# Want the OCR of DT3 & DTTG3 to come back

**URL:** <https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596>\
**Category:** Feedback\
**Tags:** dt4, devonthink3\
**Created:** [March 19, 2026, 3:51am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596 "2026-03-19T03:51:39Z")\
**Posts on this page:** 19\
**Page:** 1

<div class="post-metadata">

**Author:** ![Krzin42](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/krzin42/32/34671_2.png) [@Krzin42](https://discourse.devontechnologies.com/u/Krzin42)\
**Post date:** [March 19, 2026, 3:51am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/1 "2026-03-19T03:51:39Z")

</div>

### Summary

| View by → | DT3 | DT4 & Preview | DTTG4 | PDF Expert |
| --- | --- | --- | --- | --- |
| OCR by DT3 | Good | Good | Bad | Good |
| OCR by DT4 | Good | Bad（Incl. PDFGear、PDF Reader Pro） | Bad | Good |
| OCR by DTTG3\[1\] | Good | Good | Bad | Good |
| OCR by DTTG4\[1:1\] | Good | Good | Bad | Good |
| OCR by PDFpen | Good | Good | Good | Good |

I compared the OCR effect of different software on Chinese. Overall, the OCR effect of the 3rd version is very good, while the OCR effect and viewing effect of the 4th version are very poor, especially DTTG4.

The OCR itself (OCR Layer) might be good, but the viewing experience is not satisfactory.

It feels like ABBYY has not adapted to the new Apple OSs.

By the way, DT 3 is installed on macOS 11.

The specific situation is as follows:

#### OCR by PDFpen

PDFpenPro 12.2.3 Copyright ® 2003-2021 (The version from five years ago)  
OCR Layer✅

 ![OCR by PDFpen 12](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/7/2/7226b3aed2c2ba288668d334e0bdf080435c47bb.jpeg)

OCR by PDFpen, Review by DTTG 4✅

 ![OCR by PDFpen, Review by DTTG 4](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/3/8/38026145da6095625a52408f9148d440d47cba36.jpeg)

#### OCR by DT4

OCR Layer✅

 ![OCR by DT4](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/9/2/926bd294fa4825f9317643462b95c8a125a4d9d0.jpeg)

OCR by DT4, Review by Preview & DT4❌

OCR by DT4, Review by PDF Expert✅

OCR by DT4, Review by DTTG 4❌

 ![OCR by DT4, Review by DTTG 4](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/9/3/937d7cbdce49690bd4a099a80f727cd1c8fb2daf.jpeg)

OCR by DT4, Review by DT3✅

#### OCR by DTTG4

OCR Layer✅

 ![OCR by DTTG4 CN+EN.png](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/3/d/3d3b5aacaa163acaa662b89f897569f0a856cdb6.jpeg)

#### OCR by DT3

OCR Layer✅

 ![OCR by DT3.png](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/6/1/61443ef1dde31452e1190476a3694e3dad5a1b7d.jpeg)

OCR by DT3, Review by DT4✅

OCR by DT3, Review by DTTG4❌

 ![OCR by DT3, Review by DTTG4](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/2/f/2f080fdf76d421a46cc4256b4bf5bea101adbd3f.jpeg)

#### OCR by DTTG3

OCR Layer✅ (OCR Setting: English + Chinese)

 ![OCR by DTTG3 CN+EN](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/0/2/02aaa6b102f92dd6c837325fa9644d2185ebf10a.jpeg)

OCR Layer❌ (OCR Setting: Chinese) (DTTG4 the Same)

 ![OCR by DTTG3 CN.png](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/7/d/7d1438422d7e249d7deb09d0370bb526b731f119.jpeg)

DTTG 3&4 both have a problem: after OCR is completed, the whole page becomes larger (white part)

 ![截屏2026-03-19 11.47.20](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/f/1/f198f4f3e71c8f9892bae56b7ebfe5bf7d0d0742.jpeg)

* * *

1. The OCR setting on DTTG only opens Chinese, which is very poor. At the same time, English + Chinese is turned on, and the effect is greatly improved.

---

<div class="post-metadata">

**Author:** ![cgrunenberg](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/cgrunenberg/32/7172_2.png) [@cgrunenberg](https://discourse.devontechnologies.com/u/cgrunenberg)\
**Post date:** [March 19, 2026, 6:49am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/2 "2026-03-19T06:49:42Z")

</div>

> [@Krzin42](#):
>
> By the way, DT 3 is installed on macOS 11.

The OCR engine of DEVONthink 3 & 4 is the same. Did you use the same settings and the same macOS version for testing?

