Flux 1.1 Pro is now live

Imagine it.
Generate it.
Train it.

Five production-grade image models plus your own custom-trained LoRAs — in one creator studio. Mood board to brand-ready render in minutes, not weeks.

28,000+
Creators building
4.2M+
Images generated
840+
Custom LoRAs trained
99.2%
Generation success
cyberpunk samurai in neon alley, 85mm cinematic
Flux Dev
cyberpunk samurai in neon alley, 85mm cinematic
alpine cottage at dawn, watercolor, soft light
Playground v2.5
alpine cottage at dawn, watercolor, soft light
holographic jellyfish in black void, studio lit
Flux Schnell
holographic jellyfish in black void, studio lit
abandoned temple overtaken by moss and vines
SDXL 1.0
abandoned temple overtaken by moss and vines
Model library

Five state-of-the-art models. One prompt box.

Switch models with a click. Every model runs on the same prompt-and-settings stack, so comparing options takes seconds.

Flux Schnell
1 credit
Flux Schnell
Black Forest Labs
Rapid ideation, mood boards
Fast
Flux Dev
2 credits
Flux Dev
Black Forest Labs
Photoreal portraits and products
Medium
SDXL 1.0
2 credits
SDXL 1.0
Stability AI
Structure, composition control
Medium
Playground v2.5
2 credits
Playground v2.5
Playground AI
Editorial and aesthetic looks
Medium
Kandinsky 2.2
3 credits
Kandinsky 2.2
AI Forever
Painterly, stylized illustration
Slow
The pigment lab

Everything a generative image studio needs.

Built for creators who ship daily — from solo game devs to brand studios running client campaigns.

Five image models, one studio

Switch between Flux Schnell for lightning drafts, Flux Dev for photorealism, SDXL for control, Playground v2.5 for aesthetics, and Kandinsky 2.2 for painterly looks.

Train a LoRA on your style

Upload 8 to 20 reference images and fine-tune a custom model on your character, product, or brand aesthetic. Re-use it forever.

Prompt enhancement

Our prompt assistant rewrites your rough idea into a specific, detail-rich prompt. Feels like a pairing with an art director.

Aspect-ratio presets

1:1 for social, 16:9 for cover art, 9:16 for story, 4:3 and 3:4 for prints. No manual pixel math.

Community gallery

Explore, remix, and steal prompts from what the community is making. Every image is a click-to-copy prompt.

Collections & history

Every generation is saved with its prompt, model, and seed. Searchable. Organize into named collections for each project.

Custom LoRA training

Train a model on your own style.

Upload 8 to 20 reference images of a character, product, or aesthetic. We'll fine-tune a LoRA on Flux Dev and give you back a model that generates on-brand every time.

  1. Upload references
    Drag in 8–20 images. Varied angles, consistent subject.
  2. Train in 20–40 minutes
    1,000 steps on Replicate's flux-dev-lora-trainer.
  3. Generate anything, in style
    Use your trigger word in any prompt, anywhere.
Train your first model
Reference set · `hollowmark-hero` · 14 images
training referencetraining referencetraining referencetraining referencetraining referencetraining referencetraining referencetraining reference
Training progress
Step 842 / 1000
Trigger word: hollowmark_hero · Est. completion: 8 min
Workflow

From a rough idea to a brand-ready render.

01

Write your prompt

Start with a rough idea. Tap the Enhance button to let the assistant add cinematography, lighting, and style detail.

02

Pick a model and size

Fast drafts on Flux Schnell or high-detail renders on Flux Dev. Any aspect ratio, any guidance, any seed.

03

Train your own LoRA

Upload 8 to 20 references and train a model on your character or brand. Use it in every generation forward.

Creators in the wild

Used by indie devs, brand studios, and concept artists.

Trained a LoRA on my main character once, and I've generated 1,800+ consistent hero frames for our indie RPG in 3 weeks. Replaced a contracted concept artist we couldn't afford anyway.
Petra Lindqvist
Petra Lindqvist
Solo game dev, Hollowmark Studios
We train one LoRA per client brand and spin up social creative in minutes. 40 campaign variants for a Shopify client yesterday — would've been a week of studio time.
Marcus DeWitt
Marcus DeWitt
Creative director, Kernel & Co Brand Studio
Flux Dev + prompt enhance is uncanny for concept painting. I sketch the composition, describe the mood, and get 4 finished-looking boards in 90 seconds. It's my new morning ritual.
Aiko Tanaka
Aiko Tanaka
Concept artist, freelance (ex-Ubisoft Montreal)
Pricing

Start free. Pay when you scale.

150 credits on signup, no card required. Upgrade any time for more credits, custom LoRAs, and batch generation.

Free

Kick the tires, ship a side project.

$0/ mo
  • 150 credits on signup (~75 Flux Schnell images)
  • All 5 generation models
  • Public gallery access
  • History + 1 collection
Start free

Creator

For indie creators shipping weekly.

$15/ mo
billed annually
  • 1,500 credits / month
  • 1 custom LoRA trained / month
  • Unlimited collections
  • Private generations by default
Choose Creator
Most popular

Pro

For working professionals and teams.

$39/ mo
billed annually
  • 5,000 credits / month
  • 5 custom LoRAs / month
  • Batch generate 4 images
  • Prompt enhance + negative prompt
  • Priority generation queue
Choose Pro

Studio

For brand studios and game teams.

$79/ mo
billed annually
  • 15,000 credits / month
  • Unlimited LoRAs
  • Team gallery (5 seats)
  • Early access to new models
  • Priority support
Choose Studio
FAQ

Frequently asked

A LoRA (Low-Rank Adaptation) is a lightweight fine-tune of a base model that teaches it a specific subject, style, or aesthetic from 8 to 20 reference images. You use it by putting your trigger word in the prompt — e.g. `a photo of ohwx person standing in Times Square`. Training takes 20–40 minutes on our pipeline.