I've been playing around with fabric: https://github.com/danielmiessler/fabric
for a little while now, and it's awesome. Since the underlying software got converted to go, I decided to write my own little web UI on top of it for summarizing and keeping track of youtube videos.
Here's a little demo - You can paste in a youtube video, and it will go out and get the transcript for it.
You can then go to the video page and select and model and a pattern to use on the video - in this example I'm using the gpt-4o model and the summary type. However you can have this run against a local LLM as well (basically anything fabric supports).
My workflow here can now be that I'll just grab youtube videos that _might_ interest me. I can then get summarizes, extract_wisdom, or rate the video. Once I read through those, I can decide if the video is actually worth my time, or if I'm fine with skipping it.
Ideally I'd release this as a binary that anyone can run on their computer. Your own personal youtube time saver!
#ai #fabric #go #youtube #savetime
Erik
npub1ra…34atg
2024-09-23 13:26:05
Author Public Key
npub1rach342hnt3wa0qfqk0pey0exs2drcxfc0azs8eurwnee9vu64vs834atgPublished at
2024-09-23 13:26:05Event JSON
{
"id": "0344b121c65c98a50ca639444da0e3e2e0d41636387c02bfd74aed4bacc31dfd",
"pubkey": "1f7178d5579ae2eebc09059e1c91f93414d1e0c9c3fa281f3c1ba79c959cd559",
"created_at": 1727097965,
"kind": 1,
"tags": [
[
"imeta",
"url https://video.nostr.build/45f496b6ee251d7e4eae5e2608f712cf8a2a621fb71269a649d48d6462bfd2d9.mp4",
"m video/mp4",
"x 564431b7b03458ef2bdb188a57a24d188a6ba84d759aa26199407171a99bbae8",
"ox 45f496b6ee251d7e4eae5e2608f712cf8a2a621fb71269a649d48d6462bfd2d9",
"size 18641439"
],
[
"r",
"https://github.com/danielmiessler/fabric"
],
[
"t",
"ai"
],
[
"t",
"fabric"
],
[
"t",
"go"
],
[
"t",
"youtube"
],
[
"t",
"savetime"
]
],
"content": "I've been playing around with fabric: https://github.com/danielmiessler/fabric\nfor a little while now, and it's awesome. Since the underlying software got converted to go, I decided to write my own little web UI on top of it for summarizing and keeping track of youtube videos.\nHere's a little demo - You can paste in a youtube video, and it will go out and get the transcript for it.\nYou can then go to the video page and select and model and a pattern to use on the video - in this example I'm using the gpt-4o model and the summary type. However you can have this run against a local LLM as well (basically anything fabric supports).\n\nMy workflow here can now be that I'll just grab youtube videos that _might_ interest me. I can then get summarizes, extract_wisdom, or rate the video. Once I read through those, I can decide if the video is actually worth my time, or if I'm fine with skipping it.\n\nIdeally I'd release this as a binary that anyone can run on their computer. Your own personal youtube time saver!\n\n#ai #fabric #go #youtube #savetime\n\nhttps://video.nostr.build/45f496b6ee251d7e4eae5e2608f712cf8a2a621fb71269a649d48d6462bfd2d9.mp4",
"sig": "4dc66942703f0219f6d9509986276c8524828abdcdcd4d61beb9341e80331fd5c6cf9e5e38f4d6ccdade53f04b1c49728821e2a6335b5fb158e13bb10d2f990b"
}