TLDR: Following my recent blog post and Hacker News post (https://news.ycombinator.com/item?id=48528029). where I ran the desktop app on my M1 Max. Today, I used Edit Mind (my source-available, local-first video indexing tool) to process my YouTube library, 87 videos, 37h of footage in under 24h on a rented RTX 4090 via Docker + CUDA.
Thank you so much for your support over Hacker News, I wanna share with people of HN. a special discount code "HACKERNEWS", use it here https://shop.edit-mind.com/checkout/buy/9f18a6f0-b437-47ec-b... and get 10% OFF only for the first 5 people and it'll expires
I'm not sure if it's possible or not, the goal of my project was to utilize local models and your local machine instead of uploading your videos to the cloud. there's couple of cloud services that offer video indexing very well, I could name a few like Tweleve Labs (I'm not affiliated with them but I did a presentation at one of their webinar).
You're welcome. As of now, the project narrow down your options instead of watching 3-4 videos (5 min each, as an example), you can watch 3 clips from them instead (1s to 2.5s each). It gives you only the video moments that you need to watch for review instead of watching the full videos.
Thank you! I'm using RAG, I have every video scene indexed individually in the vector database. When I'm asking the agent, it'll use an Ollama model to understand the request, use the available search tool (searching using transcription text, faces, visual, audio or combined) something like when you use Claude or Chat GPT it'll use the web search tool to find you info online. Then, I can filter out video scenes using the Ollama to better present accurate and unique video scene, then send those video results to Davinci Resolve using their API to create a video timeline using those video clips
I agree with that, thank you for your feedback. Also, maybe you're not a video editor and you just wanna search your videos. The video editing integrations are optional and you have full control. You can switch between Adobe Premiere Pro, Final cut Pro or Davinci Resolve
Sure, I'm using (https://huggingface.co/collections/Qwen/qwen25-vl) which can help me understand action like falling down, because I can provide for example 5 frames (down scaled to 720p) to understand what is happening in this part of the video
> Aha, it makes total sense. This number sounds much more reasonable than “669 GB”, since the actual total size of processed frames would be like 10-30 GB.
The reason why is “669 GB” is the total raw footage size when I'm doing the video processing, I downscaled each frame to 720p to make the video processing much faster and I don't need full original quality in order to get accurate results (as far as I know and experiment with).
> I’m just curious tho — is there any paying options that can accelerate this kind of process? Just spin up GPU instances?
For now, I found that NVIDIA GPU for example RTX 3060 with 12GB Vram was much faster than my M1 Max. (still working on optimizing for speed and accuracy).