How to censor swearing in the captions without bleeping the audio
Say you want the captions censored and the audio left alone. Vizard Agent stars out the words in the caption track only, so a viewer scrolling with the sound off sees nothing they would have to explain, while anyone actually listening hears the conversation as it happened.
What do you type into Vizard Agent?
Say clearly which of the two layers gets censored and which doesn't. Vizard Agent will otherwise assume you want both treated the same way, because a bleep normally goes with the asterisks, and the whole point of this job is that the two come apart. Give your own word list if you have one: what a business podcast stars out isn't what a comedy channel does.
Prompt
What does Vizard Agent actually do?
Bleeping the audio changes how the whole conversation sounds and makes a relaxed exchange feel policed. Starring the word in the caption solves the problem you actually have: a screen full of profanity in somebody's feed at work. So the censoring happens in the caption generation, nowhere near the audio. In the session behind this article it:
- Transcribed a fifty-minute conversation with word-level timings, checked the client's previous renders for the house style, and found the mandatory sections the brief named.
- Fingerprinted the camera setups, calibrated a centre crop per speaker, and used speaker dominance to crop the wide two-shots onto whoever was talking.
- Built a 1080x1920 vertical layout with a 250-pixel header band for the headline above the video block, and generated kinetic captions mid-screen.
- Applied the profanity list to the caption text only, replacing the letters with asterisks, while the speech audio went through the whole pipeline untouched.
- Normalised each clip's speech to a broadcast loudness target with smooth fades, rendered the clips and a separate promotional cut, and reviewed frames for framing, captions and the header band.
When does this not work well?
Censoring a caption is a judgement about your own audience, and no setting can make it for you. Vizard Agent applies the list you supply precisely, so the list is the whole decision. Sound-on viewers still hear every word, some platforms care about the audio rather than the text, and a misheard word can be starred wrongly.
Common questions
Why not just bleep it? Because a bleep changes the tone of the conversation. Starring the caption fixes the screen without touching the sound.
Who decides the list? You do. Vizard Agent applies exactly what you supply.
Can it bleep the audio too? Yes, as a separate pass. Vizard Agent keeps the two operations independent.