In this article I explain how to speed up the creation of subtitles for your videos with automatic subtitle generator. All its information is based on the different tests I’ve done in the last 3 years.
I did my first subtitling project in 2016. At that time I had just made a video for a startup that was going to launch a Kickstarter campaign. I thought it was going to be extremely simple and that it was an opportunity to sell an additional service 🤑. I had no idea what I was getting into… At the end of this project I said to myself “I’m never doing subtitling again”.
That’s all I do today. What made me change my mind? It’s simple, I managed to optimize my creation workflow and equip myself with the right tools. I’ve managed to divide by eight the time it takes to create my subtitles.
In this article I will explain how I did it. You’ll have all the tips that will save you time creating your subtitles too ⚡
I’ll answer these questions:
- Does automatic captioning work?
- Can we use Youtube automatic subtitling?
- What is the best solution for online closed captioning?
- How do you automatically subtitle a video in a foreign language?
Here’s the plan in detail:
1 – How does automatic subtitling work?
When talking about automatic subtitling, it is important to differentiate between two needs. There is “captioning” in the original language of the video and there is “subtitling” in a foreign language. You will see that subtitling in a foreign language is more complex.
Automatic subtitles in the original language of the video
Closed captioning in the original language will use what is called voice recognition. That is to say, the solution will extract the audio from the video and then the algorithm will be able to understand and rewrite the words that are spoken.
In this step it is possible to have a dictionary that we will feed with words to allow the algorithm to widen its lexical field. This is for example very useful when using a particular or technical jargon. We can also call it a glossary.
Automatic subtitles in a foreign language with machine translation
In the case of automatic subtitling in a foreign language the algorithm will first create the subtitles in the language of the video and then on this basis it will translate it.
The quality of the machine translation will therefore depend on the quality of the captioning in the language of the video. So if you want a quality result it is best to check the original captions before translate them automatically. Every correction that is not made on the original captions will have to be made on all the foreign languages. If you only have one language it doesn’t make much difference, but if you have 15 languages it reduces your working time considerably.
How Automatic Subtitling works?
How’s it going on the algorithm side? In fact whether it’s for speech recognition or machine translation the logic is quite similar. The idea today is not to go into the technical details but to understand the logic behind it.
The result generated will depend on what is called a model. It is this model that incorporates the logic used to transform audio into text. This model will be enhanced by a set of data with which we will feed it. For example by giving it videos and then correcting the result generated by the model. The model will save its modifications and take them into account when it falls into a similar situation again.
SIf you want a much more technical presentation I invite you to read this article which is in English and which talks about machine learning and deep learning.
2 – Which solution to generate subtitles automatically?
There are different solutions available on the market. The ideal solution will depend on the need you have and on your technical skills or those of your team.
You can choose between these three main families of solutions:
- external APIs (for technical teams)
- create your own neural model (the most complex)
- use an online subtitling editor (the simplest)
External APIs for speech recognition and machine translation, if you have technical skills.
It should be noted that this solution requires technical skills.
Speech recognition API
The principle is to connect to an external service, send the video or audio to it and it will send us directly a file containing all the words with their time codes. We then have to build the subtitles from this file.
The best speech recognition APIs
As we have seen previously, each API provider will have a different model. And the model will give a more or less relevant result depending on the language you are going to use. For example Google may give you the best result for a video in English but it will not necessarily be the case for a video in Spanish. To get the best result you should test the different APIs and see which one gives the best quality depending on the videos you will have to subtitle.
Here is the list of the most powerful APIs currently used in speech recognition:
- Google Speech-to-text
Machine Translation API
For machine translation APIs we send the original text, the original language and the desired translation language. The API will then send us a text corresponding to the machine translation generated by the system.
The best machine translation APIs
Just like speech recognition APIs it is hard to say that it will give you the best result. It depends on the source language and the translation language you are going to use. Google is a leading player in machine translation, covering more than 110 languages. But it is far from being the best performer on all these languages. The little new Deepl is extremely powerful on a dozen languages including English, Spanish, Italian, German…
- Google Translate API
- Microsoft Translation API
- Translate API
- Text Translation API
- SYSTRAN.io Translation API
- MyMemory Translation API
- Deepl API
- My Translator Pro API
- Linguatools Translate API
- Yandex Translate API
- IBM Watson Language Translator API
Training your own neural network, the most complex solution.
