Hi DEV!
I write a lot of posts and kept wondering whether a short version was worth the effort. The tools that do it want a subscription, charge per video, and you feed them your text to find out.
ShortGeek does the whole chain locally:
- Point it at an article or an RSS feed
- It drafts a short script from the content
- Narrates it
- Burns in captions, timed to the narration
- Renders a vertical video
No uploads, no credits to run out of, no watermark on the export.
Why I built it
Mostly to find out where the seams are. Text to script is the easy part. Script to speech is a solved problem locally now. The part that took the work was caption timing, because you need word level timestamps out of the speech step to burn captions that actually track the voice, rather than dumping a sentence on screen and hoping.
Tech stack
- Python, fastapi 0.115.6 and uvicorn behind a desktop shell
- jinja2 for the script and caption templating
- ffmpeg for the render and the caption burn in
- Local models for the script and the narration, nothing calls out
The render queue was worth building early. A single short takes long enough that a synchronous UI feels broken, and once there is a queue you can batch a whole feed overnight.
Honest caveat
This is an assistant, not a replacement for editing. The generated script is a first draft and you will want to trim it before you publish anything. Also the installer is not code signed yet, so SmartScreen may warn on first run.
Links
- Site: https://techygeekshome.info/shortgeek/
- Source: https://github.com/techygeekshome/ShortGeek
- Video: https://youtu.be/zFTy2RzFx0M
If you have opinions on caption styling for shorts, I am all ears, that is where the polish is missing.
Top comments (0)