The Mac version has also been released as the 0.8.0 of the local translation app has been lined up on the App Store.
There are a lot of this version. While improving ease of use, it also includes some new features.There are a lot of this version. While improving ease of use, it also includes some new features.
- New feature: Zoom text (pinch in and out for iOS, ⌘ key for macOS)
- New function: Original translation clutch function that highlights the original text and translation
- New feature: Word documents, text extraction from PDFs that can extract text (Mac version implemented in 0.7.0)
- New feature: Image text OCR
- Interface: Add “Open” in the file menu for iOS and “Open” for Mac
- Interface: Display the status of the language pack in the language menu of machine translation
- Interface: Mac version with compact mode
By loading images, I grew from the stage of trying to make local translations to become a practical and meaningful app. As I write later, my own behavior of using this app is also starting to change. Of course, all the new features have been completed.Even when performing OCR from an image, you will not connect to the Internet or send an image to a server. Then, let’s take a look at them one by one.
Zoom
It is equipped with a 5-stage character size change function. The macOS version uses the familiar shortcut, the ⌘ key and the three keys =, -, and 0, and iOS changes it by pinch-in-out. You can change it immediately and at any time.
Even with presbyopia, when your eyes get used to sentences, you will want to read a lot of small characters. When you read in a hurry, it’s better to be big. That’s why we adopted an interface that can be changed on the spot. Changing the character of iOS is a bit pleasant. Please give it a try.
Translation Clutch
We decided to call it “Clutch” by installing a function that highlights and displays the contrasting part when you click or tap either the original text or the translated text. It is an image of two text fields interlocking.I used to think about the name of a new feature, but it’s fun to give a new name to a feature that I don’t often see.
From the moment I decided to make it a 2-pane translation app, I had an image of linking highlights, but I feel that it is working satisfactorily. The former Babel translates about the middle novel, but the quality is by no means high compared to the translation using cloud generation AI.Sometimes, it can be a translation that is difficult to understand. At such a time, you can tap the translation to check the original text, which is the translation clutch of the former Babel.
There is a mechanism that highlights the contrasting part for both DeepL and Google Translate, and various functions can be called.Information such as how to read, dictionaries, and other translation candidates is appreciated when you are working. But that’s it. I thought that a slightly different original translation highlight was necessary for an app used in a small screen and uncommunicable environment, so I made it a very simple function.When checking information about meetings and events, reading novels, or having a contract in mind on a narrow seat on an airplane, refer to the original text a little and illuminate the text from a different angle. It is a function that I look forward to hearing how you use it.
Text extraction and OCR
Pre-Babel can extract text from a variety of files. Typical files are the following 4 types. Text files such as .txt and .md, PDF text layers, Word .docx and .doc format files, and images.
There is no difference in the menu, but the reading method varies from file to file. Let’s explain each one briefly.
A text file
Read the text file as it is. I don’t see the extension. Import .md, .csv, or .html and translate it as a text file.It doesn’t make sense at all, but you can also read .js and .css. I want to stop it because it’s the source of the accident, but is there any good way…
Word document
Use the functions of macOS and iOS to read the text part as it is.I don’t want to read character modifications, but I read bold characters, etc., and when translated, it may change to tags, etc. You can read the file you received by email because you can read it without any problems even if you write vertically.
PDF file
Use the functions of macOS and iOS to load text layers and connect line breaks. In fact, this function was difficult. PDF text layers are arranged by cutting sentences in the manners of the created application.
This is the first sentence of “Industrial Policy for the Intelligence Age: Ideas to Keep People First” announced by OpenAI, but the highlighted part is a new line as follows.
誰もこの移行がどのように展開するか正確にはわかりません。OpenAIでは、私たちは、↩️ 人々が望むAIの未来を形作るための実質的な力を与える民主的なプロセスを通じてそれを乗り越え、さまざまな可能な結果に備えると同時に、適応能力を構築すべきだと考えています。↩️ No one knows exactly how this transition will unfold. At OpenAI, we I believe that we should overcome it through a democratic process that gives people the real power to shape the future of AI they want, prepare for various possible outcomes, and at the same time build the ability to adapt.
