This project demonstrates a fully automated, end-to-end AI video production pipeline — from character creation to final animated output — using a custom Python script to orchestrate multiple AI platforms without manual intervention.
The Challenge
Creating animated AI characters at scale through browser-based tools is slow, repetitive, and difficult to replicate consistently. Each character requires the same sequence of steps: generate an image, produce matching audio, and animate the result with lip-sync. Doing this manually for multiple characters is time-consuming and error-prone. The goal was to automate the entire back half of that pipeline via API.
The Stack
Sora — AI-generated character imagery
Vocloner — AI voice cloning and audio generation
Hedra API + Python — programmatic lip-sync animation and video generation
Creating animated AI characters at scale through browser-based tools is slow, repetitive, and difficult to replicate consistently. Each character requires the same sequence of steps: generate an image, produce matching audio, and animate the result with lip-sync. Doing this manually for multiple characters is time-consuming and error-prone. The goal was to automate the entire back half of that pipeline via API.
The Stack
Sora — AI-generated character imagery
Vocloner — AI voice cloning and audio generation
Hedra API + Python — programmatic lip-sync animation and video generation
The Process
Character images were created in Sora and voice audio generated in Vocloner. A custom Python script then handled the rest: uploading assets to the Hedra API, submitting generation requests, polling for completion, and downloading finished video files — all without touching a browser.
The Problem I Solved
Midway through production, the script stopped working. Systematic debugging revealed that Hedra had silently updated their API to require internal UUIDs for model identification rather than human-readable model names — a breaking change with no notification. By querying the API's models endpoint programmatically, I identified the correct UUID and updated the script, restoring full functionality.
The Result
A reliable, repeatable Python-based production pipeline that can generate new AI character videos by simply swapping two filenames. What previously required manual steps across multiple browser tabs now runs from a single Terminal command.
Character images were created in Sora and voice audio generated in Vocloner. A custom Python script then handled the rest: uploading assets to the Hedra API, submitting generation requests, polling for completion, and downloading finished video files — all without touching a browser.
The Problem I Solved
Midway through production, the script stopped working. Systematic debugging revealed that Hedra had silently updated their API to require internal UUIDs for model identification rather than human-readable model names — a breaking change with no notification. By querying the API's models endpoint programmatically, I identified the correct UUID and updated the script, restoring full functionality.
The Result
A reliable, repeatable Python-based production pipeline that can generate new AI character videos by simply swapping two filenames. What previously required manual steps across multiple browser tabs now runs from a single Terminal command.
Star Trek and all related characters, names, and properties are owned by Paramount Global/CBS Studios. This fan-created work is made for entertainment purposes only, with no copyright infringement intended. No profit is being made from the use of these characters or settings.
