03:03
AI video generation
The video claims that anyone can create an AI character-swap video in three easy steps using Higgsfield AI's Wan 2.2 'replace' model, needing only a self-recorded template video and a character image (which can itself be built via ChatGPT from a screenshot of that video).
A subscription to Higgsfield (or another platform offering Wan 2.2) is required before starting.
Higgsfield pricing: $9/month basic, $29/month pro plan, or a cheaper-per-month annual plan paid upfront for the year.
The process is presented as three easy steps: go to higsfield.ai, upload your image, add a template video, and become the character.
Record a template video of yourself on your phone, ideally on a tripod, against a solid, non-distracting background (a painted wall, woods, etc.), avoiding excessive movement or camera shake.
Take a screenshot from the template video, ideally one that shows your hands.
Upload that screenshot to ChatGPT and prompt it to alter the character (the video's example prompt: 'make this character older and he is wearing a old style tunic natural colored shirt').
Upload both the ChatGPT-generated character image and the template video into Higgsfield, then select the Wan 2.2 model.
Choose 'replace' mode to swap yourself with the character while keeping your own filmed background; 'animate' mode instead keeps the character image's own background and animates the character to your movements, and is said to work less well than replace.
Set resolution to 720p and generate; the demonstrated generation cost 57 credits.
The video closes by directing viewers to a description link for Higgsfield.
Higgsfield AI subscription — A paid AI platform, starting at $9/month or $29/month for the pro plan (with a discounted annual option), that hosts the Wan 2.2 model used to generate character-swap videos. Apply: Sign up for a Higgsfield plan (or another platform offering Wan 2.2) before starting, choosing monthly or annual billing based on budget.
Wan 2.2 (generator model) — The underlying AI video generator on Higgsfield described as 'the platform, the generator that allows you to replace yourself.'. Apply: After uploading your character image and template video, select the Wan 2.2 model and then choose a mode (replace or animate) before generating.
Wan 2.2 Replace mode — A Wan 2.2 mode that swaps the person in your template video with the uploaded character image while keeping your original filmed background. Apply: Choose 'replace' when you want the character to appear in your own recorded setting, as demonstrated in the video.
Wan 2.2 Animate mode — A Wan 2.2 mode that keeps the background from the character image itself (e.g., a movie screenshot) and animates that character to your template video's movements; the video states it 'doesn't work as well as replace.'. Apply: Choose 'animate' only when you want to preserve the character image's own scene/background, such as inserting your movements into a movie screenshot.
Template video recording — A self-recorded reference video, ideally on a tripod against a solid, non-distracting background, that supplies the movements the generator follows. Apply: Record yourself on your phone in front of a plain wall or simple setting that isn't too busy, mounted on a tripod, avoiding excessive movement or camera shake.
Screenshot-to-ChatGPT character creation — A technique using ChatGPT to generate the character image by editing a screenshot pulled from the template video, ideally one showing the subject's hands. Apply: Take a screenshot from your template video showing your hands, upload it to ChatGPT, and prompt it with the desired transformation, e.g. 'make this character older and he is wearing a old style tunic natural colored shirt.'
Resolution/credit settings — Higgsfield's generation settings where output resolution determines credit cost, with 720p costing 57 credits in the demonstrated example. Apply: Set resolution to 720p before hitting generate, budgeting for the associated credit cost (57 credits in this example).
The workflow deliberately chains two separate AI tools rather than using one platform end-to-end: ChatGPT handles character-image editing, while Higgsfield/Wan 2.2 handles the video generation.
Background complexity in the template video is claimed to directly affect output quality — a busy background is said to make the generated video come out worse.
Which frame you screenshot matters: the video specifically recommends choosing one where your hands are visible before sending it to ChatGPT.
'Replace' and 'Animate' are not interchangeable stylistic options but produce fundamentally different backgrounds (your own setting vs. the character image's setting), and the video explicitly ranks Animate as inferior ('It doesn't work as well as replace').
The annual Higgsfield plan is framed as a tradeoff — lower effective monthly cost in exchange for a large upfront payment for the full year.
The tutorial ends with what reads as an affiliate/referral link ('If you want a direct link to Hicksfield, check out my description'), indicating a promotional relationship with the platform.
«This is easier than you think.»
— 00:00
«2.2 is the platform, the generator that allows you to replace yourself.»
— 00:12
«here's how you create that video in three easy steps.»
— 00:30
«make this character older and he is wearing a old style tunic natural colored shirt»
— 01:39
«Replace is when you simply replace yourself with the character that you upload.»
— 02:38
«It's going to keep your same background.»
— 02:43
«I want to ideally go to 720p and hit generate 57 credits.»
— 02:49
Reception
Mixed reception with praise for the video's delivery, but significant criticisms about incomplete demonstration and product availability concerns.
The video delivers a compact, concrete three-step tutorial with specific platform and settings guidance (Higgsfield, Wan 2.2 replace vs. animate, 720p, 57 credits), but it never shows the final generated result and doubles as a paid-plan/affiliate pitch for Higgsfield.
03:03