Getting Started - Common Questions and Troubleshooting

[PurpleDoubleD]

Having trouble getting started? Check the Getting Started Guide first — it covers install, setup, chat, image gen, and video gen with screenshots.

Quick setup checklist:

1. Install Ollama from https://ollama.com (or let the app install it for you)
2. Download Locally Uncensored from Releases
3. Launch the app — the onboarding wizard guides you through everything
4. For image/video gen: the app installs ComfyUI for you in one click

Common issues:

• Ollama not detected: Make sure Ollama is running (ollama serve). The app auto-detects it on port 11434.
• Models not showing: Download models from the Model Manager (grid icon in header).
• ComfyUI not connecting: Go to Settings > ComfyUI. Check the port — the new ComfyUI Desktop App may use a different port than the default 8188. You can change the port and path in Settings.
• ComfyUI Start button not working: Fixed in v2.3.1. Update to the latest release.

Full guide with screenshots: https://locallyuncensored.com/guide/

Ask away and we will help you out!

[Underhero]

I am getting the error: ComfyUI not responding.
The only thing similar to an in-app setup wizard that I can find is when I restore defaults in settings, but it doesn't seem to do anything with the ComfyUI install.

[PurpleDoubleD]

I hear you, thank you for your Feedback. Working on it right now, check in tomorrow evening for an update :)

[PurpleDoubleD]

Hey @Underhero, just a heads up: we released v2.3.0 which completely reworks the ComfyUI setup. It now auto-detects your ComfyUI installation, offers one-click install if it is missing, and auto-starts it when you open the app. The issue you ran into should be resolved.

Would love to hear your feedback if you get a chance to try it: https://github.com/PurpleDoubleD/locally-uncensored/releases/tag/v2.3.0

[jbarkls]

Like Underhero, Comfy is not getting accessed at all. It shows the correct path, so it clearly found the installation; but it simply refuses to connect. And yes Comfy was running with no errors on its end.
Also, I have been looking all over, but it appears there is not even a basic user guide. It would benefit us if there was some level of guide to start with.
I would like to know if there is a way to change default install location, model install location, and such as those issues. I like your goals, like the package. But it is a heavy hitter resource wise, both storage, and disk hits.. thus it would be infinitely better if we could set up this whole package on a dedicated drive so as to not put such weight on OS drives.
Appreciate your work
await any further info you can pass along.

Desktop, W11 always on the latest release pipeline, Ryzen9, 128GB Ram, RTX5080 series

JB

[jbarkls]

OH, also no, running Uncensored does not auto start Comfy, and hitting its start button has no effect.... ollama does start with the start of Uncensored....

[jbarkls]

I believe I have found the problem with comfy not being loaded (with its models)....
The latest comfy has moved from browser interface to a desktop app... After a pause for thought, I started up my comfy portable (still uses the browser interface), and voila, Uncensored saw it (running) and loaded (the models, but only for the desktop comfy; not the portable).
It seems likely that your hooks look for the local port hook of comfy; which the desktop is not broadcasting in the same way as the older browser version; so you can not grab the hook. I did try opening the desktop (while running) in the browser, which was successful; restarted Uncensored, but it again did not see comfy running, nor did it grab the models.
Not sure how you can remedy this issue with comfy, the app still starts a cmd prompt but it is not visible except in the UI itself.
anyway, is there a way to wither add to uncensored's path to comfy, or change it so it sees my portable directories?

JB

[PurpleDoubleD]

Hey JB, thanks for digging into this — you were spot on with the root cause.

We just pushed an update that should fix your setup. Here is what you can do:

1. Grab the latest build from Releases
2. Go to Settings → ComfyUI (Image & Video)
3. You will see a Path field — paste your ComfyUI Portable directory there, hit Connect
4. There is now a Port field too (default 8188) — if your ComfyUI Desktop version runs on a different port, change it there and hit Set
5. Hit Start or Restart and it should pick up

The port was hardcoded in about 20 places across the app which is why the Desktop version would not connect even though it was running fine on your end. All configurable now.

Also the Start button issue you mentioned — that is fixed too, it was tied to the same hardcoded port problem.

Regarding model install paths on a dedicated drive — not in this update yet, but noted. For now pointing the Path to your portable install on the other drive should get the models loading from there.

Let us know if it works on your end.

