Runway positions Gen-4 around keeping recognizable subjects across separately generated scenes. Its product description says users can combine reference images with instructions to carry characters, objects and locations into different settings and camera views, without additional model training. For a sequence of shots, the intended benefit is continuity: a reference supplies the visual identity while the prompt specifies a new composition. This describes the offered workflow; the page’s promotional examples do not establish how reliably every subject or scene will remain consistent.
Terms in this piece · Glossary
fine-tuning — Taking a trained model and training it a bit more on your own examples so it gets better at one specific job.
