Home Creators Posts Import Register Favorites Logout

Downloads

Content

Take one photo of your AI influencer and put her anywhere, in any outfit, from any angle. Or keep the photo and change the details: colors, lighting, props. All in Krea 2 Turbo, in seconds.

What you can do with it

A whole photoshoot from one photo. One photo in, a full travel photoshoot out.

New camera angles. See her from any angle.

Put her in your scene. Drop her into any location photo.

Outfit colors. One outfit, every color.

The phone-flash look. The candid flash look that's everywhere right now.

Props and pets. Add the details that make a feed feel real.

Members get the workflow, a full gallery of before and after examples, the exact prompts behind every one, and my tips for getting results like these on the first try.

Everything you need is inside the workflow. Every model download link, which folder each file goes in, the custom nodes to install and step by step instructions are in the How To Use note on the left side of the workflow. Open it, read it top to bottom, and you're set up. If you already run my Krea 2 v4 workflows, you only need the two edit LoRAs and the Krea2Edit node pack.

Want to zoom in? The full-size 4x screenshot is attached to this post as well.

Which mode do I use?

Identity mode (switch on true): new scene, new pose, new outfit, new camera angle, or putting her into another photo. The face and body come from your photo.

General mode (switch on false): keep the photo and change something in it. Colors, lighting, adding things, swapping text or the background.

The switch also sets the right fidelity for each mode (2 for Identity, 4 for General), so you don't have to touch anything else.

What's inside

Both Krea 2 edit LoRAs (Identity Edit V2 and General Edit V2) behind a single switch.

Two-image mode to put your character into any location photo.

Optional character LoRA slot, so you can stack your own Krea 2 character LoRA on top for an even tighter face lock.

ClownsharKSampler with the same settings as my Krea 2 v4 workflows, which keeps hair texture clean.

Results match your photo's shape, plus a Before / After viewer right in the workflow.

How I made each example

Default settings on every one unless the card says otherwise. The labels show which photo went in where, and the box under each card is exactly what I typed.

1. A whole photoshoot from one photo

Mode: Identity. This is the reason Identity mode exists. One plain photo of your character goes in, and you get her on a yacht, on a balcony in Paris, by a pool in Bali. Same face, same body, new outfit, new light. One good photo turns into a week of travel posts.

Tip: Start the prompt with how much of her you want in the shot ("Full body", "waist up"). If you leave it out, Krea copies the crop of your photo.

2. New camera angle

Mode: Identity. Spin the camera around her. The back view over the shoulder, a side profile, a low angle from the floor. Krea keeps her body shape and the exact bikini while it rebuilds the shot.

Tip: Set ref_position to beside in the Krea 2 Edit Patch node when you want a big pose or angle change. On aligned it tries to keep your original layout.

3. Put her in your scene

Mode: Identity, 2 images. Got a location photo you love? A hotel suite, a club booth, a friend's car. Load it as Image 1, load your character as Image 2, and tell Krea where she goes. The room's props stay put and the room's own light falls on her, so she looks like she was actually there.

Tip: Start the prompt with what has to stay in the room ("Keep this exact hotel bed: the white sheets, the headboard, the breakfast tray..."), and place her close to the camera. The bigger her face is in the frame, the cleaner her eyes and features come out.

4. Swap the outfit colors

Mode: General. General mode changes only what you name. Turn one outfit into a whole color range for a drop or a carousel, without the face, body or room moving at all.

Tip: Name the old item and the new one ("her navy bikini to a hot pink bikini"). It's more reliable than just asking for pink.

5. The phone-flash look

Mode: General. The candid night-out flash photo is everywhere on Instagram right now, and it makes AI photos look a lot more real. Take any soft daylight shot and turn it into a harsh on-camera flash photo.

Tip: For big lighting changes, drop the General fidelity node from 4 to 2. At 4 it holds on to the original light too hard. Then describe where the light hits and where it goes dark: "hard flash on her face and chest, the room falling off into near black".

6. Add a prop or a pet

Mode: General. Puppies, coffee cups, shopping bags, a bouquet of flowers. The little real-life details make an AI influencer's feed feel lived in.

Tip: Say where it goes and how she's touching it. Krea is great at the object itself; telling it where her hands are gets you a natural pose.

Quick rules that save you re-rolls

Write it like you're describing a photo, in plain English. No trigger words needed.

Say how much of her is in the shot: full body, waist up, close-up.

Describe the new outfit and say what's gone ("the cardigan removed"). Krea likes to keep clothes from your photo.

Describe light by where it lands, not just its name.

Removing big things (like a whole person) is the one job Turbo struggles with. Say what should be there instead, or re-roll the seed.

Set the seed to fixed while you tweak a prompt.

Huge credit to conradlocke (lbouaraba), who trained both edit LoRAs and wrote the Krea2Edit nodes. The edit LoRAs are free for personal and non-commercial use; if you're using them for paid client work, the author licenses commercial use case by case. Links are in the workflow note.

The workflow file is attached below. Drop it into ComfyUI, open the How To Use note, and have fun with it.

Files

Comments

Michael

Which model do you personally recommend for editing—Qwen Workflow or Krea2 Edit? In terms of flexibility and output quality?

dreamai

same question

LoRAtech - ComfyUI Hub

Qwen absolutely in my opinion. It maintains consistency and listens to prompting much better. Krea 2 only being an “edit model” because of a LoRA that adapts it to be so definitely limits its capabilities against true edit models that were trained to be edit models.

LoRAtech - ComfyUI Hub

Qwen absolutely in my opinion. It maintains consistency and listens to prompting much better. Krea 2 only being an “edit model” because of a LoRA that adapts it to be so definitely limits its capabilities against true edit models that were trained to be edit models.