[HN Gopher] Mistral OCR 3
       ___________________________________________________________________
        
       Mistral OCR 3
        
       Author : pember
       Score  : 665 points
       Date   : 2025-12-18 15:01 UTC (2 days ago)
        
 (HTM) web link (mistral.ai)
 (TXT) w3m dump (mistral.ai)
        
       | pzo wrote:
       | there has been so many open source OCR in the last 3 months that
       | would be good to compare to those especially when some are not
       | even 1B params and can be run on edge devices.
       | 
       | - paddleOCR-VL
       | 
       | - olmOCR-2
       | 
       | - chandra
       | 
       | - dots.ocr
       | 
       | I kind of miss there is not many leaderboard sections or arena
       | for OCR and CV and providers hosting those. Neglected on both
       | Artificial Analysis and OpenRouter.
        
         | pzo wrote:
         | what I like in MistralOCR is that they have simple pricing
         | $1/1k pages and API hosted on their servers. With other OCR is
         | hard to compare pricing because are token based and you don't
         | know how many tokens is the image unless you run your own test.
         | 
         | E.g. with Gemini 3.0 flash you might seem that model pricing
         | increased only slightly comparing to Gemini 2.5 flash until you
         | test it and will see that what used to be 258 per 384x384 input
         | tokens now is around 3x more.
        
           | gunalx wrote:
           | But they doubled the price g for this new mistralocr3 model
           | to 2$
        
           | amelius wrote:
           | Simple would be to bill per character.
           | 
           | Now I have to figure out how large a page can be.
        
         | hereme888 wrote:
         | https://www.codesota.com/ocr
        
         | culi wrote:
         | Someone posted a project here about a month ago where they
         | compare models in head-to-head matchups similar to llmarena
         | 
         | https://www.ocrarena.ai/leaderboard
         | 
         | Hasn't been updated for Mistral but so far gemeni seems to top
         | the leaderboard.
        
           | jeffbee wrote:
           | OCR developers from decades past must be slapping their
           | foreheads now that it seems users will wait a whole minute
           | per page and be happy.
        
             | delaminator wrote:
             | What they are happy about is accurate OCR.
             | 
             | Getting the wrong answer really quickly is not the best
             | goal.
        
             | culi wrote:
             | You can also sort by latency. dots.ocr has the lowest at
             | 3.8s/page. And although it doesn't fare very well against
             | much larger slower models, it's still streets ahead of
             | traditional OCR techniques
        
           | andai wrote:
           | How can something have a very high ELO but a very low win
           | rate?
        
             | BlackLotus89 wrote:
             | You don't loose any elo if your opponent is much stronger
             | than you. Remis could in theory play a part as well.
        
           | pplonski86 wrote:
           | very nice comparison! I'd like to see on what examples OCR
           | engines fail
        
         | andai wrote:
         | I spent like three hours trying to get one of these running and
         | then gave up. I think the paddleOCR one.
         | 
         | It took an hour and a half to install 12 gigabytes of pytorch
         | dependencies that can't even run on my device, and then it told
         | me it had some sort of versioning conflict. (I think I was
         | supposed to use UV, but I had run out of steam by that point.)
         | 
         | Maybe I should have asked Claude to install it for me. I gave
         | Claude root on a $3 VPS, and it seems to enjoy the sysadmin
         | stuff a lot more than I do...
         | 
         | Incidentally I had a similar experience installing open web
         | UI... It installed 12 GB of pytorch crap.. I rage quit and
         | deleted the whole thing, and replicated the functionality I
         | actually needed in 100 lines of HTML.... Too bad I can't do
         | that with OCR ;)
        
           | CamperBob2 wrote:
           | gemini-cli is good for this sort of thing. You can just tell
           | it "Find out why xyz.py doesn't run" and let it crunch. It
           | will try reasonably hard to get you out of Python dependency
           | hell, and (more important) it generally knows when to give
           | up.
           | 
           | But yes, in general, you want to use uv. Otherwise, the next
           | Python application you install WILL break the last one you
           | installed.
           | 
           | I suppose you could use gemini-cli as a substitute for proper
           | Python virtual environment management, always letting it fix
           | whatever broke since the last time you tried to run the
           | program, but that'd be like burning down a rainforest to
           | toast a marshmallow.
        