If you have significant technical resources in your team the other solution is to build your own model powered by your data or open source data.
You can for example rely on an open source ASR such as CMUSphinx, Kaldi or more recently Common Voice by Mozilla.
This article details the differences in the complexity of implementation and the quality of the result: open source solution for automatic transcription.
Please note that if you also want to subtitle in a foreign language you will have to create another template based on machine translation. The more languages you have for subtitles, the more complex and time-consuming it will be to create and maintain your own solution.
For the open source translation solution you can use the OpenNMT solution which is one of the most advanced projects.
The simplest online automatic subtitling solution.
If you don’t have the technical skills and want to opt for the easiest solution to generate your automatic subtitles, you’ve come to the right place.
Usually it is a solution available on a website that allows you to upload your video and automatically the platform will generate subtitles in the desired language. Some offer automatic translation and others only subtitles in the original language of the video.
Let me present you the different solutions in 3,2,1…
3 – What are the best solutions for automatic online subtitling?
Automatic Youtube subtitling
YouTube’s automatic subtitling platform is one of the most widely used. It has the advantage of being directly integrated into YouTube, but it is also free of charge. If subtitle quality is not a priority then it is more than enough. But if you want professional subtitles then in this case you might spend some time correcting the machine errors. The online interface is ergonomic but requires a lot of keyboard/mouse input in order to edit subtitles. We did the test with a timer and it takes as much or more time to correct the automatic subtitling of Youtube than to rewrite it entirely. But for those looking for a free solution, this is the best. For those who are interested you have the step by step detail to use the Youtube automatic transcript in this article the best subtitling solutions on the market.
Please note that to use this solution you will have to upload your video to YouTube. But be aware that you can put it in private to prevent it from being visible to everyone.
Automatic Facebook subtitling
Facebook took a while to offer its online captioning solution. But inevitably in competition with YouTube, they had to come up with something. At the moment it’s not as good as YouTube. So if you only post on Facebook it can do the job but the other choices on this list are usually better.
Checksub : online automatic subtitling solution
By 2017 I had tested all existing subtitling solutions. Automatic solutions did not allow me to create subtitles faster. In fact, I was spending more time correcting the automatic subtitling rather than creating subtitles from scratch. So I came up with a solution that really saves time with technology. On April 1, 2020 we decided to make this solution available to everyone at www.checksub.com.
Over the last 2 years we have imagined and designed the dream solution for video professionals, communication agencies, translation agencies, e-learning trainers,… The objective is to use all available technologies to create quality subtitles.
The strength of the solution is based on 3 elements:
- the best speech recognition and machine translation APIs are used
- the online subtitle editor is powerful and very easy to use
- it is a collaborative platform for working with translators, clients or other partners
How Automatic Closed Captioning Works :
- Upload your video
- Indicate, the desired language(s)
- Reread automatic subtitles
- Export the video or subtitle source file
The Checksub platform allows you to create subtitles in 128 languages.
Kapwing is mainly made for videos for social networks. The automatic English transcript works very well. In the other languages the result is not so good. Kapwing’s automatic subtitling interface is very simple. This quickly becomes a limit if you want quality subtitling. We can also regret the absence of the audio spectrum to adjust the time codes precisely.
Rev : pour le sous-titrage anglais uniquement
Rev is a transcription and captioning company that has become known for its manual (human) captioning service. For some time now it has been offering automatic subtitling. This service is only available for English videos. If your video is in another language you will not be able to use it. The result of the English transcription is of very good quality. Their online editor is very basic. Just like Kapwing, it is an interface that is meant to be quite simple and this is done at the expense of accuracy. This solution is therefore suitable if you have a project in English that is intended for social networks and the accuracy of the time codes is not a priority for you. In terms of price Rev is in the high end of the market.
