Introducing Reel Farmer Paste a YouTube URL, get TikTok/Reel with animated captions in one command. Fully local. > How it works: 1. yt-dlp downloads the video 2. LLM reads the transcript and picks the most clip-worthy moments 3. FFmpeg cuts each clip, silence removed automatically 4. Whisper.cpp runs locally... for word-level timestamps 5. Remotion renders animated captions 6. Combines everything into 1080x1920 reels Every stage checkpoints to SQLite. In case of a crash, it'll pick up exactly where it left off. MiniMax (official) M2.7 wrote the entire code.show more

Code:

bruh literally saw reelfarm and said "imma do that but add -er to the end"

nice on, im going to add it to Which is here by the way

automate it with openclaw, post this slop on ig/tiktok/shorts

@arjunaditya_ paneer

@arjunaditya_ yeah, can't believe i'll get paid to try out the latest AI for free and build side projects.

thanks, i'll use this if i ever need smthg like this for smthg in my life lol

i was building something similar - will be releasing it soon :-

I built a clipper bot using Claude, but I faced an issue when the character in the YT video isn't centered in the frame, the reels look horrible

ask claude to center crop the video and put it in the bottom half of the screen. other styles - fit, fill, stretch, tile, span.

local first is cool until your laptop cries, whats the quality hit vs cloud agents

you show that fake prof that spreads propaganda and division on your video. What a clown.

this is the first time i'm seeing a prof jiang hater🤣

what kind of fuck that video? Subway Surfers game above and video of someone explaining something in below? bruh wtf

the video and clip was chosen by ai

tbh something like brainrot title would have been better

There's a desktop app for that

checkpointing every stage to sqlite is the kind of boring choice that makes the whole thing actually usable

time to reel farm

brilliant choice of video

the video and the clip are randomly chosen by opencode 🤣

ok, one question how does it identify the "picks the most clip-worthy moments"??

llm decides that.

best use of llms

I think we can do the same with PIL as well instead of remotion?

yes, you can.

the sqlite checkpoint recovery is the part nobody's appreciating enough. every video pipeline breaks halfway through and you lose everything. this just picks up where it left off. that's the kind of boring engineering decision that separates tools people actually use from tools people star on github and never open again.

Prof. Jiang FTW

i watch all his videos 😭

Damn , this is so cool 👀

thanks payal

@shydev69 is there a chance this can be made into a web app

this is not intended to be used by humans. set it up locally and let your ai agent make reels. do a manual review before you post.

this is the most brain rot video in existence lol

Is remotion actually good ?

better than adding captions with html canvas. this project also good:

ohhu

Are you using an offline model for transcription and captions, or transcription using the OpenAI Wispr API? And if locally, which model is the best suitable for this task?

this is whisper base running on my macbook pro m4 24 gb. the fans run full speed when it runs.

Crazy krdia ye toh

mythical

I did this write up its al about engineering the solution now.

this is fabulous!

@anonthedev , slayshot xd

NICE man!! I remember once made something using ffmpeg and wisper. you upload a study lecture on it, it extract audio and transcribe it, then generate notes quiz and other things around that lecture !!

damnn, this is cool man

This is insanely clean. Full local pipeline + checkpointing is a huge win.