             | andai wrote:
             | Actually, I just remembered, this was inside uv!
        
       | Tiberium wrote:
       | From a tweet: https://x.com/i/status/2001821298109120856
       | 
       | > can someone help folks at Mistral find more weak baselines to
       | add here? since they can't stomach comparing with SoTA....
       | 
       | > (in case y'all wanna fix it: Chandra, dots.ocr, olmOCR, MinerU,
       | Monkey OCR, and PaddleOCR are a good start)
        
         | belval wrote:
         | I've worked on document extraction a lot and while the tweet is
         | too flippant for my taste, it's not wrong. Mistral is comparing
         | itself to non-VLM computer vision services. While not
         | necessarily what everyone needs, they are a very different
         | beasts compared to VLM based extraction because it gives you
         | precise bounding boxes, usually at the cost of larger "document
         | understanding".
         | 
         | Its failure mode are also vastly different. VLM-based
         | extraction can misread entire sentences or miss entire
         | paragraphs. Sonnet 3 had that issue. Computer vision models
         | instead will make in-word typos.
        
           | wills_forward wrote:
           | Why not use both? I just built a pipeline for document data
           | extraction that uses PaddleOCR, then Gemini 3 to check + fix
           | errors. It gets close to 99.9% on extraction from financial
           | statements finally on par with humans.
        
             | jadbox wrote:
             | This is The Way. Remember AI doesn't have to replace
             | existing solutions but can tactfully supplement it.
        
             | vrc wrote:
             | I did the opposite. Tesseract to get bboxes, words, and
             | chars and then mistral on the clips with some reasonable
             | reflow to preserve geometry. Paddle wasn't working on my
             | local machine (until I found RapidOCR). Surya was also very
             | good but because you can't really tweak any knobs, when it
             | failed it just kinda failed. But Surya > Rapid w/ Paddle >
             | DocTr > Tesseract while the latter gave me the most
             | granularity when I needed it.
             | 
             | Edit: Gemini 2.0 was good enough for VLM cleanup, and now
             | 2.5 or above with structured output make reconstruction
             | even easier.
        
           | zerocrates wrote:
           | Is DeepSeek's not VLM?
        
         | logicprog wrote:
         | I'd want to see a comparison with Qwen 3 VL 235B-A22B, which is
         | IME significantly better than MinerU.
        
         | vinckr wrote:
         | after clicking on your link I browsed twitter for a minute and
         | damn that place has become weird (or maybe it always was?)
        
           | crystal_revenge wrote:
           | As someone who has been on Twitter since 2007, it's radically
           | changed in the last few years to the point of being
           | unrecognizable.
        
         | ozgune wrote:
         | Also, do you know if their benchmarks are available?
         | 
         | In their website, the benchmarks say "Multilingual (Chinese),
         | Multilingual (East-asian), Multilingual (Eastern europe),
         | Multilingual (English), Multilingual (Western europe), Forms,
         | Handwritten, etc." However, there's no reference to the
         | benchmark data.
        
         | nerbert wrote:
         | On the OP link, they compare themselves to the capabilities of
         | leaderboard AI's and beat them.
        
       | petcat wrote:
       | It seems like Mistral is just chasing around sort of "the
       | fringes" of what could be useful AI features. Are they just
       | getting out-classed by OAI, Google, Anthropic?
       | 
       | It seems like EU in general should be heavily invested in
       | Mistral's development, but it doesn't seem like they are.
        
         | VWWHFSfQ wrote:
         | I think there is a lot of broad support, but they're just kind
         | of hamstrung by EU regulation on AI development at this stage.
         | I think the end game will ultimately be getting acquired by an
         | American company, and then relocating.
        
           | tensor wrote:
           | I hope the EU blocks any acquisitions by American companies.
           | The west needs to start protecting its strategic assets.
        
           | bootsmann wrote:
           | Do you have any source on this other than vibes based on"EU
           | bad" sentiment?
        
         | tensor wrote:
         | Form processing is vastly more useful than meme generation.
         | When people need to do real work this is the sort of tool they
         | are going to reach for.
        