In the middle of the sentence, there is a new line after should, so the translation is strange.This example is still good, but in a negative sentence, if a line with not is expelled from another sentence, it may be translated into a completely opposite sense. Therefore, the former Babel decided to connect the loaded line breaks with a simple rule. Just before opening the business, the symbol at the end of the sentence (. .!?If there is no), the line break will be linked, the lines that appear in the headline will not be connected, the lines with only numeric symbols will not be connected, and the lines that are extremely shorter than the surrounding area will not be connected… It is a rule like this. Now it’s much better.
Translation
No one knows exactly how this transition will unfold. At OpenAI, we believe that we should overcome it through a democratic process that gives us the real power to shape the future of AI that people want, prepare for various possible outcomes, and build the ability to adapt at the same time.
Even so, vertical PDFs don’t work well. It is easy to be imported one by one, so it is connected and loaded, but if there is a ruby in it, the paragraph will collapse, and if it is a tier, the upper and lower lines may be connected.I don’t know what I can do… Anyway, the PDF format I tried is only a small part. If you are in trouble because you can’t read it, please contact the customer support form. I will do my best if it is a compatible format.
Image OCR
Using Apple’s image recognition function Vision Framework, OCR from images is also installed. This is quite powerful. Before the previous Babel, it is possible to import and translate from screenshots to photos and finally handwritten (relatively carefully written).
While introducing the operation on the iPhone, we will introduce how OCR works. First, select a photo in the Photos app and open the sharing panel.
Vision Framework read the Hangul written on a slanted, wrinkled paper almost accurately. The Vision Framework also solves the problem of unnatural line breaks in PDF.Text read as a block naturally connects line breaks. I’m so grateful.
By the way, the slightly disturbing black print is Korean writer Hong Ji-eun (who is also an anthology mate on July 7th. ” Ninety-nine Beasts”, it’s interesting!)It is said that it was printed and distributed when he went to the demonstration against the coup. I also think that the former Babel, which does not rely on communication and servers, will be the most powerful program in such a place.
Image OCR is slightly changing my behaviour.Recently, the number of screenshots that cannot be translated in X’s posts has increased, but even in such a case, if you press and hold the image and send it to the front Babel, you can read it in Japanese. Therefore, I have read more texts that I left behind.Even if you try to read it on the site, “Truth Social” is too heavy to read.
Interface touch-up
In addition to new features related to loading, we are also working on the interface.In the Mac version, we have created a compact mode like the iPhone so that it can be used without stress on the Macbook. The layout when the iPad is horizontal has also been modified so that the full screen can be used.
What was a bit troublesome was the machine translation language menu. Apple Translatio Framework cannot be used unless it is downloaded in advance. After I couldn’t communicate, “Wow, I didn’t get in!” Even if you think so, it’s too late.So I thought there was no way to know in advance, so I thought it was to display an icon that prompts the download after the language name. I haven’t downloaded the Spanish selected on the screen below yet. If you are going to the Spanish-speaking world, tap and download it.
By the way, I was tempted to wear a national flag when I was making this language menu, but there are few countries where the language and country match, so I stopped. When South Americans choose their mother tongue, it is not good to think about it that they have to choose the flag of the ancestors (or invaders) of the country that has crossed the Atlantic Ocean.
In the future
Thanks to the loading from PDF and image OCR, the former Babel has become a special app. When reading the text of screenshots posted on X, it naturally loads into the previous babel, and now it is also used to read untranslated receipts.There is no need to bother thinking about how to use it.
I would like to put an end to this, but I’m actually thinking of going one more thing. You have to do your best until you put on the “eyes” on the iconic Babellobo.I don’t know if I can make a live replacement like Google Lens, but I’m thinking of implementing it up to the translation of menus where the position of characters is important. If that happens, it will become a must-have app for travelling, but it should take a little time. Please wait patiently.
