How to generate a two-shot cinematic scene from character reference images
Give Vizard Agent images of your characters and describe the scene beat by beat. Vizard Agent crops close-ups to study each character properly, generates the first shot, then extracts its final frame and uses it as the opening frame of the second, so the two clips read as one continuous piece rather than as a cut between two guesses.
What do you type into Vizard Agent?
Write it like a scene description rather than a prompt, in order: what happens, then what changes. The detail that pays off most is the physical specifics, such as the posture, the direction of the gaze and exactly when the weather changes, because those are the things Vizard Agent can render and check.
Prompt
What does Vizard Agent actually do?
Generated video is good for a few seconds at a time and loses the thread beyond that. The way around it is not a longer generation but two shots, where the second starts from the exact frame the first ended on. Continuity between generated shots cannot be asked for in words, so it is handed over as a frame. In the session behind this article it:
- Extracted frames from the uploaded material every few seconds and tiled them into a grid to identify the characters.
- Extracted high-resolution frames, cropped the character regions and looked at them close up, one at a time.
- Checked the generation tool's options, uploaded the character references and generated the first shot, then tiled its test frames to review composition and quality.
- Took the last frame of the first shot, uploaded it and generated the second shot from it.
- Reviewed the second shot's frames for continuity, joined the two losslessly and mixed music under the generated rain.
- Delivered a 720x1280, H.264, 24fps file of 10.24 seconds with AAC audio, then analysed it to check the motion, the join and the sync of picture and sound.
When does this not work well?
Chaining shots holds continuity far better than one long generation, but not perfectly. Faces drift between shots, and close-ups show it first. Hands and props are the weak point: an umbrella held in shot one may be held differently in shot two. And the turn has to be visual. "She lifts her head" can be generated, while "he realises she is sad" cannot.
Common questions
How many reference images do I need? One clear image per character. Vizard Agent crops close-ups from them to study the details before generating anything.
Can it do more than two shots? Yes. Each new shot starts from the last frame of the previous one, so Vizard Agent can extend the chain as far as the scene needs.
How long can each shot be? About five seconds each holds up well. Longer generations drift further from the reference.