How do you make AI influencer videos?

To make AI influencer videos, you start from a still image of your character and animate it with one of 3 tools: Kling (rented; the course rates it fastest and highest quality), Wan 2.2 (open-weight, runs on a rented GPU) or LTX (cheap batches, lower quality). The character comes first: create her and lock her face before you make a single clip, or every video shows a slightly different woman.

What do you need before you make videos?

Three things: a character whose face never changes, good still images of her, and source clips you’re allowed to use. The stills come from her character LoRA. Video models animate what you give them, so a drifting face in the still becomes a drifting face in the clip.

1. A locked face. One original character, trained into a LoRA. Our consistent AI character guide covers how.
2. Stills. Images of her in the outfit and place you want the video to show, made with the LoRA in the course prompt format.
3. Source clips you have rights to, if you want to copy a movement: a clip you filmed of yourself, or licensed stock.
4. A way to pay. Credits for Kling, or GPU hours on RunPod for Wan 2.2 and LTX.

Which video tool should you use?

The course uses a simple decision line. In the course’s view Kling is the fastest and gives the highest quality, but you rent it and pay per second. Wan 2.2 is an open-weight model you run yourself on a rented GPU. LTX makes cheap batches at lower quality. Pick by budget and volume, not hype.

The quality ranking is the course’s view, not a benchmark. Speed, setup, versions and costs side by side are in our Kling vs Wan vs LTX comparison.

How do you turn a trend clip into a video of her?

Rebuild the first frame of a clip with your character in an image-editing model, then let Kling’s motion control copy the movement from the clip onto that frame. Use only clips you filmed of yourself or licensed stock: never a real creator’s video or anyone else’s clip. That’s the course’s method for dances and trends.

The steps, the course’s exact prompt, Kling’s clip rules and the credit costs are in our Kling motion control guide. For dance trends specifically, see how to make an AI dance video.

How do you make videos on your own rented GPU?

Run Wan 2.2 or LTX on a RunPod pod. You pay for GPU time instead of credits, and you control the model, the workflow and the files. The trade-off is setup time: pods take a while to load, and you need enough VRAM for the model you pick.

Wan 2.2. The newest Wan with downloadable weights (Alibaba’s Wan-AI page on Hugging Face, checked 2 Oct 2026). The course generates on an RTX 5090 and trains video LoRAs on an RTX PRO 6000.

LTX. Lightricks’ open-weight LTX-2 family; the newest weights on its Hugging Face page are LTX-2.5 (checked 2 Oct 2026). From the course’s runs: the pod takes 20 to 40 minutes to load and the first generation about 10 minutes, so you batch about 20 source clips per session instead of making one at a time. If the workflow has two prompts, describe the outfit with the same words in both.

Does her image LoRA work for video?

No. A LoRA only works with the model it was trained for, and image and video models are different models. For Wan 2.2 you train a separate video LoRA of her; the course’s came out at about 300 MB and was trained on an RTX PRO 6000. Kling works from images instead.

Kling’s motion control guide takes a still of her plus the clip, and on VIDEO 3.0 you can bind a face Element made from her images (checked 2 Oct 2026). Her look comes from those images, so the image LoRA’s job in video is upstream: it makes the stills, rebuilt frames and close-ups the video tools start from.

What videos should an AI influencer make?

Short vertical clips that match what real creators post this week. The course’s trend-chasing Instagram account runs on dance trends, POV, get-ready-with-me, day-in-the-life and “rating my outfits”. Keep fast trend content on that account and the polished reels on her main account, one quality reel every 2 days.

Keep her clothes casual for growth content. The course’s rule for a new main account is to start with a dress, not swimwear.

Do you have to label AI influencer videos?

Yes. TikTok’s community guidelines require you to label AI-generated content that shows realistic-looking people or scenes, with a caption, sticker, watermark or its AI-generated label; unlabelled content can be removed or restricted (guidelines effective 24 Sep 2026). On Instagram, turn on the AI-generated profile label.

TikTok’s rules are on its integrity and authenticity page. For Instagram, see our guide to the AI-generated profile label. The character is always presented as AI, in every video.

How do videos make money?

Two ways in the course. Videos grow her audience on Instagram and TikTok, which feeds subscriptions on Fanvue. And short product videos, AI UGC, can be sold to brands: the course makes three 8-second clips, about 24 seconds in total, then assembles them in CapCut.

The course prices beginner UGC at $50 to $150 a video (as of October 2026). The whole picture, with costs and what pays first, is in our guide to making money with an AI influencer.

What can go wrong?

• Her face changes in the video. The still was off, or the character was never locked. Fix the image side first.
• Kling’s result is short or broken. The source clip has cuts, camera moves or very fast motion. Kling’s guide says fast, complex moves can shorten the output.
• The body proportions look wrong. The image and the clip don’t match: a half-body image with a full-body clip. Rebuild the first frame of the clip instead of using a random portrait.
• The RunPod bill is bigger than expected. LTX pods take 20 to 40 minutes to load, and that time is billed. Batch your clips, then stop and terminate the pod.
• You used a real creator’s video. That’s their work and their likeness. Delete it and film your own.
• The post gets restricted. Realistic AI video without a label breaks TikTok’s rules. Label every clip.