Ocr4AI: Turn images and PDFs into text for AI, with an OCR API

Hi Product Hunt, I'm Jang. I'm hunting Ocr4AI, I'm not the maker. It is built by Nam Hoàng Nguyên, who shared it on Unikorn (https://unikorn.vn/u/namhnz, product page https://unikorn.vn/p/ocr4ai), the Vietnamese community for new products. I'm hunting it because it targets one clear job: getting text out of images and PDFs cheaply before an AI model reads it.

What the product says it does:

• Images and PDFs in, Markdown or plain text out, in the browser or via an API

• OpenAI-compatible API: point the OpenAI SDK at the base URL, use model ocr4ai-exact or ocr4ai-fast, and set ocr_format to text, markdown or bounding_box

• OCR in 80+ languages

• Asynchronous jobs, page streaming for PDFs and a Playground, per the API docs

• Limits listed in the docs: PDFs up to 30 MiB and 100 pages, images up to 10 MiB

Pricing, from the site: not free. Accurate mode is $0.08 per million input tokens and $0.08 per million output tokens, Fast mode is $0.02 and $0.02, and cached input is $0.002. New accounts get $0.10 of welcome credit once, and accounts that have never topped up are limited to 100 accepted images per day.

Good to know before you try it: the maker is open about the trade-off. The home page says to prioritize speed and low cost over perfect accuracy, because an AI can still understand the content when a few accents or characters are misrecognized. The Unikorn post lists $0.1 per million tokens, while the site now shows the two modes above, so I used the site's numbers. I haven't benchmarked the accuracy.

Maker, if you see this: you can claim this page from the product page on Product Hunt (More, then Claim this page; Product Hunt checks an email address on your product's domain), and I'm glad to help however I can. What documents would you run through it first?