---

<div class="post-metadata">

**Author:** ![Krzin42](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/krzin42/32/34671_2.png) [@Krzin42](https://discourse.devontechnologies.com/u/Krzin42)\
**Post date:** [March 19, 2026, 7:15am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/3 "2026-03-19T07:15:17Z")

</div>

Can I install DT4 & 3 on macOS 26 at the same time now? Will it affect the database or cause issues?  
Now I’m using 4.2.2

---

<div class="post-metadata">

**Author:** ![cgrunenberg](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/cgrunenberg/32/7172_2.png) [@cgrunenberg](https://discourse.devontechnologies.com/u/cgrunenberg)\
**Post date:** [March 19, 2026, 8:47am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/4 "2026-03-19T08:47:59Z")

</div>

> [@Krzin42](#):
>
> Can I install DT4 & 3 on macOS 26 at the same time now? Will it affect the database or cause issues?

I would recommend to install DEVONthink 3 in the Applications folder of a dedicated user account.

---

<div class="post-metadata">

**Author:** ![Krzin42](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/krzin42/32/34671_2.png) [@Krzin42](https://discourse.devontechnologies.com/u/Krzin42)\
**Post date:** [March 19, 2026, 3:13pm UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/5 "2026-03-19T15:13:59Z")

</div>

I’ve created a new user and am about to install DT3 and stopped 😅. Let’s not take this risk.  
At least, it is certain that files that have undergone DT4 OCR cannot be properly annotated in DTTG4 and DT4 either. (Chinese PDFs)

---

<div class="post-metadata">

**Author:** ![cgrunenberg](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/cgrunenberg/32/7172_2.png) [@cgrunenberg](https://discourse.devontechnologies.com/u/cgrunenberg)\
**Post date:** [March 19, 2026, 3:19pm UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/6 "2026-03-19T15:19:30Z")

</div>

Did this really happen after upgrading to DEVONthink 4 or after upgrading to macOS 26? Tahoe is not known for being rock solid.

---

<div class="post-metadata">

**Author:** ![BLUEFROG](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/bluefrog/32/135_2.png) [@BLUEFROG](https://discourse.devontechnologies.com/u/BLUEFROG)\
**Post date:** [March 19, 2026, 5:18pm UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/7 "2026-03-19T17:18:14Z")

</div>

> [@Krzin42](#):
>
> I’ve created a new user and am about to install DT3 and stopped 😅. Let’s not take this risk.

Just to clarify: The previous comment wasn’t suggesting a methodology for your testing. It was giving you a recommendation on appropriately separating version 3 and 4.

---

<div class="post-metadata">

**Author:** ![Krzin42](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/krzin42/32/34671_2.png) [@Krzin42](https://discourse.devontechnologies.com/u/Krzin42)\
**Post date:** [March 20, 2026, 1:06am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/8 "2026-03-20T01:06:39Z")

</div>

I don’t quite remember. Because there was a time lag between the emergence of DT3, DT4, DTTG4 and macOS 26, and I didn’t use the OCR function of DT much before.

At first, I found that the files (Chinese PDFs) after OCR with DT3 couldn’t be annotated by DT3 itself (I reported this issue a few years ago, unless ForceEditablePDFS was enabled). Therefore, in the past, I mainly used PDFpen 12 for OCR.

After buying the M4 Mac last year, I found that the PDFpen 12 couldn’t be installed normally on the new machine. Therefore, alternative solutions need to be considered. [The new version of Nitro PDF too expensive](https://discourse.devontechnologies.com/t/ocred-pdf-becomes-abnormal-big-on-dt4-after-annotating/85716/7).

Later, I spent $70+ on PDF Reader Pro and used its 3.x version for OCR. DT could annotate, but it didn’t work after upgrading to 4.x.

At this point, I no longer wanted to spend money on this hassle. Therefore, I opened ForceEditablePDFs, but found that its compatibility with Chinese in Ver.4 (DT & DTTG) was too poor. DTTG 4 initially couldn’t recognize Chinese properly. After the feedback, you have made significant improvements and can now perform OCR normally (thank you very much), but it still can’t annotate normally at present.

---

<div class="post-metadata">

**Author:** ![Krzin42](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/krzin42/32/34671_2.png) [@Krzin42](https://discourse.devontechnologies.com/u/Krzin42)\
**Post date:** [March 20, 2026, 1:08am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/9 "2026-03-20T01:08:50Z")

</div>

Thank you for the clarification.  
Yes, I understand that there are too many variables to test on different macOS versions. It would be best to test on the latest system – control variables.

I just have never used the function of “multi-user” and I’m worried that there will be unexpected problems. Therefore, stop continuing the test. If there is a new system in the future, I will try to install DT3 first and uninstall it after testing.

---

<div class="post-metadata">

**Author:** ![cgrunenberg](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/cgrunenberg/32/7172_2.png) [@cgrunenberg](https://discourse.devontechnologies.com/u/cgrunenberg)\
**Post date:** [March 20, 2026, 7:24am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/10 "2026-03-20T07:24:22Z")

</div>

Could you share a document (before/after OCR) and a screenshot of your OCR settings? Thanks in advance!

