YeOldeCut / README.md
atlasmoth's picture
Add launch post to README
68b9926
|
Raw
History Blame Contribute Delete
1.87 kB
---
title: YeOldeCut
emoji: 🎬
colorFrom: red
colorTo: yellow
sdk: gradio
sdk_version: "6.18.0"
app_file: app.py
python_version: "3.12"
pinned: false
suggested_hardware: a10g-small
short_description: Turn rejection texts into dramatic medieval roast films.
tags:
- build-small-hackathon
- gradio
- text-to-video
- open-weights
- minicpm
- track:wood
- sponsor:openai
- sponsor:openbmb
- achievement:offgrid
- achievement:offbrand
- achievement:demo
---
# YeOldeCut
Turn painful rejection texts into dramatic medieval roast films.
YeOldeCut takes a breakup, job rejection, investor pass, ghosting message, or screenshot and turns it into a short MP4: faux-medieval script, generated scene art, narrated voiceover, background score, subtitles, title card, final card, and a share caption.
## Demo Video
<video src="https://res.cloudinary.com/dzhtwka6d/video/upload/v1781563099/YeOldeCut_Demo_q1pv5e.mp4" controls width="100%"></video>
[Watch the demo video](https://res.cloudinary.com/dzhtwka6d/video/upload/v1781563099/YeOldeCut_Demo_q1pv5e.mp4)
## Launch Post
[Read the launch post on X](https://x.com/c__osuji/status/2066660708855497177)
## Build Small Positioning
- Track: Thousand Token Wood
- Core output: downloadable MP4 for demos and social sharing
- Small/open model stack: Gemma director model, FLUX.2 Klein 4B scene generation, OpenBMB VoxCPM2 narration from the MiniCPM family, and MusicGen background score
- Sponsor hooks: OpenBMB/MiniCPM-family narration, Black Forest Labs image generation, and Codex-attributed development for the OpenAI track
- Off-grid note: the app does not call hosted model APIs at runtime; generation runs through downloaded open-weight models in the Space process.
## Local Development
```bash
uv sync
uv run python app.py
```
The app writes generated artifacts under `outputs/<run-id>/`.