Veed.io : for online video editing
Veed.io is above all an online video editor. But they also offer an automatic subtitling function. The principle is similar to that of other platforms, you upload your video and you have a “subtitles” menu so that the platform automatically generates subtitles in the original language. They are the only ones besides Checksub to offer automatic subtitle translation. Veed.io is an ideal solution for those who need to modify the editing of their video. If you are looking for a solution to manage mainly your subtitle creations this is not the most suitable solution.
Authôt : For video transcriptions
With Authôt it is possible to create a transcript of your video. But you should know that the tool is rather designed for the automatic transcription of audio documents. As soon as you open the interface, you will notice that the editor is less adapted to subtitling than its competitors. The subtitles are not displayed on the video, it is not possible to work on cutting the script into subtitles and finally it is impossible to modify the time codes.
Knowing that their rates are higher than other platforms, this solution does not seem to be suitable for the creation of subtitles. But this solution is interesting if you exceptionally have a subtitle project to do, you are used to using Authôt for your transcription and you already have credit. In this case you don’t need to use another platform. But if your need for subtitling is becoming more and more important it would be a good idea to test other solutions.
4 – Tips to know :
Prepare your video file
Before starting to work on subtitling, it is essential that the editing be finalized. If after creating the subtitle you edit the video, the subtitle time codes may no longer match the video. This will force you to readjust all the subtitles or start from scratch. So even if you want to save time, wait until you have the final cut. If your changes do not affect the editing of the video at all (color, over-printing of text, …) it is not embarrassing and you can then start subtitling before the finalized version. But as soon as you modify the editing of the video you risk having staggered subtitles.
Optimize the download of your video
All the automatic subtitling solutions presented above require you to send the video to an external server for processing. If your video is too heavy and you don’t have the fiber, you may spend a lot of time uploading it to the server. The automatic subtitling platform will also take longer to prepare your project with a heavy video. So if you don’t need to generate an export from the platform by overlaying subtitles, it’s better to have a low definition version of your video to save time. In addition, if you have a slow Internet connection it will also save you time when loading the video on the platform. To compress your video you can use the Handbrake software available on this link: https://handbrake.fr/
In the case of a translation, correct the original subtitles.
As I explained earlier, machine translation will use subtitles in the original language. So the better the subtitles in the original language, the better the machine translation will be. So you have the choice to correct the original subtitles or spend more time on the foreign language subtitles. This is a tip that is especially valuable if you need to create subtitles in different foreign languages. Every change you make to the original subtitle saves you from having to redo it for each foreign language.
Recovering or exporting your video
Creating subtitles is good, using them is better. Depending on the use you want to make of the subtitles you will be able to export them in different formats. For example if you want to add subtitles to a YouTube or Facebook video you will use the SRT format. If you want to customize the style of your subtitles it is necessary to generate a new export of the video by embedding the subtitles in the image. Checksub allows you to generate an export by embedding subtitles in the video. Alternatively you can use a solution like Submerge to embed subtitles in a video. Here is a tutoriel complet to learn how to use Submerge to add subtitles to your video.
Add subtitles to a video already uploaded in Youtube, Facebook, Vimeo, Wistia,…
If the video you have just subtitled is already available on different platforms there is no need to upload your video again. If for example you want to keep the views on YouTube you should avoid creating a new video that will make you start from scratch on all the statistics. In this case it is best to upload the source file of the automatic subtitles, usually in SRT format, and then import it directly to the platform in question. This is an option that is available on almost all video hosting platforms that have a video player. It works with YouTube, Facebook, Vimeo, Wistia and many others. The procedure, however, is specific to each form. We plan to do a tutorial to show how to do it. But for the moment the best is to look in the FAQ space of your hosting platform.
5 – To remember
As we have seen, there are many solutions, not all of which will necessarily save you time in subtitling. It is therefore important to choose the most suitable one. As a professional, the existing solutions did not suit me. That’s why I imagined and created Checksub. So if you are a video professional, a translation agency, a global company, … I invite you to test this solution because I imagined it for you.
Thank you for your attention and I hope this article will have helped you save a lot of time on the creation of your subtitles ⚡