Call a pipeline, not a model
Feed an image to a general multimodal model and you will get something HTML-shaped. You will still have to build tall-image slicing, asset cropping, text and number validation, render comparison, retries and refunds yourself. The img2htm API ships that whole production system behind one endpoint.
POST /api/v1/reconstruct
Authorization: Bearer i2h_...
image_file=@page.jpg&mode=full&format=jsonWhat the response contains
html,css— a publishable documentassets— separated image URLscontent— sections and texts as structured JSONquality— accuracy scores and suggested review areas
See the API documentation for async mode, exponential backoff and rate limits.
核心优势
- REST API, one call
- JSON, HTML and ZIP output formats
- Async jobs with status polling
- Automatic credit refund on failure
- Samples for cURL, Node, Python, PHP and more
- Per-key rate limits and usage
核心优势
REST API, one call
JSON, HTML and ZIP output formats
Async jobs with status polling
Automatic credit refund on failure
Samples for cURL, Node, Python, PHP and more
Per-key rate limits and usage
1endpoint to start
8code samples
50/mofree preview calls
工作原理
上传图片
上传 JPG、PNG 或 WebP，最大 25 MB。每 15,000 纵向像素消耗 1 点。
AI 分离与重构
理解标题、正文、价格、表格和列表的层级，并还原为真实文本。
HTML 校验并发布
生成在桌面、平板和手机上自然重排的语义化 HTML 与 CSS。