AI influencer and LoRA glossary

A social media character whose face and photos are generated by AI. On this site that always means an original character, never a copy of a real person, and always disclosed as AI.

How to create an AI influencer in 2026, step by step

Short for AI OnlyFans management: running subscription accounts for AI-generated models. OnlyFans needs a verified human behind every account, so this work mostly happens on platforms like Fanvue that openly allow AI creators with disclosure.

AI OnlyFans in 2026: what's allowed, and why we use Fanvue

Open-source software by Ostris for training LoRAs, with a web interface. It's what our LoRA trainer template runs.

Short product videos in the style of a real customer review, made with an AI character instead of a human creator. Brands buy them for ads and social posts.

Instagram's label for accounts whose main persona is AI-generated. Since 31 August 2026, accounts that feature an AI person without it get less reach.

Instagram's AI-generated profile label: who needs it and how to turn it on

The big image model you generate with or train on top of, like Z-Image Turbo or Krea 2. A LoRA only works with the base model it was trained for.

A short text file next to each training image that describes it. For a character LoRA, captions describe what should stay changeable, like pose, outfit, place and light, and never her face.

How many images do you need to train a character LoRA?

A setting that controls how strictly the model follows your prompt. Fast turbo models usually run at CFG 1.

Keeping one character's face identical across every image and video. It's the difference between a recognisable person and a stream of random faces.

How to keep your AI character's face the same in every photo

A LoRA trained on images of one character, so the base model can draw her face on demand. You call her with her trigger word.

How to keep your AI character's face the same in every photo

Someone who handles fan messages on a subscription platform like Fanvue. Many creators chat themselves at first and hire help on a percentage once the account grows.

A saved copy of a LoRA partway through training, for example at step 2000 or 2500. You compare checkpoints and keep the one that looks most like her without overcooking.

A free, visual tool for running image and video models. You load a workflow of connected blocks and press Run, no code needed.

The temporary disk of a RunPod pod. It's wiped when the pod stops, so anything you want to keep goes on the volume.

The set of images, with captions, that a LoRA learns from. For a character, that's about 50 images of her: different angles, outfits, places and light.

How many images do you need to train a character LoRA?

Saying clearly that a character or a piece of content is AI-generated. Fanvue requires it in the bio, captions or a watermark, and Instagram has a profile label for it.

Are AI influencers legal? The 2026 rules, platform by platform

A subscription platform for creators that allows AI-generated creators, as long as they're clearly disclosed. It's the platform the course builds for.

Fanvue for AI creators: rules, ID check, fees and setup

A file browser and terminal in your web browser. On our RunPod templates it runs on port 8888, and it's where you upload and download files.

An open-weight image model family from Krea, released in two versions: Raw and Turbo, the faster one.

A small add-on file that teaches a base model something new, like one character's face or a style. It loads on top of the model instead of replacing it.

How strongly a LoRA is applied when you generate, usually between 0.8 and 1.0 for a character. Lower is looser, higher is a stronger likeness.

Google's image editing model, available as a paid API. Some optional steps in the course can use it instead of a free open-weight model.

Storage on RunPod that lives on its own and can attach to different pods. It keeps billing while it exists, even when no pod is running.

An AI model whose files you can download and run on your own hardware or rented GPU. Each one comes with its own licence, so check it before you sell work made with it.

A rented GPU computer on RunPod that runs a template. It bills by the hour while it runs.

RunPod for AI images: templates, GPUs and what it costs

Paid content a creator sends in messages or posts, bought for a one-off price on top of the subscription.

The text you give an image model to describe what to make. With a character LoRA, the prompt starts with her trigger word and a short hair-and-eyes line, then describes the shot, outfit, place and light, never her face shape or body.

Free AI influencer prompt generator

An open-weight image editing model from Alibaba's Qwen team. It changes an existing image from a text instruction, like a new angle, outfit or background.

A training setting that sets how much a LoRA can learn. Higher rank holds more detail and makes a bigger file.

A cloud service where you rent GPUs by the hour. It's how you run big image models without owning an expensive PC.

RunPod for AI images: templates, GPUs and what it costs

A saved setup for a RunPod pod: which software it runs, its disks and its ports. Deploying a template gives you a ready-to-use machine in a few minutes.

The method a model uses to turn noise into an image step by step. Workflows come with one set, and you usually leave it alone.

The number that sets the random starting noise for an image. Same prompt, same seed, same settings gives the same image, which is how you compare LoRA checkpoints fairly.

A RunPod GPU that only runs when a request comes in and shuts down when idle. It's what powers our Discord generator and the Telegram bot in video 4.

How many times the trainer updates a LoRA. More steps learn more, up to the point where the LoRA starts copying its dataset.

Deleting a RunPod pod for good, including its container disk and volume. Stopping alone isn't enough to end the bill, because storage keeps charging.

RunPod for AI images: templates, GPUs and what it costs

A made-up word a LoRA is trained to link with your character, like a name only the model knows. You put it at the start of every prompt to call her.

How many images do you need to train a character LoRA?

The memory on a graphics card. Big models need a lot of it, which is why you rent GPUs with 24 GB to 96 GB instead of using a laptop.

A ComfyUI setup saved as a .json file: all the blocks, models and settings for one job. Load it, press Run.

A fast open-weight image model from Alibaba's Tongyi lab. It makes realistic portraits in a few steps, and it's what the free AI Empire Discord generator uses.