Offline support

Ainavirin

I've been a user of Eudaimonia for a while now.

Long story short: for many personal (political/ethical) reasons, I don't want to give money to online AI providers.
But I want to support Eudaimonia, and I was wondering if you would plan to add another type of subscription for local-focused users to enable subscriber-only features (like local intelligent image generation locally and larger galleries).

What would be the amount of VRAM needed to run a decent model for image generation?

Mesh1401

Same here! I second this request.

KairosJS

I fully understand the political motivations to restrict AI to local usage. I need to think about a cheaper "local usage" tier that would give access to subscriber advantages without access to online text and image generation.
Is that what you had in mind?

Regarding the VRAM: using Illustrious within ComfyUI for 1024×1024 images takes between 6–8 GB of VRAM. Running an LLM locally will also take between 6–16 GB of VRAM (obviously depending on the model and quantization). This means that running both at the same time will probably require 24 GB of VRAM.
Local TTS will also need around 2–4 GB of VRAM.

I know that there are solutions to reduce the amount of VRAM needed for ComfyUI and LLMs, but I'm not familiar with them, so I can't really help there.

ElectroLover

Hi ! @Ainavirin @Mesh1401
I have a 5070 and i was able to use the Site with Local TTS / LLM / Image generation !

i use Local TTS From th site
Image generation with 20 step and face enhancer with "babesByStableYogiPony_v60FP16.safetensors"
For LLM i Use : "DisyaMistral-qwq-12b-merge-ggufqwq-12b-test-q5_k_m.gguf" you can find it the net

all seems to work really fine with not lag.
Maybe You can try

Mesh1401

ElectroLover Thanks for your kind suggestion. Yes, it all works fine for me too (24 GB vram). But that was not the (first) point of the request. The request was for a subscription tier for local-focused users to enable subscriber-only features. :-)