Image obtained through Pasadena, TX Police Department Public Records
AI TOOLS
‘CANDYMAN’ is the pilot episode for an anthology true-crime series that centers around stories that are almost too crazy to believe. Through obtaining various public records (evidence, images, videos, etc.) we create the world of the story. First writing the script, then developing the visual style. Once that style is locked in, we train character models, shot-list the script, and generate the imagery for storyboards - which will be used for the image-to-video input.
The series will contain all kinds of stories that are each told through their own visual style which leverages AI generative tools — think ‘Love, Death, & Robots’ for true-crime mysteries.
candyman
Episodic Web Series (Various Animation Styles)
Documentary / Crime Drama
‘CANDYMAN’ CLIP
CHARACTER MODEL
Taking the archival photos we had of Ronald O’Bryan, we trained a character model that would align with our animation style - in practice: turning the real Ronald O’Bryan into a animated, caricature of himself.
Once we find a look that feels right, we create a “character bible”: polished sheets with front/¾/profile turns and a full expression range. Each image is up-res’d and hand-enhanced for continuity—skin detail, hair edges, fabric grain—so the dataset teaches the right cues. We then fine-tune a LoRA on this curated set to lock visual consistency across new prompts, lenses, and scenes. The result is a production-ready digital actor who stays on-model in stills and video.
LOCATION DESIGN
Once we have our character, next step is to place him in the scene. To do so, we determine the visual style, lighting style, and overall architectural style of the set. Then, we design the locations accordingly. For ‘CANDYMAN’ we wanted to stay true to the 1970’s Texas architecture and feel. As you can see, we test out various options for locations, seeing which works best with character placement, lighting, and other quality assurance parameters.
