I really like that it maintains the layout. For the tests i've done, i like the API and its fast enough Review collected by and hosted on G2.com.

I would like more control for the page markers and a way to filter out some of the artifacts for text in random places. Review collected by and hosted on G2.com.

What I dislike is minimal — LLM Whisperer is already the cherry on top of my workflow. The only thing I miss is the ability to recognize and interpret color coding inside tables, especially when extracting complex PDF tables. It’s not a dealbreaker at all, but having that feature would make the tool even more powerful for data-heavy use cases. Review collected by and hosted on G2.com.

What I dislike is minimal — LLM Whisperer is already the cherry on top of my workflow. The only thing I miss is the ability to recognize and interpret color coding inside tables, especially when extracting complex PDF tables. It’s not a dealbreaker at all, but having that feature would make the tool even more powerful for data-heavy use cases. Review collected by and hosted on G2.com.

I am using the OCR on archaeological and numismatic data, and so far, it has been the most reliable tool to extract data from those books. I also test it on other old book formats to extract data, and the preservation of Layout has been a real lifesaver in working with those old scans, since here the missing column and row lines are represented simply by tab stops or multiple whitespaces. Therefore, the layout preservation is an important tool in digitising those older books. Review collected by and hosted on G2.com.

So far, I have not encountered problems with the software itself, but the online playground takes a bit to calculate the pages (if they have a higher resolution). I would assume with an API it would take less time, but for me, the playground is more convenient. Also, sometimes it fails to recognise that act as placeholders and autofills in columns, but that is probably due to the old font type (from the 80s). Review collected by and hosted on G2.com.

I believe the most important features of LLM Whisperer would be its precise text and character recognition ability, along with the ordered and coherent way it presents the information it has managed to retrieve. Review collected by and hosted on G2.com.

A drawback of LLM Whisperer is its page limit, as it currently supports processing documents of only up to 100 pages, which can be restrictive for larger files. Review collected by and hosted on G2.com.

What I like best and find most helpful about LLMWhisperer is how it is able to extract text in a manner that remains faithful to how the document is actually structured and meant to be read. Instead of flattening PDFs into unreliable text streams, it retains layout, spatial relationships and ordering, which makes a huge difference when dealing with complex documents such as multi-column reports, tables and regulatory PDFs. This accuracy on the structural level means that there are less errors downstream, numbers or headers are not misplaced and checking and auditing is much more reliable. The greatest benefit is that it gives you a reliable base for any system in which correctness is more important than speed, particularly in technical or compliance-heavy workflows. Review collected by and hosted on G2.com.

What I do not like is that although the quality of the extraction is excellent, the output still needs a lot of downstream logic to determine the meaning and validate results for domain-specific use cases. It is not an end-to-end solution and additional effort is required by the teams in post-processing, validation, and orchestration to make the data safe to operate. For large scale or highly heterogeneous document sets, cost predictability and handling edge cases can also mean additional monitoring and tuning. Overall it's a powerful and focused tool but one that works best for technically capable teams that are willing to do the work of building around it rather than expecting a fully finished solution out of the box. Review collected by and hosted on G2.com.

The most unique feature of the Unstract LLM Whisperer extract API is its approach to text output. Unlike many other extraction tools that force everything into a Markdown format Whisperer focuses heavily on layout preservation.

In my experience, the layout preserving mode is incredibly effective. It maintains the visual relationship between elements like tables, columns, and nested sections. When feeding this output into an LLM, the reasoning accuracy is significantly higher because the model can actually "see" the original structure of the document through the text alignment. Review collected by and hosted on G2.com.

The main trade-off for this high-fidelity output is the file size. Because the API preserves the layout and spacing so precisely, the resulting text files are naturally larger than a compressed Markdown version. If you are processing massive volumes and are extremely cost-sensitive about storage or token windows, this is something to account for, but in my opinion, the accuracy gains far outweigh the extra bytes.

Note: You can also post process the text output and only send to LLM what is important for you Review collected by and hosted on G2.com.

LLM Whisperer extracts data from our PDF documents quickly and accurately. Proper handling of checkboxes is of upmost importance to us - LLM Whisperer nails it every time! I also appreciate the generous daily extraction limit of 100 documents available in their free version. We started using it daily in our healthcare business. I am looking forward to integrating LLM Whisperer into our automated workflow using the provided API. Review collected by and hosted on G2.com.

Free version does not seem to provide an easy way to copy the extracted clean text for further processing, hopefully the API integration will make it easier for us in the future. Review collected by and hosted on G2.com.

Consistent quality and very fast processing times. It has ability to process as many files in parallel as you need Review collected by and hosted on G2.com.

Nothing so far. It's a good experience but I don't quite understand the intent of the Unstruct Cloud platform. LLMWhisperer as a standalone product is great Review collected by and hosted on G2.com.

For me personally, the easy integration into my existing systems is one of the most important points and is therefore very user-friendly (from a developer's perspective). The speed of processing is unbeatable while maintaining unbeatable quality. I can control in the request which information and in what format I want to receive it back. The costs are also unbeatable in this segment. I needed the support once and it was prompt and very helpful. The use is daily. Review collected by and hosted on G2.com.

There is hardly anything I don't like, but I would wish for servers in Europe and data processing agreements to be concluded according to EU law. Review collected by and hosted on G2.com.

This tool performs exceptionally well and is tailored specifically for LLM inputs. Before settling on this one, I experimented with several other OCR solutions, but this proved to be the most effective. The space-aware layout feature is particularly useful. For the most part, it simply works—the document outputs generally match expectations. Errors such as mismatched lines or incorrectly parsed text are rare. Also, the fact that it returns bounding boxes for each line makes it very easy to show citations for ai-generated results, which is helpful for verification. Review collected by and hosted on G2.com.

1-2 times a week, the service is down when making a request. This is a problem, especially since I'm trying to develop products for skeptical users. Also, my application requires parsing tables that sometimes have text (handwritten or typed) that spans multiple columns of the table, and the response sometimes breaks that text up in weird ways. I imagine it is difficult to confidently parse this kind of writing into a clean text document, but better performance here would be nice. Review collected by and hosted on G2.com.