I Open-Sourced TopicMiner
Turn One Long Video Into Dozens of Branded Shorts
Most video repurposing tools are either too shallow, too closed, or too annoying to customize.
So I open-sourced TopicMiner.
TopicMiner is built for a simple but painful workflow: you have a podcast, webinar, interview, or long-form video, and you want to turn it into short, branded, vertical clips without manually chopping everything up for hours.
It handles:
- local transcription
- VTT caption generation
- agentic video editing
- branded short-form clip creation
- Stripe payments
It’s built on Backboard.io, which means it’s not just a one-off toy project — it’s part of a bigger architecture for building serious AI-native apps.
What TopicMiner does
TopicMiner helps turn long-form content into short-form assets that are actually usable for distribution.
Think:
- podcast to TikTok clips
- webinar to YouTube Shorts
- interview to branded social snippets
- long videos to multiple cutdowns with captions
The goal wasn’t “AI for the sake of AI.”
The goal was to make content repurposing feel less like punishment.
Why I open sourced it
Because a lot of people want to build AI video workflows, but most examples out there are:
- thin wrappers
- black boxes
- impossible to self-host
- hard to extend
- not wired for real product flows like billing
I wanted to put out something more real.
TopicMiner shows how to build an actual app around AI workflows — not just a demo, but a product-shaped system with media processing, transcription, editing logic, and payments.
What’s under the hood
At a high level, TopicMiner includes:
- local-first transcription workflows
- caption generation with VTT
- agent-driven editing flow
- branding-oriented short clip pipeline
- Stripe integration for payments
If you’re building in the intersection of:
- AI agents
- creator tools
- media workflows
- transcription pipelines
- OSS SaaS
…this repo should be useful.
Why this matters
Short-form distribution is one of the biggest leverage points in content right now.
But the workflow is still painfully manual.
A lot of creators and teams are sitting on hours of valuable long-form content that never gets repackaged because the process is too slow. TopicMiner is my take on closing that gap with an open-source stack you can actually inspect and build on.
Built on Backboard.io
TopicMiner is one example of what that stack can support: AI-assisted workflows, product logic, and real app infrastructure in the same system.
Repo
If you want to check it out, fork it, or break it:
Top comments (0)