           | sbuttgereit wrote:
           | Yep. I saw the title and got excited.... this is a particular
           | problem area where I think these things can be very
           | effective. There are so many data entry class tasks which
           | don't require huge knowledge or judgement... just clear
           | parsing and putting that into a more machine digestible form.
           | 
           | I don't know... feels like this sort of area, while not
           | nearly so sexy as video production or coding or (etc.)... but
           | seems like reaching a better-than-human performance level
           | should be easier for these kinds of workloads.
        
         | lawlessone wrote:
         | >It seems like EU in general should be heavily invested
         | 
         | Maybe, i think it will be to our benefit when the bubble pops
         | that we are not heavily invested, no harm investing a little.
        
         | BoredPositron wrote:
         | I guess it's better to do the same stuff everyone else is
         | doing?
        
         | bee_rider wrote:
         | Following the leaders too closely seems like a bad move, at
         | least until a profitable business model for an AI model
         | training company is discovered. Mistral's models are pretty
         | good, right? I mean they don't have all the scaffolding around
         | them that something like chatGPT does, but building all that
         | scaffolding could be wasted effort until a profitable business
         | model is shown.
         | 
         | Until then, they seem to be able to keep enough talent in the
         | EU to train reasonably good models. The kernel is there, which
         | seems like the attainable goal.
        
           | qwytw wrote:
           | >Mistral's models are pretty good, right
           | 
           | Are they? IIRC their best model is still worse than the gpt-
           | oss-120B?
        
             | amarcheschi wrote:
             | Devstral 2 should be above
             | https://mistral.ai/news/devstral-2-vibe-cli
             | 
             | Though I haven't checked other benchmarks and they only
             | report swe
        
               | acters wrote:
               | Devstral 2 is free from the API. That has to be a bigger
               | point to what makes it better. The price to performance
               | ratio is practically better in every way. Does it matter
               | if the performance is slightly worse when it is
               | practically free?
        
               | qwytw wrote:
               | Yes, but if it's actually competitive that won't last
               | that long. Mistral will do the same as google (cut their
               | free tier by 50x or so) if they ever catch up.
               | Financially anything else would make no sense.
               | 
               | Of course currently Mistral has an insane free tier, 1
               | billion tokens for each(?) of their models per month.
        
             | tomalbrc wrote:
             | Calling it oss is a farce
        
           | menaerus wrote:
           | They can't hire the best talent because the most experienced
           | people will not leave their homes to chase a high-risk role
           | with questionable remuneration by relocating their whole life
           | to Paris or London.
           | 
           | This goes to show how leaders in Mistral don't quite get that
           | they are not special as they seem to think they are.
           | Anthropic or OpenAI also require their talent to relocate but
           | with stakes that are at least a high reward - $500k or $1M a
           | year is a good start that is maybe worth investing into.
        
             | bee_rider wrote:
             | If somebody is in the EU already that calculation
             | completely flips. We have a strong software startup
             | industry in the US, would it really be that surprising if
             | there was more unallocated talent in the EU, at this point?
        
         | IMTDb wrote:
         | > It seems like EU in general should be heavily invested in
         | Mistral's development, but it doesn't seem like they are
         | 
         | The EU is extremely invested in Mistral's development: half of
         | the effort is finding ways to tax them (hello Zucman tax), the
         | other half is wondering how to regulate them (hello AI act)
        
           | District5524 wrote:
           | Zucman taxes rich individuals (100mEUR+), not Mistral. AI Act
           | rules are not that difficult to comply with by GPAI model
           | providers as long as the model doesn't become systemic
           | risk... They have to spend a lot more time on PR and
           | handshaking with French politicians than on AI compliance.
           | They probably don't even have a single FTE for that... So
           | that's just prejudice I believe.
        
         | lm28469 wrote:
         | We're too busy with real life to bother with generating SVGs of
         | pelicans on bicycles sorry, but feel free to dump billions on
         | chatbots
        
         | _menelaus wrote:
         | Mistral is pursuing pursuing B2B use cases. Thats because
         | they're releasing open models and the big thing about B2B is
         | they HATE sending their data off-prem. OCR'ing and organizing
         | old docs is a huge feature in B2B. Mistral's strategy seems
         | smart to me.
        
       | film42 wrote:
       | Is open router still sending all OCR jobs to Mistral? I wonder if
       | they're trying to keep that spot. Seems like Mistral and Google
       | are the best at OCR right now, with Google leading Mistral by a
       | fair bit.
        
