Run locally in Windows?

#7
by dan40 - opened

I love this, and really need it installed locally. I went through what I could in the "Run Locally" instructions and all I got were errors. I was able to clone the repository, and pip run the requirements.txt file, but then it failed. I run Docker, ollama and openwebUI and that works just fine, but this is driving me crazy. I have 40,000+ images to get captions for and uploading all of my images here would seem unfeasible so a local install would be better. Any ideas?

I might add, that I have different ollama models installed and System prompts to help me get the desired results, but it would appear joy-caption does that all on its own and creates mostly accurate descriptions. It only has a problem with identifying right and left if the subject is human. Sometimes it will say "left hand" when it's actually the right hand, and vice-versa.

dan40 changed discussion status to closed
dan40 changed discussion status to open

Hey Dan,

If you want, you can use Captionator.ai. Just create a free no credit card required account, and you can upload all 40,000. It will probably take a couple hours to caption them all, but you can organize into libraries, edit the captions, export in a variety of formats including one text file per image, etc. Technically free accounts are limited to 2500 images per month, but right now we are still in beta and so there is no limit. Abuse to your hearts content.

https://captionator.ai/videos/demo.mp4

Heya Dan,
Another way to use joycaption offline on your device is by using the quantized llama-based joycaption model Here. You can use this with llama.cpp. I think ollama has a model download function builtin, so it should be easy to install.

Let me know if you have any questions regarding this model.

I was able to pull it with ollama, but this is what happened. Did I do it right?
Screenshot 2025-06-16 105500.png

Hey Dan,

If you want, you can use Captionator.ai. Just create a free no credit card required account, and you can upload all 40,000. It will probably take a couple hours to caption them all, but you can organize into libraries, edit the captions, export in a variety of formats including one text file per image, etc. Technically free accounts are limited to 2500 images per month, but right now we are still in beta and so there is no limit. Abuse to your hearts content.

https://captionator.ai/videos/demo.mp4

Thank you so much for this! If I can't get joycaption working, I may fall back on this idea!

I was able to pull it with ollama, but this is what happened. Did I do it right?
Screenshot 2025-06-16 105500.png

Hmmm, not exactly. I'm not sure what the problem is...

Ok, I pulled it with Ollama but how do I get some kind of GUI to upload images? Cant figure it out.

Sign up or log in to comment