AQC One Year On: Proven Accuracy, Now Tailored to You
Proven by every customer who tried it, and now customizable with your own guidelines.
One year ago, we launched Automated Quality Control (AQC) by MediaCAT AI. Today, 100% of customers who have used AQC confirm that it delivers higher translation accuracy than subtitles produced without it, and every one of them continues to use it.
When we first introduced AQC, we described it as an intelligent layer on top of MediaCAT's domain-trained machine translation. A year of real production work has turned that promise into a proven result.
And AQC has grown. With the new Advanced mode, you can now give MediaCAT AI your own instructions in plain language, so every QC pass follows your style guide, your client's rules, and your audience's expectations.
The problem: "correct" is not the same as "ready to air"
Modern machine translation rarely gets the words completely wrong. The real cost of subtitle localization sits in the gap between a correct translation and a subtitle a viewer never notices.
That gap shows up in the same places on almost every project:
Cultural blind spots. Idioms, greetings, and local references translated word for word lose their meaning.
Stiff phrasing. Lines that are technically accurate but read like a textbook, not like people talking.
Inconsistency. A character who speaks casually in episode 1 suddenly sounds formal in episode 3, or a name is spelled two ways.
Poor readability. Awkward line breaks and lines too long to read in time.
Client-specific rules. Every distributor and platform has its own style guide, and keeping track of all of them by hand is slow and error-prone.
Human QCers fix these issues one line at a time. AQC fixes them automatically before the file ever reaches a reviewer.
What AQC does better, with examples
AQC reads each line in the context of the scene, the characters, and the whole file, then refines the MT output. Here is what that looks like in practice.
1. It localizes meaning, not just words
A literal translation can be grammatically perfect and still confuse the viewer. AQC recognizes idioms and cultural expressions and finds the equivalent the audience actually uses.
2. It catches what "correct" MT misses
Standard MT translates one sentence at a time, so it guesses whenever the answer lives in a previous line. AQC reads the surrounding dialogue and fills in what the sentence alone doesn't say.
3. It keeps characters consistent
In Spanish, the choice between the formal 'usted' and the casual 'tú' defines how two characters relate to each other. AQC follows that relationship across the whole file, so close friends don't suddenly start using formal speech.
4. It makes subtitles easier to read
AQC places line breaks where the sentence naturally pauses, and restructures the sentence when a clean break isn't possible.
5. It respects your glossaries
Character names, brand names, and key terms from your MediaCAT glossaries are applied consistently in every AQC pass, so refinement never comes at the cost of terminology.
New: Tell MediaCAT AI exactly how you want your QC done
AQC now comes in two modes. Recommended applies MediaCAT AI's proven QC out of the box. Advanced adds an Extra guidelines field where you write your own instructions in plain language. No templates or technical setup are needed.
Think of it as briefing a senior QCer before they start a project. Whatever you would tell that person, you can now tell AQC:
Keep character and brand names in their original spelling.
Use an ellipsis (…) when a sentence continues into the next segment.
This is a children's animation. Keep the vocabulary simple and avoid mild profanity.
The audience is in the UK. Use British spelling and vocabulary.
The two lead characters use casual speech with each other throughout the series.
Write all numbers under ten as words.
Each guideline is applied to every line of the file, together with your glossaries. The result is a QC pass shaped around your client's style guide, your platform's delivery specs, and the tone of your content, not a one-size-fits-all edit.
For localization teams handling many clients, this means one tool can follow many different rulebooks, and reviewers spend less time enforcing rules and more time on creative judgment.
Try AQC on your next project
Getting started takes less than a minute:
Open your project in MediaCAT and turn on Perform AI Quality Control by MediaCAT (AQC).
Choose Recommended for MediaCAT AI's default QC, or Advanced to add your own Extra guidelines.
Run AQC and compare the original MT output with the AQC version side by side.
Every customer who has tried AQC has seen the difference, and every one of them has kept using it. See what it does for your content.