         | numlocked wrote:
         | (I work at OpenRouter) If you send a PDF to our API we will:
         | 
         | 1. Use native PDF parsing if the model supports it
         | 
         | 2. Use this Mistral OCR model (we updated to this version
         | yesterday)
         | 
         | 3. UNLESS you override the "engine" param to use an alternate.
         | We support a JS-based (non-LLM) parser as well [0]
         | 
         | So yes, in practice a lot of OCR jobs go to Mistral, but not
         | all of them.
         | 
         | Would love to hear requests for other parsers if folks have
         | them!
         | 
         | [0]
         | https://openrouter.ai/docs/guides/overview/multimodal/pdfs#p...
        
           | dimitri-vs wrote:
           | Chandra
        
           | vikp wrote:
           | Hey, I'm the founder of Datalab (we released Chandra OCR). I
           | see someone requested it below - happy to help you all get
           | setup. I'm vik@datalab.to
        
           | siquick wrote:
           | That links gives an error and so does
           | https://openrouter.ai/docs/guides/overview/multimodal/pdfs
        
       | hereme888 wrote:
       | I'm reading worse performance than many OSS offerings like
       | Paddle, MinerU, MonkeyOCR, etc:
       | 
       | https://www.codesota.com/ocr
        
         | nextworddev wrote:
         | Thanks for sharing this site
        
         | rafram wrote:
         | Their handwriting benchmark is not useful. The test cases
         | aren't even handwritten!
         | 
         | https://www.codesota.com/ocr/best-for-handwriting
        
           | dr_dshiv wrote:
           | Do you know of any good handwriting eval/benchmark? I haven't
           | been able to find one.
        
           | kwikiel wrote:
           | That's just the illustration. But this is misleading - I will
           | fix it asap and show real examples. I've run the mistral ocr
           | on other benchmark
        
       | vasco wrote:
       | Gave it a birth registry from a Portuguese locality from 1755
       | which my dad and I often decipher to figure out geneology and it
       | did a terrible job.
       | 
       | Regular Gemini Thinking can actually get 70-80% of the documents
       | correct except lots of mistakes on given names. Chatgpt maybe
       | understands like 50-60%.
       | 
       | This Mistral model butchered the whole text, literally not a word
       | was usable. To the point I think I'm doing something wrong.
       | 
       | The test document: https://files.fm/u/3hduyg65a5
        
         | zzleeper wrote:
         | Oh god, I'm sure I wouldn't come close to 50%; that's so hard
         | to read
        
           | vasco wrote:
           | It's tough but my dad is quite good at it. He has books of
           | common abbreviations and agglutinations from different
           | centuries. After you get used to it it's faster and very fun.
           | 
           | We were mind blown how good Gemini was at it.
        
             | ilamont wrote:
             | I am too. Gemini 3.0 fast on old scrawled diary entries in
             | English from 100+ years ago got them 95% right. It also
             | added historical context when I prefaced the images with
             | the identity of the writer, such as summaries of an old
             | military unit history in Europe post-WW1 it got from a very
             | obscure U.S. Army archive.
             | 
             | Huge timesaver.
        
         | observationist wrote:
         | Just gave it a shot with Grok 4.1 thinking - do you have the
         | ground truth translation to compare? I've tried 4 different
         | times, with slight tweaks adding information from your
         | description, and it's given me a range of interpretations. It'd
         | be nice to see if any of them got close - a couple were more
         | like pulpy telenovela plots, lol.
         | 
         | The model might need tuning in order to be effective - this is
         | normal for releases of image mode models, and after a couple
         | days, there will be properly set up endpoints to test from, so
         | it might be much better than you think. Or it could be really
         | bad with turn of the 19th century portugese cursive.
        
         | amelius wrote:
         | Forgivable, as that's a quite atypical document, I'd say.
        
           | vasco wrote:
           | Not atypical enough for Gemini is my point. Also its one of
           | the most common hand written document types in existance
           | since at the time almost nobody other than the local priest
           | knew how to write and birth and marriage certificates were
           | probably the only written documents in whole towns and
           | villages. This is the same throughout Europe at least.
        
         | CamperBob2 wrote:
         | Quick tip: when you digitize a page, put a sheet of black paper
         | behind it. That keeps the ink on the other side from bleeding
         | through.
        
