Emily Higgins' Five-Tool Workflow for Cinematic AI Videos

Creator @digitalsamaritan breaks down the exact production workflow behind Emily Higgins' AI-generated videos, which consistently pull millions of views thanks to their cinematic visual style. The walkthrough covers five tools chained together in a specific order, from script to final animated output.

The Workflow

The process starts in Claude, where Higgins writes her script and brainstorms which visuals should accompany each section. In the video, @digitalsamaritan holds up a screen showing a Claude conversation with reference images and detailed prompts -- the kind of back-and-forth where you define the concept before touching any image generator.

From there, Higgins moves to Phleem (a visual inspiration tool) rather than jumping straight into image generation. The purpose is to collect cinematic reference material -- mood, framing, color palette -- before committing to any AI-generated output. The frame showing Phleem displays a grid of mixed references: portraits, sketches, diagrams, and film-style compositions that establish the visual language for a piece.

Once she has her references, she brings them back into Claude to build a full storyboard. This is the step where the concept gets structured shot by shot, with the cinematic tone locked in from the Phleem references.

The storyboard then goes into Hailuo (referred to as "Hicksfield" in the audio) to generate multiple image shots that maintain a consistent cinematic theme. The frame at this stage shows a dark-themed generation interface with a grid of output images -- all sharing the same soft, hand-drawn aesthetic with consistent lighting and composition.

Finally, anything that needs motion gets animated, turning static shots into video clips ready for editing.

Why It Works

The workflow is linear and each tool has a single job. Claude handles language and structure. Phleem handles visual direction. Claude again handles storyboarding. Hailuo handles generation. An animation tool handles motion. No single tool is asked to do everything, and the visual research step (Phleem) prevents the generic look that plagues most AI video content.

As @digitalsamaritan puts it: "stupidly simple but a very clever execution."

Key Takeaways

• Write the script in Claude first, then brainstorm visuals in the same conversation before generating anything.
• Use a visual inspiration tool (Phleem) to establish cinematic references instead of going straight to image generators.
• Bring those references back into Claude to build a structured storyboard, shot by shot.
• Generate images in Hailuo using the storyboard to maintain visual consistency across all shots.
• Animate only the shots that need motion as the final step.
• The quality gap in AI video content comes from the pre-production steps (scripting, reference gathering, storyboarding), not from which generator you pick.