Higgsfield rebuilt its AI Influencer studio, and the new part is this: the character you design can now be dropped into video that already exists, not only into clips generated from scratch. For a short-form creator that means publishing without ever stepping in front of a camera. The service has 30 million users — and it deliberately makes its characters look a little off, so viewers can tell at a glance they are watching AI rather than a deepfake.
What changed
Semafor reported the update on October 7, 2026: you build a character, then insert it into existing footage. Higgsfield published its own write-up of the rebuilt studio on October 3.
Before, an AI character lived only inside generated clips. Now it works on real material: you bring footage — yours or a reference video — and either the performer or individual objects in the frame get replaced.
Semafor puts Higgsfield at 30 million users, with a $1 billion annual run rate the company announced in September 2026.
How the character is built
There are no prompts. Instead of describing a person in words, you click through menus — 19 settings in total, split into a base panel and advanced tabs.
| Settings group | What's inside |
|---|---|
| Base | character type (average, bold, extreme), gender, origin, skin and eye color, age, skin conditions |
| Face | eye type and details, mouth, teeth, ears, horns, skin material, patterns |
| Body | build, height, proportions, extra limbs |
| Style | hair, accessories, rendering style |
Skin conditions include freckles, scars, birthmarks, vitiligo and albinism. Uploading your own photo is optional — if you do, the studio keeps the facial features from that shot. Every generation returns two frames: a close-up portrait and a full-body shot.
The cost is shown before you confirm. Generating a character in the new studio runs 1.12 credits — roughly $0.06 at the current credit rate.
How the character gets into existing footage
That part runs through Genjutsu, and it has two modes.
Motion transfer
You supply a reference video with the movement you want — film it yourself or use existing footage — and the character repeats that movement while keeping its own look. Talking-head formats use lipsync. There are also ready-made motion templates if you have no reference of your own.
Objects swap
The second mode replaces things rather than people: a product, an outfit, a prop inside the shot. This one leans toward brand work rather than face-to-camera content.
The technical limits:
| Parameter | Value |
|---|---|
| Reference video length | 4–30 seconds |
| Character or product images | up to 30 |
| Output resolution | 480p, 720p, 1080p |
Separately, Soul ID training takes 20+ portraits and then makes your character available across other Higgsfield tools.
What it costs
Plans run from $15 to $129 per month, and credits convert at 20 credits per $1. So a character costs about six cents; the real money goes into video generation, where the price depends on the model, the resolution and the clip length.
| Item | Price |
|---|---|
| Creating a character | 1.12 credits, ≈ $0.06 |
| Credit rate | 20 credits per $1 |
| Subscriptions | $15 to $129 per month |
Why the characters look deliberately strange
This is the most interesting detail in the story. Higgsfield characters read as realistic but not quite of this world: geometric haircuts, elongated necks, deliberately extreme proportions. That is on purpose — the studio does not want the character mistaken for a real person.
A Higgsfield spokesperson puts it this way: "A flawless face is the fastest way for a character to read as generic AI."
For a creator there is a practical takeaway. An obviously synthetic host is a feature, not a flaw: the audience knows it is watching a drawn character, so nobody feels tricked by a deepfake.
What this gives a short-form creator
- You never have to be on camera. "I don't want my face in the frame" is the most common blocker for new creators. A character removes it, and you can still drive the performance from your own footage.
- One recurring host builds a brand. A consistent character is recognizable in the feed in a way faceless clip compilations are not.
- Objects swap is the brand play. If a brief requires the product on screen, you can place it into footage you already shot instead of reshooting the scene.
- The 4–30 second reference fits short form exactly. This is not built for long monologues.
- Do the math first. The character is pennies; video generation is where the budget goes. Decide how many clips a month you need, then pick the plan that covers it.
- Disclose it at upload. Platform rules require labeling realistic AI-made content, and platforms add those labels themselves when they detect generation.
The other side: memecoins and AI labels
Semafor notes plainly that some AI influencers already pull millions of views and thousands of followers — and that people are using them to promote memecoins. Worth knowing: the format is young and its reputation is mixed.
The second thing is labeling. Mark your uploads as AI-generated when the content is realistic; that costs you nothing compared to reduced reach or a removed video. We covered how to check footage and images for AI in this guide.
What we don't know yet
- Which plans include both Genjutsu modes — Higgsfield only says model availability depends on your subscription.
- The exact price of a video generation: it depends on model, resolution and length, and is only shown right before you confirm.
- Whether the service embeds an invisible provenance marker in the file — the tool description doesn't say.
- How consistently the same character holds up across clips without Soul ID training.
FAQ
How do I create an AI influencer for free?
Not in Higgsfield, strictly speaking: the studio runs on credits, and credits come with plans starting at $15 per month. The character itself is about $0.06; video generation is the real cost. For free experiments with AI video, see our roundup of free AI video generators.
Can I put an AI character into a video I already filmed?
Yes — that is the update. Motion transfer moves the performance from your footage onto the character; Objects swap replaces props and products in the frame. The reference clip can be 4 to 30 seconds.
Do I have to upload my own face?
No. The photo upload is optional and a character can be assembled entirely from the menus. If you do upload a shot, the studio preserves those facial features.
Do I need to label videos made with an AI character?
Yes — flag them as AI-generated at upload. Platforms require it for realistic synthetic content and detect generation on their own anyway.
How is an AI character different from a deepfake?
A deepfake imitates a specific real person. This is the opposite: the character is invented and intentionally stylized so viewers immediately see it is not a real human.
Read next
- How to check if an image is AI generated: SynthID Detector is open to all
- Best AI video editors in 2026
- How to get more views on Reels: Instagram's Edits AI reads your stats
A character only answers the question of who is on screen — you still get paid per view. On Prime Oracles, creators pick up brand briefs under "Content Rewards" and earn per view on their own videos, whether the person in the frame is you or a character you drew.