A medium shot in filmmaking frames the subject from the waist up, showing more detail than a long shot but still keeping some background visible. It's often used to focus on characters while they are talking or interacting with their surroundings. This shot type helps balance between showing the subject’s emotions and some of the environment, making it a versatile and common shot in movies.
When you use this LoRA trained on movie trailers to generate a medium shot, it will apply the stylistic choices from those trailers. This could include focusing on character expressions, body language, and how they interact with the space around them. The model has likely learned how to balance framing the character while maintaining some background details, giving the generated image that polished, cinematic look typical of medium shots in trailers.
Extensive 7000+ movie trailers dataset.
Computer vision was used to identify the extreme close up shots on movie trailers I used python to extract and tag images from trailers frames for LORA training.
Data set source:
https://paperswithcode.com/dataset/movieshots#:~:text=Subject%20Centric%20Lens-,MovieShots%20is%20a%20dataset%20to%20facilitate%20the%20shot%20type%20analysis,and%20movement%20types%20of%20shot.
Check point: stable diffusion XL
Trigger word: include (long-shot) in your prompt