Nov 2024 – Jan 2025
The hand-off
December 2024: one image plus a prompt becomes a cast of references plus a prompt.
In one month at the end of 2024, three tools started accepting several reference images at once: a face, an outfit, a location. Instead of hoping a character looked the same in the next shot, you could show the model exactly who and what you meant.
Tutorials1
Runway Frames and Kling Elements
Theoretically Media · YouTube · Jan 2025
How it worked
- Kling 1.6 shipped with first and last frame conditioning and, through Elements, up to four subject images (people, animals, objects, scenes) combined with a prompt; the global rollout landed in January 2025.
- Pika 2.0 Scene Ingredients let you upload as many reference images as you like for character, object, wardrobe and setting, then prompt the shot.
- Runway Frames was the pre-echo: an image model built around consistent worlds feeding Gen-3 image-to-video.
- The consistency toolkit underneath (DreamBooth, LoRA, IP-Adapter, ComfyUI graphs) is what artists reached for whenever the hosted model could not hold a face.
Example Films2
Related on the timeline2