           | vasco wrote:
           | You can tell that to the national archives!
        
       | tecoholic wrote:
       | > Mistral OCR 3 is ideal for both high-volume enterprise
       | pipelines and interactive document workflows.
       | 
       | I don't know how they can make this statement with 79% accuracy
       | rate. For any serious use case, this is an unacceptable number.
       | 
       | I work with scientific journals and issues like 2.9+0.5 and
       | 29+0.5 is something we regularly run into that has us never being
       | able to fully trust automated processes and require human
       | verification every step.
        
         | MallocVoidstar wrote:
         | Where are you seeing 79% accuracy? 79% only occurs on the page
         | as a win rate, not an accuracy
        
           | g947o wrote:
           | And I believe the number is 74%, compared to OCR 2.
           | 
           | What matters is whether this is better than
           | competition/alternatives. Of course nobody is just going to
           | take the output as is. If you do that, that's your problem.
        
             | skygazer wrote:
             | 79% win over OCR2 was just for English.
        
           | tecoholic wrote:
           | Right! I didn't know the difference. Does it mean for 79 out
           | of 100 documents they produce 100% accurate OCR, I doubt it.
           | The win rate sounds like a practical approximation of
           | accuracy here to me.
           | 
           | If I am wildly off, I am happy to learn.
        
             | argsnd wrote:
             | 79 out of 100 documents Mistral OCR 3 provides better
             | output than Mistral OCR 2.
        
             | ricardobeat wrote:
             | 79% of the time it beats the previous model.
             | 
             | The previous version already achieved up to 99% accuracy in
             | multiple benchmarks, already better than most OCR software.
        
               | tecoholic wrote:
               | Thank you.
        
         | knrz wrote:
         | Those are tricky! We've found https://www.datalab.to/ to be
         | good for this @ thesynthesis.company
        
       | GZGavinZhao wrote:
       | Does it handle math expressions (those rendered from LaTeX) well?
       | I've been looking for a good OCR model to transcribe my math
       | textbooks into markdown (obviously ignoring the images and
       | figures) with LaTeX as math expressions, and none of the current
       | OCR models work reliably enough.
       | 
       | EDIT: you can try it yourself for free at
       | https://console.mistral.ai/build/document-ai/ocr-playground once
       | you create a developer account! Fingers crossed to see how well
       | it works for my use case.
        
         | RagnarD wrote:
         | Please post an update on how well it works for you.
        
         | loaf_api wrote:
         | I've just finished processing thousands of documents using the
         | Gemini Pro 3 vision model and it outperformed every OCR and
         | image model I've tested by a long shot, perfect markdown with
         | latex for the math every time.
        
           | pacman1337 wrote:
           | what prompt are you using?
        
           | lysecret wrote:
           | 3 flash is also insanely good even slightly outperforms 3 pro
           | for me.
        
         | nerbert wrote:
         | Just need to open the link to answer that question.
        
       | singularity2001 wrote:
       | No one mentioning the possibly most beautiful css effect on the
       | Internet??
        
         | jbk wrote:
         | How so?
        
       | singularity2001 wrote:
       | Not OS / free weights right?
        
       | jesuslop wrote:
       | I am testing it as a replacement of MathPix, first few tests look
       | rather decent. In python for windows:
       | https://pastebin.com/uyiFHKdJ (alpha version prototype). Launches
       | windows snip tool, waits for clipboard image, calls Mistral,
       | retrieves markdown and puts it as text in the clipboard, ready to
       | be pasted in Typora, Obsidian, or other markdown editor.
        
       | 7thpower wrote:
       | My main beef with mistral is that they don't bother to respond to
       | customer inquiries for products the hide behind "reach out for
       | pricing" terms, so even if they were better than SoTA it wouldn't
       | really matter.
        
         | 650REDHAIR wrote:
         | I absolutely loathe dealing with sales people.
         | 
         | I will pay a premium for an inferior product or service if it
         | means I don't have to deal with sales people.
        
           | 7thpower wrote:
           | Agreed. In this case the offering just fit neatly into a non
           | core stack we had designed and displaced a bunch of stuff
           | didn't want to build ourselves.
           | 
           | I also hate dealing with sales people and am not going to
           | reach out to them via another avenue as they will try and
           | posture as if they're doing us a huge favor (in contrast to
           | me begging gdb for gpt4 api access).
        
