Steply

the problem

Written SOPs need another round of work before they can become training videos.

my part in the build

Prototyped Steply at Polaris Labs.

I combined Python, language-model agents, text-to-speech, and ffmpeg in a prototype for turning procedure documents into narrated training videos.

how it fits together

The main flow, from input to something useful.

1. 01

   Read the procedure

   An SOP or procedure document supplies the steps that need explaining.

2. 02

   Prepare narration

   Language-model agents and text-to-speech turn the written material into a narrated sequence.

3. 03

   Assemble the video

   ffmpeg brings narration and visuals together into the video output.

a few decisions behind the interface

Use the procedure as the input

The prototype starts with an existing SOP, so its workflow centers on converting documented steps into a different format.

Separate narration from assembly

Language-model processing, text-to-speech, and ffmpeg each have a defined role in the pipeline: preparing the explanation, generating audio, and assembling the video.

what changed, and what you can check

The prototype connects document processing, generated narration, and video assembly. The public demo site presents that document-to-video workflow.

The public site illustrates the SOP-to-training-video concept and links to a demo. The implementation stack and my prototype contribution are documented in this portfolio.