Sensei · AI influencer post
Post your AI creator in four places, same face
Sensei's AI influencer recipe makes one portrait of your creator (or uses one you saved), then shoots every slide of the carousel with that portrait as the reference. To test the idea I ran those steps myself on Seedream 5, the model Sensei uses for stills, for Noor, a lifestyle creator who does not exist.
Opens Sensei with the AI influencer post recipe ready to fill in: who the creator is and what the post is about. Sensei shows the plan and price before it spends anything.




What I asked for
Your creator: Noor, late twenties, warm brown skin, dark curly shoulder-length hair, small gold hoops, oversized cream cardigan over a white tee What the post is about: a slow Sunday. Farmers market, flowers at home, journaling on the sofa
- Slides
- 4 (portrait + 3)
- Credits
- 10 (2.5 per photo)
- Render time
- About 3.5 minutes per slide, run side by side
- Model training
- None
The usual advice: train a model first
Most guides to AI influencers start with a LoRA: collect 20 to 30 images of your character, train for half an hour, then prompt the trained model. That was the right call a year ago, when a reference photo alone gave you a cousin of your character, not your character.
What I saw with one portrait
I made Noor's portrait first, then three more photos with only that portrait as the reference and a different place in each prompt. Her curls, hoops and cardigan are in every slide. Her face reads as the same person at a glance and on a second look, across a café, a crowded market, a kitchen and a sofa. Ten credits, no training, no dataset.
- Where it drifted: on the sofa slide she is barefoot in dark trousers. I never said what she wore below the cardigan, so the model chose. Name the whole outfit if it matters.
- Where it held: the slides with her face at three quarters (market, kitchen) are as close as the straight-on portrait.
- What surprised me: the market slide has a believable crowd behind her and none of the strangers borrowed her face, which is the classic failure.
When a trained model still wins
If you post the same creator every day for months, train one. PixelDojo has LoRA trainers for that, and a LoRA keeps small details (a scar, a tattoo, the exact jawline) steadier across hundreds of posts than any single reference does. It also lets you post without keeping the portrait around.
For a launch, a test account, or a brand mascot you are not sure about yet, one portrait and a carousel recipe gets you posting today. Save the creator in Character Studio and Sensei will find them by name next time.
Made and written by Brian Lovett, founder of PixelDojo. Tested .
Questions people ask
Do I need to train a LoRA to keep my AI influencer consistent?
Not to start. One good portrait as the reference held Noor's face, hair and accessories across all three new slides in my test. Train a LoRA when you post the same creator daily and need small details to stay exact.
Can I use a character I already made?
Yes. Name a saved Character Studio character in the recipe (or write @Name) and Sensei uses their photos as the reference for every slide.
What does a carousel cost?
My four slides cost 10 credits on Seedream 5, 2.5 per photo. Sensei shows the plan and the price before it makes anything, and you approve it.
Does Sensei write the caption too?
Yes. The recipe ends with a short caption in the creator's voice and three hashtags, which you can edit before posting.
Can I make video posts of the same creator?
Yes. Ask Sensei to animate any slide, or make a talking clip with your creator as the presenter in Marketing Studio.
More to make
Pick a different job
Start with one portrait you love. Everything after it is a reference away.
Make a post with Sensei