       | temp0826 wrote:
       | My current holy grail is my attempt to convert a Shipibo (an
       | indigenous Peruvian language)-to-Spanish dictionary into a
       | Shipibo-to-English dictionary. The pdf I have (available freely
       | on archive.org) isn't a great scan (though I think it'd be a heck
       | of a lot easier than some of the handwritten examples they show).
       | Layout (2-columns) along with header/footers can cause some
       | headaches, but it is all Latin script. This seems to fall on its
       | face pretty badly (not even a couple of pages in), so my search
       | continues. (The other major problem I'm having is trying to
       | separate out Shipibo definitions/examples from the Spanish ones,
       | and _only_ translating the Spanish to English...so pretty complex
       | I guess. I 've been taking fresh stabs at this project every few
       | months when I see OCR/LLM news pop up and continue to be
       | disappointed)
        
         | culi wrote:
         | I'm assuming you're interested in studying Ayahuasca
         | traditions?
         | 
         | I recently learned that traditionally in Shipibo culture,
         | ayahuasca was never meant to be given to "the normal mind".
         | Instead the maestras would be the ones taking the ayahuasca in
         | order to help guide them into diagnosing people dealing with
         | various sicknesses.
         | 
         | These maestras were also ranked by how many different plants
         | they'd done a dieta on. A dieta is kinda similar to fasting.
         | You can't shower with soap, you can't have sex, you can't have
         | too much salt/seasoning, can't be exposed to too much smoke,
         | can't have alcohol, etc. And you use that specific plant
         | throughout your time. Basically you want to eliminate any
         | conflicting variables so you can experience the plant as purely
         | as possible to understand its effects. Traditionally these
         | dietas could last over a year but modern day maestros typically
         | do them for just a few weeks.
         | 
         | I don't really have a point to this. Just found it fascinating
         | how deeply and strictly they study certain plant medicines and
         | wanted to share
        
           | temp0826 wrote:
           | Yes essentially. I've got a few resources cobbled together
           | over the last few years but it'd be really nice to have this
           | reference (my Spanish isn't the best, and running to the
           | translator for a definition can be a little annoying). Also
           | to share with fellow learners/apprentices I know. There are a
           | couple of classes out there (which are actually geared more
           | toward the ceremonial/icaro language, not purely
           | conversational Shipibo, which is a bit simpler as you don't
           | need to worry as much about conjugation and other
           | complexities) which I might look into eventually.
           | 
           | (Fwiw I've accumulated a couple years worth of dieta under my
           | belt and am well aware of the restrictions! It's indeed very
           | fascinating, been pretty serious about it the last few years
           | and I've barely scratched the surface)
        
             | canucker2016 wrote:
             | Couldn't you use your smartphone and Google Lens (on
             | Android, Google app on iOS includes Google Lens
             | functionality) to translate the Spanish to English?
             | 
             | FYI - Lens on Android does in-place language translation
             | including attempting to use the same/similar font that the
             | original language is written/printed.
             | 
             | Unfortunately, I don't think Lens can be used in an
             | automated batch translation mode to convert an entire
             | book/multiple pages
        
           | mkaic wrote:
           | I suppose both of us watched the same Youtube video by Metta
           | Beshay (i think that is his name?)
        
             | temp0826 wrote:
             | I actually did too lol. I was pleasantly surprised because
             | it was actually decent and realistic about the situation (a
             | lot of people get this romantic idea about going to the
             | jungle to live and learn with the indigenous and have an
             | "authentic" experience, and this does a pretty good job if
             | dispelling that).
        
         | knadh wrote:
         | Once you have managed to get the data out and structured, you
         | may want to check out dict.press. It's a dictionary publishing
         | and management tool (which I maintain). Multiple widely used
         | Indian dictionary projects run on it.
        
