Software that generates images on your own computer, instead of on a company’s servers, has one unusual demand: a strong graphics card. A fast processor and a new laptop do not guarantee one. This guide explains the three numbers that decide whether local AI will run on your PC, and how to find them in two minutes.
Why the graphics card does the work
An image model is a very large table of numbers, and making a picture means doing billions of small calculations with that table. A graphics card (GPU) is built to do huge numbers of simple calculations at the same time, which is why it is many times faster at this than the main processor.
The card has its own memory, called video memory or VRAM. The model has to fit inside it while it works. That is why VRAM, not speed, is the first thing to check: a model that does not fit either fails with an out-of-memory error or falls back to much slower methods.
The three numbers
| Part | What it does here | What Influencer Pro asks for |
|---|---|---|
| Video memory (VRAM) | Holds the model while it generates | 8 GB at least, 12 GB recommended, on an NVIDIA card |
| System memory (RAM) | Holds the app, the browser and models being loaded | 16 GB at least, 32 GB recommended |
| Free disk space | Stores the model files, which are large | About 30 GB for Essential, 80 GB for Complete |
The maker matters as well as the size. Most local AI software is written first for NVIDIA cards, and Influencer Pro needs one.
How to check your PC
- Graphics card and VRAM. Press Ctrl+Shift+Esc to open Task Manager, choose Performance, then GPU. The name of the card is at the top right. Lower down, Dedicated GPU memory shows how much is in use and how much there is in total; the total is your VRAM. If you see two GPUs, the one with dedicated memory is the one that counts.
- RAM. In the same window, choose Memory. The total is at the top right.
- Disk. Open File Explorer, choose This PC, and read the free space under each drive.
“Shared GPU memory” in Task Manager is not VRAM. It is ordinary RAM that the graphics system can borrow, and it is far too slow for image models. A PC with only integrated graphics has no dedicated memory to speak of, and will not run this kind of software.
What to expect from your card
These figures come from Influencer Pro’s own manual, which is written around a 12 GB card:
- A photo takes about one to two minutes on a 12 GB card.
- Training a character LoRA, which teaches the model one face, takes about two to four hours on a 12 GB card.
- Voice generation needs 4 to 6 GB of video memory to itself.
- AI-only video clips are short on a 12 GB card: a few seconds, useful as cutaways.
With 8 GB the same things are slower and tighter. Start with the Essential package, generate one image at a time, and pick the smaller model where the app offers one. With less than 8 GB, local image generation of this kind is not practical.
One heavy job at a time
Video memory is shared by everything that uses the card. If the image engine, the voice tool and the caption writer are all loaded, each has less room, and one of them will run out. The fix is simple: stop the engines you are not using before you start a heavy job. Close games and video editors too, since they hold video memory while open.
Laptops
- Check the exact card. Laptop versions of a card often have less memory than the desktop card of the same name.
- Plug it in. On battery, most laptops limit the graphics card heavily.
- Give it air. Long jobs keep the card at full load. Use a hard, flat surface, and expect the fans to run.
Disk space and drive type
Model files are large: single files of several gigabytes are normal, and a full set runs to tens of gigabytes. Two practical points. Install on an SSD if you can, because models load from disk into memory each time an engine starts, and a hard disk makes every start slow. And leave headroom: generated images, videos and training files grow over time, so the minimum figure is where you start, not where you stay.
Keep the driver current
AI software relies on recent features of the graphics driver. If generation fails on a card that should be large enough, update the NVIDIA driver from NVIDIA’s own site or app, then restart the PC before trying again.
If your PC is not enough
You have three honest options: a desktop graphics card upgrade, which is usually the cheapest route to 12 GB; an online service that runs the models for you, where your images pass through someone else’s servers; or waiting, since models that need less memory appear regularly. What does not work is hoping that a strong processor will make up for a small card.
When the numbers check out, the setup guide takes you through the install.