https://github.com/real-yfprojects/uelis-words
https://github.com/real-yfprojects/uelis-words
Last synced: about 2 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/real-yfprojects/uelis-words
- Owner: real-yfprojects
- License: mit
- Created: 2023-11-13T10:55:45.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2023-11-27T10:38:54.000Z (over 1 year ago)
- Last Synced: 2025-02-01T07:41:19.559Z (4 months ago)
- Language: Rust
- Size: 1.51 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 2
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# uelis-words
## Install
- `whisper-rs` (deps)
- model from [here](https://huggingface.co/ggerganov/whisper.cpp), saved to `whisper.model` in root dir
- don't use `large` (?), use `large-v2` (install with `make large-v2` in `whisper.cpp`)
- directly use whisper? (more options, e.g. as vtt (detailed word timestamps), initial prompt to capture special stuff (e.g. erms, "special" words))
## Workflow
See installation(s) in `convert.sh`
1. login with credentials with a `POST` request at `https://video.ethz.ch/lectures/d-infk/2023/autumn/252-0025-01L.series-login.json`
2. fetch all lectures at `https://video.ethz.ch/lectures/d-infk/2023/autumn/252-0025-01L.series-metadata.json`
3. get more info for each new lecture at `https://video.ethz.ch/lectures/d-infk/2023/autumn/252-0025-01L/{}.series-metadata.json`
4. download the lowest resolution `.mp4` (`.mp3` quality is the same) to `videos/`
5. transcribe using `large-v2` from `whisper`, `vtt` format, language set to `de`, output to `outputs/`
## Run locally
1. download and install rust (stable should be enough)
2. `. env.sh` (load username & pwd into env vars)
3. `cargo run --release` (output in `videos/*.mp4`)
4. `pip install -U openai-whisper`
5. `bash transcribe.sh` (output in `outputs/*.vtt`, other format?)