           | temp0826 wrote:
           | Will take a look, I assumed there'd be some tools along those
           | lines, thanks for the suggestion
        
         | b112 wrote:
         | I applaud your efforts, but that seems difficult to me. There's
         | so much nuance in language, and the original spanish
         | translation would even be dependent upon locale-destination of
         | the original dictionary. Which would also be time based, as
         | language changes over time.
         | 
         | And that translation is likely only a rough approximation, as
         | words don't often translate directly. To add in an extra layer
         | (spanish -> english) seems like another layer of imperfect (due
         | to language) abstraction.
         | 
         | Of course your efforts are targeting a niche, so likely people
         | will understand the attempt and be thankful. I hope this
         | suggestion isn't too forward, but this being an electronic
         | version, you could allow some way for the original spanish to
         | be shown if desired. That sort of functionality would be quite
         | helpful, even non-native spanish speakers might get a clearer
         | picture.
         | 
         | What tools are you using to abstract all of this?
         | 
         | If the spacing and columns of the images are consistent, I'd
         | think imagemagick would allow you to automate extraction by
         | column (eg, cutting the individual pages up), and OCR could
         | then get to work.
         | 
         | For the Shipibo side, I'd want to turn off all LLM
         | interpretation. That tends to use known groupings of words to
         | probabilistically determine best-match, and that'd wreak havoc
         | in this case.
         | 
         | Back to the images, once you have imagemagick chop and sort,
         | writing a very short script to iterate over the pages, display
         | them, and prompt with y/n would be a massive time saver. Doing
         | so at each step would be helpful.
         | 
         | For example, one step? Cut off header and footer, save to dir.
         | Using helpful naming conventions (page-1, and
         | page-1-noheader_footer). You could then use imagemagick to
         | combine page-1 and -age-1-noheader_footer side by side.
         | 
         | Now run a simple bash vet script. Each of 500 pages pops up,
         | you instantly see the original and the cut result, and you hit
         | y or n. One could go through 500 pages like this in 10 to 20
         | minutes, and you'd be left with a small subset of pages that
         | didn't get cut properly (extra large footer or whatever). If
         | it's down to 10 pages or some such, that's an easy tweak and
         | fix for those.
         | 
         | Once done, you could do the same for column cuts. You'd already
         | have all the scripts, so it's just tweaking.
         | 
         | I'm mentioning all of this, because combo of automation plus
         | human intervention is often the best method to something such
         | as this.
         | 
         | Anyhow, good luck!
        
           | temp0826 wrote:
           | Thanks for the suggestions, I do appreciate it. I was being
           | pretty brief with my post but I really have spent a lot of
           | time and tried this from a number of angles. I've had good
           | luck with non-LLM tools to do the initial OCR, but it's not
           | context aware especially about column/page breaks (like I
           | mentioned it's kind of a dirty scan, and if the breaks happen
           | on a Shipibo part it barfs a bit. Good for a rough search at
           | least).
           | 
           | I would _love_ to create a json version of it that would
           | essentially have a bunch of fields for each word (Shipibo
           | /Spanish/English word/definition/example, type of word, etc).
           | It's further complicated by how words can be modified in
           | Shipibo (it's actually a very technical language- words can
           | have any number of prefixes and suffixes tagged on to change
           | their meaning and their precision. In their "icaros", the
           | healing songs they sing in ceremony, the most technical use
           | of the language is considered to be the most beautiful.
           | Essentially poetry from their "medical" jargon).
           | 
           | I've done some human-in-the-loop attempts but still come up
           | short in one way or another (I end up getting frustrated and
           | throwing my hands up after seeing how much time I dump on
           | it). So I figure this will remain a good test as the tools
           | (and my prompting abilities) get better. It's definitely not
           | urgent for me.
        
       | speff wrote:
       | This might be a good place to check the options available for OCR
       | in-place translations. I took a look at OCR3, but it doesn't seem
       | to support my use-case. It looks more tailored towards data
       | extraction for further processing.
       | 
       | I've got some foreign artbooks that I would like to get
       | translated. The translations would need to be in place since the
       | placement of the text relative to the pictures around it is
       | fairly important. I took a look at some paid options online, but
       | they seemed to choke - mostly because of the non-standard text
       | placements and all.
       | 
       | The best solution I could come up with is using Google Lens to
       | overlay a translation while I go through the books, but holding a
       | camera/tablet up to my screen isn't very comfortable. Chrome has
       | Lens built in, but (IIRC) I still need to manually select
       | sections for it to translate - it's not as easy to use as just
       | holding my phone up.
       | 
       | Anyone know of any progress towards in-place OCR/translations?
        
         | claar wrote:
         | If you don't mind a paid solution, try DEEPL. I also use Word's
         | built in document translation to good effect.
        
