Channel Grow
All courses
The Base Selfie Members only8 min

Generate the Base Selfie with FLUX.1

Keyword-style prompting, seed rolling, and LoRA weight tuning: how to get a V1 selfie that actually looks like you before any realism layers go on.

Casual AI-generated selfie of a young man in a white shirt at home with natural window light
What a good V1 looks like: right person, right outfit, casual framing — realism comes in the next section. (AI-generated illustration)

Goal of this lesson: one image — call it V1 — that looks like you, dressed how you want, framed how you want. Realism comes later; identity comes first.

Step by step:

  1. Open the FLUX.1 with LoRAs playground in Text to Image mode.
  2. Under Loras, paste the link to your Digital Avatar LoRA.
  3. Write a short, keyword-style prompt. The working example: A selfie of [your trigger word], buzz cut, short hair, white shirt, shot on an iPhone, IMG_1018.HEIC, iPhone photo.
  4. Hit randomize on the Seed field, then Generate.
  5. Not you, or wrong outfit? Randomize the seed and go again. Still not you after several rolls? Nudge the Avatar LoRA weight up or down by 0.1 and retry.
  6. When an image lands, write down its seed. A fixed seed is what lets you add the realism LoRAs in the next section without the whole composition changing underneath you.
Fal.ai Loras section with a LoRA URL pasted into the Path field
Step 2 in the playground: your Digital Avatar LoRA link goes in the Loras → Path field.
Fal.ai Seed field showing a numeric seed with a randomize button
Step 4: the Seed field. Hit the randomize button until the face is yours — then keep that number.

Why keywords beat prose here: a short prompt concentrates the model's attention on your face and clothing instead of spending it on background detail. Long flowery prompts come back in the editorial pack — for the base selfie, terse wins.