---

<div class="post-metadata">

**Author:** ![Krzin42](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/krzin42/32/34671_2.png) [@Krzin42](https://discourse.devontechnologies.com/u/Krzin42)\
**Post date:** [March 20, 2026, 8:42am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/11 "2026-03-20T08:42:35Z")

</div>

The conclusion is the same as the Summary above.  
Special notes:

1. DT4 performs better on documents with similar fonts (such as provided standard files). If there are significant font, boldness, italics differences, etc. on the same page, the effect will be very poor (you can refer to the highlighted yellow parts).
2. DTTG4 can only provide annotations in English (with notes “HL by DTTG4”).
3. The file size becomes very large after DTTG4 OCR (2 times and 10 times)

The total files are quite large (34M) and cannot be attached here. Are there any other ways I can provide it to you?

 ![DT4 OCR Setting](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/4/2/42f97a251ba6ac87ee1b4439c294bdb038fda38f.png)  
 ![DTTG4 OCR Setting](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/c/6/c65cf6b9bd2dc66a2bf05012afd8467b387b60e1.png)

---

<div class="post-metadata">

**Author:** ![aedwards](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/aedwards/32/413_2.png) [@aedwards](https://discourse.devontechnologies.com/u/aedwards)\
**Post date:** [March 23, 2026, 10:01am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/12 "2026-03-23T10:01:15Z")

</div>

1. I couldn’t find any issue with the highlighted sections of the documents you sent, the layout and selection of the characters was as expected.

1. Currently Annotations are not available for some languages as Apple’s PDFKit will corrupt the text layer on saving the file.

2. We are currently working on improvements to the DTTG OCR.

---

<div class="post-metadata">

**Author:** ![Krzin42](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/krzin42/32/34671_2.png) [@Krzin42](https://discourse.devontechnologies.com/u/Krzin42)\
**Post date:** [March 23, 2026, 10:44am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/13 "2026-03-23T10:44:56Z")

</div>

It seems that our results are different. Is there anything I can do to figure out the reasons why I am different from you? Or is this just normal?

PS: This video was recorded after re-importing the file sent to you into DT4

I’m glad to hear that DTTG OCR will be improved. I’m looking forward to it.

---

<div class="post-metadata">

**Author:** ![aedwards](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/aedwards/32/413_2.png) [@aedwards](https://discourse.devontechnologies.com/u/aedwards)\
**Post date:** [March 23, 2026, 10:59am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/14 "2026-03-23T10:59:33Z")

</div>

Which version of macOS are you using? I tested on Tahoe 26.3.1

---

<div class="post-metadata">

**Author:** ![Krzin42](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/krzin42/32/34671_2.png) [@Krzin42](https://discourse.devontechnologies.com/u/Krzin42)\
**Post date:** [March 23, 2026, 12:13pm UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/15 "2026-03-23T12:13:36Z")

</div>

Tahoe 26.3.1(a)  
M4 Mac mini

---

<div class="post-metadata">

**Author:** ![aedwards](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/aedwards/32/413_2.png) [@aedwards](https://discourse.devontechnologies.com/u/aedwards)\
**Post date:** [March 23, 2026, 3:20pm UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/16 "2026-03-23T15:20:53Z")

</div>

That is very odd, the text selection shouldn’t be different for the same file on different machines. is your Mac system language Chinese Simplified, English or another language?

---

<div class="post-metadata">

**Author:** ![Krzin42](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/krzin42/32/34671_2.png) [@Krzin42](https://discourse.devontechnologies.com/u/Krzin42)\
**Post date:** [March 23, 2026, 11:04pm UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/17 "2026-03-23T23:04:22Z")

</div>

Chinese Simplified  
English

 ![截屏2026-03-24 07.03.38](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/4/f/4fc0813c70a12e9cd8be5d22b1fe0eb85cdb177f.png)

Another Mac Only Chinese Simplified

 ![截屏2026-03-24 08.22.13](https://devontech-discourse.s3.dualstack.us-east-1.amazonaws.com/uploads/original/3X/8/8/88fcb51f01a7ba5b9bc13f9108a370416bb8822d.png)

The result is the same.

---

<div class="post-metadata">

**Author:** ![aedwards](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/aedwards/32/413_2.png) [@aedwards](https://discourse.devontechnologies.com/u/aedwards)\
**Post date:** [March 24, 2026, 9:31am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/18 "2026-03-24T09:31:55Z")

</div>

I can reproduce your issue with Chinese Simplified as the primary language. It appears to be an issue with Apple’s PDFKit as the incorrect selection happens in Preview too.

---

<div class="post-metadata">

**Author:** ![Krzin42](https://discourse.devontechnologies.com/user_avatar/discourse.devontechnologies.com/krzin42/32/34671_2.png) [@Krzin42](https://discourse.devontechnologies.com/u/Krzin42)\
**Post date:** [March 24, 2026, 10:09am UTC](https://discourse.devontechnologies.com/t/want-the-ocr-of-dt3-dttg3-to-come-back/86596/19 "2026-03-24T10:09:23Z")

</div>

Yes, the performance of the preview is the same as that of DT4. It seems that this will be a difficult problem involving Apple’s PDFkit.  
Thank you for taking the time to test.