           | speff wrote:
           | I don't mind paying for one, though I do remember trying
           | DEEPL without much success. Can't remember the problem
           | offhand, but one of the services I tried just gave me a
           | generic error when I uploaded the PDF. My view at the time
           | was that it had a conniption and just gave up.
           | 
           | Wonder if Word uses the same system Edge has. I remember Edge
           | was also good, but like Chrome's Lens, I'd need to highlight
           | sections for it to get translated. Edge also OCR'd everything
           | very well - just didn't do the translation part
           | automatically.
        
         | haraldooo wrote:
         | I'm fairly confident this is solvable quite well with "just two
         | api calls". Are examples of those books available online?
        
           | speff wrote:
           | Sure - there are some good examples in the product pictures
           | for this book: https://www.amazon.com/hands-Takami-Kagami-
           | teaches-power/dp/...
        
       | constantinum wrote:
       | At instances where data accuracy is of paramount importance, i
       | think a hybrid route of non-llm ocr for data parsing and LLMs for
       | structured data extraction is the safe passage to tread on. Seen
       | better results for LLMWhisperer(OCR)[1] and Latest Gemini.
       | 
       | [1] - https://pg.llmwhisperer.unstract.com/
        
       | vintermann wrote:
       | I appreciate having an OCR interface rather than having to chat
       | with a bot, but unfortunately chatting with Gemini 3 gives far
       | better results than this. I gave it the document Gemini 3 got a
       | surprisingly good result on:
       | 
       | https://urn.digitalarkivet.no/URN:NBN:no-a1450-rk10101508282...
       | 
       | and the output wasn't even recognizably Danish.
       | 
       | Just out of pity I gave it a birthday card from my sister written
       | in very readable modern handwriting, and while in managed to make
       | the contents of that readable, the errors it made reveals that it
       | has very little contextual intelligence. Even if ! and ? can be
       | hard to tell apart sometimes, they weren't here, and you do not
       | usually start a birthday letter with "Happy Birthday brother?"
        
         | suspended_state wrote:
         | > got a surprisingly good result
         | 
         | > the output wasn't even recognizably Danish
         | 
         | How would you know that it's good then?
        
           | kadoban wrote:
           | I believe you misread. My reading is that Gemini 3 gave a
           | good result on a certain input, so they gave the same input
           | to this model and the result was poor.
        
             | suspended_state wrote:
             | You're correct.
        
         | hoyd wrote:
         | Voynich manuscript next :-)
        
       | amelius wrote:
       | Can we have an open source tool that uses the same API, and that
       | you can just instruct to use Mistral or any other service if you
       | think the open source tool has quality issues for a particular
       | text?
       | 
       | This makes more sense to me, as I find that FOSS OCR is quite
       | okay for most usecases.
        
       | i_am_not_groot wrote:
       | Finally a way to read doctor's prescriptions
        
       | jwr wrote:
       | Sadly, only available through a hosted API. I don't see how this
       | is useful for OCR, unless you are OK with uploading your
       | confidential documents to "the cloud"?
       | 
       | I'm still hoping for improved locally hosted models:
       | qwen3-vl:30b-a3b-thinking-q4_K_M is already really good.
        
         | dan-robertson wrote:
         | Businesses sign contracts about what happens when the data is
         | uploaded. Ultimately your purpose is to make money more than
         | maximally locking down your IP.
        
       | Western0 wrote:
       | I need solresol in any language. It are constructed for discusion
       | and negotiation on war
        
       | stri8ted wrote:
       | What languages does it support? I can't find this info anywhere
       | on the page.
        
       | ethin wrote:
       | So I tried this on the NVMe specification (I have a huge library
       | of PDFs) and it worked decently, though the output had some
       | oddities:
       | 
       | - Parts of the table of contents were headings
       | 
       | - I didn't like how tables were links to separate markdown files.
       | 
       | In theory, I could recombine everything into one document, but
       | that would require complicated Markdown parsing and manipulation
       | and I wasn't even sure how to go about that given how free-form
       | the resulting text was. I also haven't gone through the entire
       | document (it's 784 pages) to check to make sure it's correct
       | compared to what pdftotext or acrobat could create, so there's
       | that too.
        
       ___________________________________________________________________
       (page generated 2025-12-20 23:01 UTC)