[jbarkls]

Awesome! I had unloaded completely the grab what I found in discussion about a portable... But found it was pretty much same, with same issues.. Not often we get such responsiveness, so a big KUDOS sent your way. I'll grab the updated install and give it a spin sometime this weekend if I can get DOD dumb ass techs to follow instructions for once... Let you know when I've given it flyby. JB

…

On Sat, Apr 11, 2026, 7:13 AM David ***@***.***> wrote: Hey JB, thanks for digging into this — you were spot on with the root cause. We just pushed an update that should fix your setup. Here is what you can do: 1. Grab the latest build from Releases <https://github.com/PurpleDoubleD/locally-uncensored/releases> 2. Go to *Settings → ComfyUI (Image & Video)* 3. You will see a *Path* field — paste your ComfyUI Portable directory there, hit *Connect* 4. There is now a *Port* field too (default 8188) — if your ComfyUI Desktop version runs on a different port, change it there and hit *Set* 5. Hit *Start* or *Restart* and it should pick up The port was hardcoded in about 20 places across the app which is why the Desktop version would not connect even though it was running fine on your end. All configurable now. Also the Start button issue you mentioned — that is fixed too, it was tied to the same hardcoded port problem. Regarding model install paths on a dedicated drive — not in this update yet, but noted. For now pointing the Path to your portable install on the other drive should get the models loading from there. Let us know if it works on your end. — Reply to this email directly, view it on GitHub <#6 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AHG4XLQXSRI3CHUMVTT475D4VIZGDAVCNFSM6AAAAACXGRQRW6VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTMNJSGY3TIMY> . You are receiving this because you commented.Message ID: <PurpleDoubleD/locally-uncensored/repo-discussions/6/comments/16526743@ github.com>

[tlitookilakin]

I downloaded llama via the model manager, but after the download completed, the model isn't listed in LU. I can see that it is in the LU model folder, it's just not showing up in the UI. Additionally, I had comfyui already installed, and while it was successfully detected by LU, the model I already had installed in it is not being listed. (possibly due to using a non-default model location?)

Also had some additional questions- is it possible to manually install models, and is there an option to select a different model directory? By default it is installing on my C drive which does not have much space on it. I would much rather use my secondary drive, as it is both larger and faster.

[PurpleDoubleD]

hey @tlitookilakin, three clean questions, three answers:

1. llama pulled via model manager, not showing up: usually a race where ollama reports the pull done before the manifest is live on the /api/tags endpoint. LU polls for a while but not forever. workaround: hit refresh on the model dropdown in the header, or restart the app. if it still doesn't show after a restart, paste the output of ollama list from a terminal — that tells us whether it disappeared on the daemon side.

2. existing comfyui install with a custom model path: LU scans the default ComfyUI/models/ folder only. if your models live elsewhere, the easiest fix is a symlink / junction into that folder, or edit extra_model_paths.yaml in your comfyui root so comfyui itself picks them up — then LU picks them up automatically. a setting for an explicit second scan path is on the roadmap but not in yet.

3. moving model directory to a secondary drive: no first-class setting yet but absolutely a wanted feature. quick workarounds on windows:

   • text models (ollama): set OLLAMA_MODELS=D:\whatever\ollama before ollama starts — pulls land there
   • comfyui models: add a second entry in ComfyUI/extra_model_paths.yaml pointing at your D: drive
   • LU app data: sits in %APPDATA%\Locally Uncensored (roaming) — chat history etc., pretty small, less worth moving

manual install works, just drop .gguf / .safetensors files into the right subfolder. ollama manifests are finicky so ollama pull is the easier route for those, but ollama create with a modelfile works for local ggufs.

filing "custom model root in settings" as a real feature request. thanks for the organized ticket.

[tlitookilakin]

paste the output of ollama list from a terminal — that tells us whether it disappeared on the daemon side.

The list is empty- it shows only the list headers and no entries... so it's probably an issue with ollama. Thanks for the tip.

comfyui itself picks them up

I should clarify- the model does show up in comfyui; I configured the path during comfyui's installation and it worked as expected within comfyui. However, after checking the comfyui folder, that yaml file that you specified does not exist, so it's likely storing the path in a configuration file somewhere else that LU is not detecting. I will add that yaml file though, and see if that helps LU find the folder.

text models (ollama): set OLLAMA_MODELS=D:\whatever\ollama before ollama starts — pulls land there

Ah, that was my mistake. I did add that to the env, but I think ollama was minimized to the system tray instead of turned off when I did that. I didn't realize it did that when I closed the window until later, so probably manually moving the model and restarting everything will fix it. Fiddling with the paths while it was running is probably also what screwed up the model manager.

Thank you very much for the detailed and informative response!