every formula from every page — automatically. This is useful for textbooks, research papers, or any multi-page document with mathematical content. + +## How to Extract from a PDF + +1. Click the upload zone or drag and drop your PDF file. +2. TexPixel detects all pages and identifies formula regions. +3. Each recognized formula is listed in the result panel. +4. Copy individual formulas or export the entire document as DOCX. + +## What Gets Extracted + +TexPixel identifies formulas in PDFs regardless of whether they were: +- Typeset in LaTeX (rendered as vector math) +- Embedded as images (scanned pages) +- A mix of both + +For vector PDFs (generated from LaTeX or Word), recognition accuracy is typically 95%+. For scanned/image PDFs, accuracy follows the same image quality guidelines as regular image uploads. + +## Supported PDF Types + +| Type | Description | Accuracy | +|---|---|---| +| Vector PDF | Created from LaTeX, Word, or typesetting tools | 95–99% | +| Scanned PDF (high quality) | 300 DPI scan of printed text | 90–97% | +| Scanned PDF (low quality) | < 150 DPI or poor contrast | 60–80% | +| Photo PDF | Photographed pages embedded as images | 75–90% | + +## File Limits + +- **Max file size:** 20 MB +- **Max pages:** 50 pages per upload (Pro plan: unlimited) +- **Processing time:** ~2–5 seconds per page + +For documents exceeding these limits, split the PDF into smaller chunks before uploading. + +## Exporting PDF Results + +After extraction, you can export in several ways: + +- **Copy individual formula** — click any recognized formula to copy its LaTeX +- **DOCX export** — download the full document with formulas as native Word equations +- **Batch copy** — copy all formulas as a list (Pro feature) + +## Tips for Better PDF Results + +- **Use the original PDF**, not a re-scanned copy — vector PDFs give the best results +- **Avoid password-protected PDFs** — these cannot be processed +- **Crop pages** if a PDF has wide margins with no content — smaller pages process faster +- **Split by chapter** for very large documents to stay within page limits + +## Common Issues + +**"No formulas found"** +The PDF may be encrypted, have formulas stored as complex vector paths, or use non-standard encoding. Try converting the page to a PNG image and uploading that instead. + +**Formulas recognized but garbled** +This often happens with