Spaces:
Sleeping
Sleeping
title: Digest Everything Gpt | |
emoji: 💻 | |
colorFrom: green | |
colorTo: green | |
sdk: gradio | |
sdk_version: 3.24.1 | |
app_file: main.py | |
pinned: true | |
license: mit | |
# DigestEverythingGPT | |
DigestEverythingGPT provides world-class content summarization/query tool that leverages ChatGPT/LLMs to help users | |
quickly understand essential information from various forms of content, such as podcasts, YouTube videos, and PDF | |
documents. | |
The prompt engineering is **chained and tuned** so that is result is of high quality and fast. It is not a simple single | |
query and response tool. | |
Please leave me a star🌟 if you like this project! | |
# Live website | |
https://huggingface.co/spaces/michaelthwan/digest-everything-gpt | |
# Showcases | |
**Example of summary** | |
- "OpenAssistant RELEASED! The world's best open-source Chat AI!" (https://www.youtube.com/watch?v=ddG2fM9i4Kk) | |
![final_full_summary](/img/final_full_summary.png) | |
**DigestEverythingGPT's final output will adopt to video type.** | |
- For example, for the video "17 cheap purchases that save me | |
time" (https://www.youtube.com/watch?v=f7Lfukf0IKY&t=3s&ab_channel=AliAbdaal) | |
- it shown the summary with and specific 17 things correctly. | |
![n_things_example](/img/n_things_example.png) | |
**LLM Loading in progress screen - chained prompt engineering, batched inference, etc.** | |
![in_process](/img/in_process.png) | |
**Support for multiple languages** regardless of video language | |
![multi_language](/img/multi_language.png) | |
More showcases: [SHOWCASE.md](SHOWCASE.md) | |
# Features | |
- **Content Summarization**: | |
- Automatically generate concise summaries of various types of content, allowing users to save time and make | |
informed decisions for in-depth engagement. | |
- Chained/Batched/Advanced prompt engineering for great quality/faster results. | |
- **Cross-Medium Support**: | |
- DigestEverythingGPT is designed to work with a wide range of content mediums. | |
- Currently, the tool supports | |
- YouTube videos [beta] | |
- podcasts (in progress) | |
- PDF documents (in progress) | |
- **Interactive "Ask" Feature** (in progress): | |
- Users can pose questions to the tool and receive answers extracted from specific sections within the full content. | |
# Installation | |
Use python 3.10+ (tested in 3.10.8). Install using requirement.txt then launch gradio UI using main.py | |
``` | |
pip install -r requirements.txt | |
python main.py | |
``` | |
# License | |
DigestEverything-GPT is licensed under the MIT License. | |
# Acknowledgements | |
- chatgpt_academic for gradio code framework | |