How to fill the gaps in your own footage with generated shots that match
Upload the footage you actually filmed and tell Vizard Agent what is missing from it. Vizard Agent studies every scene you shot, pulls style references out of your own frames, generates the shots you never got, checks each one against your footage for likeness, and cuts them together as one piece.
What do you type into Vizard Agent?
Describe the finished video, not the gap. Vizard Agent watches what you uploaded and works out for itself which beats are missing, so a description of the whole piece gets you a better result than a shot list of what to generate. Upload everything, including the takes you think are unusable, because they set the style.
Prompt
What does Vizard Agent actually do?
Your own frames are the thing every generated shot has to match, so almost half the run is spent looking at the footage you already have, scene by scene. Your footage stays the spine and only what is missing gets generated, because mostly real reads as complete and mostly generated reads as generated. In the session behind this article it:
- Checked the project's past files and deliveries, then probed the footage, extracted frames and measured the audio.
- Built a contact sheet of every frame, transcribed and shot-logged the footage, and looked at each scene individually: the opening, the desk, the close-ups, the reaction, the writing.
- Measured the audio per scene and detected the cuts.
- Extracted style references from the user's own frames and searched for music and stock.
- Generated the missing shots alongside the music and a stock timeline, then checked review frames from every generated clip against the real footage for likeness before cutting.
- Delivered 1080x1920 at 30fps, 25.74 seconds long.
When does this not work well?
Generated shots inherit their believability from how close they sit to real ones. A recognisable face in a generated shot is the hard case, while hands, backs and objects are far safer. Unusual lighting needs more correcting. Nothing can generate a real event you missed, and if the result could be mistaken for filmed footage, disclosure is your call.
Common questions
Will the generated shots match my lighting? Vizard Agent pulls style references from your own frames, which is what keeps them close.
Can it generate a shot with me in it? Vizard Agent can, and that is the hardest case. Hands, over-the-shoulder and object shots are far more reliable.
Should I say the video contains generated shots? If it could be taken for entirely filmed footage, yes. Vizard Agent makes the join close, and the disclosure is yours to make.