very low DPI scans. Try using a PDF scanner app to rescan at 300 DPI before uploading. + +**Processing is slow** +Large PDFs with many pages can take 30–60 seconds. This is normal. The result will appear when processing is complete. + +--- + +[Upload a PDF and extract formulas →](/app) diff --git a/content/docs/zh/copy-to-word.md b/content/docs/zh/copy-to-word.md new file mode 100644 index 0000000..19d744e --- /dev/null +++ b/content/docs/zh/copy-to-word.md @@ -0,0 +1,66 @@ +--- +title: 导出到 Word +description: 将识别的公式直接导出到 Microsoft Word 中作为可编辑方程 +slug: copy-to-word +date: 2026-03-25 +tags: [导出, Word, DOCX] +order: 4 +--- + +# 导出到 Word + +TexPixel 可以将识别的公式直接导出到 Microsoft Word 中作为原生可编辑方程——而不是图片。这意味着导出后你可以在 Word 中继续编辑公式。 + +## 如何导出到 Word + +1. 上传公式图片并等待识别完成。 +2. 点击结果面板中的**导出**按钮。 +3. 从文件导出选项中选择 **DOCX**。 +4. 下载文件并在 Microsoft Word 中打开。 + +下载的 `.docx` 文件包含以原生 Word 方程(OMML 格式)表示的公式,Word 使用内置方程编辑器渲染。 + +## 为什么使用 DOCX 导出? + +| 方式 | Word 中可编辑 | 正确渲染 | 复制粘贴 | +|---|---|---|---| +| 截图/图片 | 否 | 是 | 否 | +| LaTeX 字符串 | 否(无插件) | 否 | 是 | +| DOCX 导出 | **是** | **是** | N/A | + +DOCX 格式非常适合以下情况: +- 提交 Word 格式的作业或报告 +- 与不使用 LaTeX 的同事共享公式 +- 导出后继续编辑公式 + +## 插入到现有文档 + +如果你想将公式插入现有 Word 文档而不是新建文档: + +1. 在 Word 中打开下载的 `.docx` 文件。 +2. 选中方程并复制(`Ctrl+C` / `Cmd+C`)。 +3. 粘贴到目标文档(`Ctrl+V` / `Cmd+V`)。 + +Word 在粘贴时保留方程格式。 + +## 混合内容(文字 + 公式) + +如果上传内容包含普通文字和公式的混合(例如教材页面),请使用 DOCX 导出——这是唯一能正确处理混合内容的格式。LaTeX 和 MathML 导出仅适用于纯公式结果。 + +> **注意:** 对于混合内容结果,LaTeX/MathML 导出不可用。请使用 DOCX 获取包含文字和方程的格式正确文档。 + +## 兼容性 + +DOCX 导出与以下软件兼容: +- Microsoft Word 2016 及更高版本(Windows 和 Mac) +- Google 文档(导入时方程渲染为图片) +- LibreOffice Writer(部分支持) + +## 提示 + +- 粘贴到 Word 后,双击方程打开方程编辑器进行修改。 +- 如果公式与预期不同,请尝试上传更高分辨率的图片以获得更准确的识别结果。 + +--- + +[尝试将公式导出到 Word →](/app) diff --git a/content/docs/zh/image-to-latex.md b/content/docs/zh/image-to-latex.md new file mode 100644 index 0000000..d64f73a --- /dev/null +++ b/content/docs/zh/image-to-latex.md @@ -0,0 +1,80 @@ +--- +title: 图片转 LaTeX +description: 如何使用 TexPixel 将任意公式图片转换为干净的 LaTeX 代码 +slug: image-to-latex +date: 2026-03-25 +tags: [LaTeX, 教程] +order: 2 +--- + +# 图片转 LaTeX + +TexPixel 的核心功能是将公式图片——来自照片、扫描件或截图——直接转换为可以粘贴到任何地方的 LaTeX 代码。 + +## 使用方法 + +1. **上传图片** — 将 JPG 或 PNG 拖拽到上传区域,或点击浏览文件。也可以直接从剪贴板粘贴。 +2. **AI 处理** — 模型检测公式区域,运行 OCR,在不到一秒内生成结构化 LaTeX。 +3. **复制结果** — 点击 LaTeX 输出旁的复制按钮,直接粘贴到 Overleaf、VS Code、Word 或任意 LaTeX 编辑器。 + +## 输入要求 + +| 要求 | 详情 | +|---|---| +| 文件格式 | JPG、PNG | +| 最大文件大小 | 10 MB | +| 推荐分辨率 | 150 DPI 或更高 | +| 背景 | 白色或浅色背景效果最佳 | + +## 支持识别的内容 + +TexPixel 可处理多种数学内容: + +- **代数** — 方程、不等式、多项式 +- **微积分** — 导数、积分、极限 +- **矩阵** — 2×2 到大型数组 +- **希腊字母** — α、β、γ、Σ、Π 等 +- **上下标** — `x_i^2`、`a_{n+1}` +- **分数** — `\frac{a}{b}`、嵌套分数 +- **根号** — `\sqrt{x}`、`\sqrt[n]{x}` + +## 示例 + +上传二次公式图片,输出: + +```latex +x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a} +``` + +上传积分图片: + +```latex +\int_0^\infty e^{-x^2}\, dx = \frac{\sqrt{\pi}}{2} +``` + +## 获得最佳结果的技巧 + +- **使用清晰图片** — 避免模糊、阴影或低对比度 +- **紧密裁剪** — 背景越少,焦点越准确 +- **白纸深色墨水** — 手写公式的理想条件 +- **避免旋转图片** — 保持公式水平 +- **每张图片一个公式** — 对于复杂的多部分作业,分别裁剪每个公式 + +## 局限性 + +- 非常淡或铅笔书写的公式准确率可能较低 +- 公式外的手绘箭头或注释标记可能被忽略 +- 非常大的矩阵(10×10 以上)可能准确率降低 + +## 复制选项 + +识别完成后,可以多种格式复制输出: + +- **LaTeX** — 原始 LaTeX 字符串 +- **MathML** — 用于网页嵌入 +- **Markdown** — 行内 `$...$` 或块级 `$$...$$` +- **纯文本** — Unicode 近似表示 + +--- + +准备好了吗?[立即上传公式图片 →](/app) diff --git a/content/docs/zh/ocr-accuracy.md b/content/docs/zh/ocr-accuracy.md new file mode 100644 index 0000000..f9e1d3e --- /dev/null +++ b/content/docs/zh/ocr-accuracy.md @@ -0,0 +1,79 @@ +--- +title: 识别准确率 +description: 了解 TexPixel 识别准确率及如何获得最佳效果 +slug: ocr-accuracy +date: 2026-03-25 +tags: [准确率, 技巧] +order: 5 +--- + +# 识别准确率 + +TexPixel 在数学公式识别方面达到行业领先的准确率——但准确率在不同输入类型之间并不统一。本指南解释影响准确率的因素以及如何最大化识别效果。 + +## 按公式类型的准确率 + +| 公式类型 | 典型准确率 | +|---|---| +| 印刷体公式(教材、论文) | 95–99% | +| 清晰手写公式 | 88–95% | +| 扫描文档(300 DPI+) | 93–98% | +| 白板照片 | 82–92% | +| 低分辨率图片(< 72 DPI) | 60–80% | + +这些是大致范围,实际结果在很大程度上取决于图片质量。 + +## 影响准确率的因素 + +### 图片质量 + +这是最重要的单一因素。模糊、低分辨率或光线不佳的图片效果始终不如清晰扫描件。 + +- **分辨率** — 建议 150 DPI 或更高,文档理想为 300 DPI +- **对比度** — 白色背景上的深色墨水为模型提供最清晰的信号 +- **清晰度** — 避免运动模糊或对焦不准 + +### 公式复杂度 + +简单的单行方程识别准确率接近完美。更复杂的结构可能偶有错误: + +- 多行方程组 +- 大矩阵(6×6 或更大) +- 深度嵌套分数(3 层以上) +- 非标准符号或自定义符号 + +### 手写风格 + +印刷体(打字)公式优于手写体,但当以下条件满足时,TexPixel 能很好地处理手写: + +- 字母清晰成形且不连笔(印刷体,而非草书) +- 变量写成明显不同的大小(例如 x 和 × 清晰区分) +- 符号间距一致 + +### 降低准确率的因素 + +- **旋转图片** — 倾斜的公式更难解析 +- **重叠元素** — 划掉的内容、注释或符号附近的箭头 +- **纸上铅笔** — 对比度低;上传前可尝试增加图片亮度/对比度 +- **一张图片多个公式** — 裁剪到你需要的具体公式 +- **装饰字体** — 花体或风格化数学书写 + +## 提高识别效果 + +如果识别出错,按以下顺序尝试: + +1. **提高图片分辨率** — 用 300 DPI 扫描代替 150 DPI +2. **改善对比度** — 使用图片编辑器提高亮度和对比度 +3. **紧密裁剪** — 去除周围文字和空白 +4. **矫正图片** — 上传前纠正旋转 +5. **重新拍摄** — 更好的光线、更近的距离、更清晰的对焦 + +## 反馈错误 + +发现 TexPixel 持续识别错误的公式类型?请告知我们——准确率反馈直接改进模型。 + +联系我们:[support@texpixel.com](mailto:support@texpixel.com) + +--- + +[上传公式测试识别准确率 →](/app) diff --git a/content/docs/zh/pdf-extraction.md b/content/docs/zh/pdf-extraction.md new file mode 100644 index 0000000..be140c8 --- /dev/null +++ b/content/docs/zh/pdf-extraction.md @@ -0,0 +1,75 @@ +--- +title: PDF 公式提取 +description: 使用 TexPixel 自动从 PDF 文档中提取并转换公式 +slug: pdf-extraction +date: 2026-03-25 +tags: [PDF, 提取] +order: 6 +--- + +# PDF 公式提取 + +TexPixel 可以处理完整的 PDF 文档,自动从每一页提取所有公式。这对教材、研究论文或任何包含数学内容的多页文档非常有用。 + +## 如何从 PDF 提取 + +1. 点击上传区域或将 PDF 文件拖拽到其中。 +2. TexPixel 检测所有页面并识别公式区域。 +3. 每个识别的公式列在结果面板中。 +4. 复制单个公式或将整个文档导出为 DOCX。 + +## 提取内容 + +无论 PDF 中的公式是如何生成的,TexPixel 都能识别: +- 用 LaTeX 排版(渲染为矢量数学) +- 嵌入为图片(扫描页面) +- 两种混合 + +对于矢量 PDF(由 LaTeX 或 Word 生成),识别准确率通常为 95% 以上。对于扫描/图片 PDF,准确率遵循与普通图片上传相同的图片质量准则。 + +## 支持的 PDF 类型 + +| 类型 | 描述 | 准确率 | +|---|---|---| +| 矢量 PDF | 由 LaTeX、Word 或排版工具创建 | 95–99% | +| 扫描 PDF(高质量) | 印刷文字的 300 DPI 扫描 | 90–97% | +| 扫描 PDF(低质量) | < 150 DPI 或对比度差 | 60–80% | +| 照片 PDF | 嵌入为图片的拍照页面 | 75–90% | + +## 文件限制 + +- **最大文件大小:** 20 MB +- **最大页数:** 每次上传 50 页(专业版:无限制) +- **处理时间:** 每页约 2–5 秒 + +对于超出限制的文档,上传前将 PDF 分割成较小的部分。 + +## 导出 PDF 识别结果 + +提取后,可以多种方式导出: + +- **复制单个公式** — 点击任意识别的公式复制其 LaTeX +- **DOCX 导出** — 下载包含原生 Word 方程的完整文档 +- **批量复制** — 将所有公式复制为列表(专业版功能) + +## 提高 PDF 识别效果的技巧 + +- **使用原始 PDF**,而非重新扫描的副本——矢量 PDF 效果最佳 +- **避免密码保护的 PDF**——这类文件无法处理 +- 如果 PDF 有很宽的空白边距,**裁剪页面**——较小的页面处理更快 +- 对于非常大的文档,**按章节分割**以保持在页数限制内 + +## 常见问题 + +**"未找到公式"** +PDF 可能已加密,公式可能以复杂矢量路径存储,或使用了非标准编码。尝试将页面转换为 PNG 图片后再上传。 + +**公式已识别但内容乱码** +这通常发生在非常低 DPI 的扫描件上。尝试在上传前使用 PDF 扫描应用以 300 DPI 重新扫描。 + +**处理速度慢** +包含多页的大型 PDF 可能需要 30–60 秒。这是正常的,处理完成后结果会显示。 + +--- + +[上传 PDF 提取公式 →](/app) diff --git a/docs/superpowers/plans/2026-03-26-landing-refactor.md b/docs/superpowers/plans/2026-03-26-landing-refactor.md new file mode 100644 index 0000000..0dda0e2 --- /dev/null +++ b/docs/superpowers/plans/2026-03-26-landing-refactor.md @@ -0,0 +1,1523 @@ +# Landing Page Refactor Implementation Plan + +> **For agentic workers:** REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (`- [ ]`) syntax for tracking. + +**Goal:** Replace all marketing home components with the UI from `texpixel-landing.html`, preserving exact visual fidelity while integrating with React Router, AuthContext, and existing app infrastructure. + +**Architecture:** Extract the reference HTML's `