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.
主なメリット
- 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
主なメリット
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
仕組み
画像をアップロード
JPG・PNG・WebPをアップロード。最大25 MB。縦15,000 pxごとに1クレジット。
AIが分離・再構築
見出し・本文・価格・表・リストの階層を理解し、実際のテキストとして復元します。
HTMLを検証して公開
デスクトップ・タブレット・モバイルで自然にリフローするセマンティックなHTMLとCSSを生成します。