Open Shorts
An open-source, local-first alternative to Agent Opus / OpusClip. A general-purpose “text to podcast/video short” pipeline that anyone can self-host. Built on the Jamie Pine stack (Voicebox + SpaceUI) with Ollama for LLM work.
What Agent Opus charges for
Agent Opus (opus.pro) is a cloud-based AI video agent that converts ideas, text, links, audio, or blogs into on-brand videos. It combines scriptwriting, storyboarding, asset collection, motion graphics, voiceover, and editing from a single prompt. Paid, closed-source, per-credit billing, data goes to their servers.
What Open Shorts provides instead
The same pipeline, self-hosted. No per-video credits, no subscription, no data leaving your machine. Running on Blackthorn or any hardware with Ollama and Voicebox.
Pipeline:
- Input — a topic, a blog post, a transcript, a script, raw notes
- Script generation — Ollama model produces a structured script (monologue, dialogue, interview format, explainer)
- Segment identification — key moments, hooks, natural break points for short-form clips
- Voice rendering — Voicebox REST API, with voice profiles per speaker, expressive tags, effects
- Assembly — multi-track audio composition, optional visual layer (motion graphics, captions)
- Output — full episode + individual Shorts, with VTT captions, ready for distribution
Relationship to Lugh
What Lugh teaches about building this pipeline transfers directly:
- Stage 3 (script generation) → general script generation for any content type
- Stage 4 (TTS via Voicebox) → voice rendering with multi-speaker support
- Shorts Mode → segment identification and Short extraction from longer content
- Episode Anatomy → structural patterns for engaging audio (cold open, segments, wrap-up)
Lugh is the educational application. Open Shorts is the generalized tool. Same infrastructure, different purpose.
Technology stack
| Component | Tool | Role |
|---|---|---|
| LLM | Ollama (local models) | Script generation, segment identification, key moment extraction |
| Voice | Voicebox (REST API) | TTS, voice cloning, multi-track, effects, expressive tags |
| UI | SpaceUI components | Frontend for the creation workflow |
| Orchestration | Spacebot Workers (or standalone) | Multi-step pipeline execution |
Contribution back to the ecosystem
This is the first “content production pipeline” that ties Voicebox and SpaceUI together into an end-to-end workflow. Jamie Pine built the instruments; Open Shorts builds the recording studio. The project would live as a standalone open-source tool that happens to use the same stack as Guildhall.
Status
Idea stage. Name is placeholder. Implementation depends on Lugh’s pipeline being proven first — the educational use case validates the patterns before generalizing them.