Get a quote
Palexgroup - Case Studies - Medical Webinar Video Subtitling

Medical Webinar
Video Subtitling
for YouTube using
Machine Translation

This case study is about collaboration between a leading medical technology company and Palex Group on the task of preparing and translating the subtitles for neurosurgery and radiosurgery webinar on a limited budget. Our task was to localize webinar from Spanish into English for YouTube channel.

The challenge

Our long-term client, an international digital medical technology company, is engaged in educational work, creating and promoting training videos for doctors, patients and caregivers.

Client’s inquiry: to translate specialized neurosurgery and radiosurgery webinar from Spanish into English and to create English subtitles.

Special requirement: to save the budget as much as possible while keeping the appropriate quality.

Misc: machine translation (MT) was allowed as a possible option for the first time in our cooperation.

Highly complicated medical content and video format

The high speed of speech in webinar with plenty of niche medical terms in Spanish. Translation should be understandable to English speaking medical professionals and should not sound awkward.

Focus on budget saving

The fact that human effort should be minimized and use of tools is encouraged leads to higher quality risks. Speech-to-text auto-recognition tools can skip part of source text during transcription. MT use is allowed but it is unsafe to use it for medical content.

The solution

Risk-free comprehensive full-cycle process

  • Transcription of the source text using automated speech‐to‐text tool
  • Manual review of the transcription output
  • Translation and review by 2 linguists
  • Subtitles integration into the video
  • Linguistic sign‐off of the English subtitles integrated in the video
Too bulky and expensive for the purpose!
The cost of this process is equal to the cost of producing the same webinar in English from scratch.

Thrifty and nifty 2-step fully automated process

  • Speech‐to‐text (STT) auto‐recognition
  • Spanish → English Machine Translation
Too risky and wild!
So attractive and low-cost but. . . there is a high risk of a messy and awkward result even with great progress such automated tools have been showing recently. Fluoroscopy can transform into flowers copying: beautiful but misleading.

Is there a balance between bearable output and nice price?

Challenge was accepted!
Palex elaborated 3 scenarios for testing.

Fully automated 2‐step solution.

1 Step
STT auto‐recognition
2 Step
Spanish → English Machine Translation

Source improved 3‐step solution.

1 Step
STT auto‐recognition
2 Step
Rough human finetuning of Spanish transcription
3 Step
Spanish → English Machine Translation

Source and target improved 4‐step solution.

1 Step
STT auto‐recognition
2 Step
Rough human finetuning of Spanish transcription
3 Step
Spanish → English Machine Translation
4 Step
Human Machine Translation Post Editing of English text

What STT tool would give us the best Spanish source text?

The better the quality of the initially generated
text — the less improvement efforts are
required down the process.
1 We shortlisted STT tools presented on the market in 2021 under the scope of our criteria (basic price and additional payments during the work, auto-recognition technical quality, embedded MT engine, captions type, pool of languages, etc).
Captions Hub
2 Palex Linguistic Lead chose the best part of the webinar for testing with the 3 selected tools.
3 Spanish linguist analysed STT samples from the 3 preferred tools and Veed showed the best result.

Which scenario shows reasonable
translation quality?

01. Fully automated 2‐step solution.

  • STT auto‐recognition
  • Spanish → English Machine Translation
Completed fully via Veed.

02. Source improved 3‐step solution.

  • STT auto‐recognition
  • Rough human finetuning of Spanish transcription
    Human finetuning is necessary to improve the input quality for MT.
    It was quite a challenge to decide on the criteria of what mistakes
    should be marked and corrected as “critical issue” and what could be ignored.
    Thanks to our dedicated
    and professional Spanish team we prepared correction guideline.
  • Spanish → English Machine Translation
✌️ Significantly improved quality compared to 2-step solution is approved by linguistic experts.

03. Source and target improved 4‐step solution.

  • STT auto‐recognition
  • Rough human finetuning of Spanish
  • Spanish → English Machine Translation
  • Human Machine Translation Post Editing of English text
    "Light MTPE" guide was prepared to define the critical
    issues to be corrected.
💡Three native English experts including
3rd party independent linguist compared
3-step and 4-step solution results.
Surprisingly translation quality improved
insignificantly while we expected MTPE
to bring more value.

The results

Type of
17 days
The base for comparing.
Fully automated
2‐step solution
1 day
not appropriate
Automated STT tool is a useful aid.
Properly selected tool decreased the
volume of work. However, STT tool’s
operation is not perfect and good
amount of source text can be skipped
which leads to manual transcription
and additional costs in any scenario.
Source improved
3‐step solution
3 days
Human speech-to-text finetuning took
about 10-15% of budget and showed
the best effect on final MT output and
less PE effort. This is why we decided
to go with this option.
Source and target
4‐step solution
5 days
“Light MTPE” took about 20-25% of
budget and did not improve the quality
much for this type of text.
9 tech solutions tested
3 workflows elaborated
1 best solution selected
75% budget saved
14 days faster
Project quote for client lowered to
25% compared to initial one and
the project was completed
14 days faster.

Expert says

This solution is recommended for the specialized multimedia content which is viewed by the professionals.
It is also a good fit for limited budget requests when you want to have the translation free from critical/awkward errors. You should not expect ideal result but can be sure that the end user who reads the subtitles will understand the content.

All this coupled with substantial money and time savings present an attractive and justified solution.

Vlada Klimova

Project Manager


Multimedia Localization Palex has expertise and dedicated Multilanguage team for about 20 medical spheres and is ready to support you with international expansion. Medical Expertise

Reliable Partner

Crystal-clear reputation
Experienced player on the market (since 2002)
Deep understanding of localization solutions
Localization Engineering and QA departments
$3M-liability insurance
Talented team committed to support client’s mission
ISO 17100 and ISO 9001 certified