Flux
“Best for people and hands” — a portrait artist who studies every fingertip
What is Flux?
Flux is a family of image generation models from Black Forest Labs in Germany. It renders tricky details like faces and hands naturally, so it is widely used for photo-like images with people.
- 1A model well regarded for rendering people and hands
- 2The official ways to use it are the playground and API at bfl.ai
- 3No free plan — you pay per image, as you use it
What sets it apart
These are the differences we found by running the same material through several tools in our training sessions.
People look natural
AI images usually get screened out of external materials because of faces and hands. Flux is known for fewer redo attempts on exactly these, and it is the first candidate for promotional shots that include people. You spend less time counting fingers.
Choose the price and quality you need
The official price list covers several tiers, not a single model. The lightweight klein tier starts at $0.014 per image and the business-grade pro tier at $0.03, so you can generate drafts cheaply and move only the final version up to a pricier tier.
Can run on your own servers
Some models have open weights, so you can run them on your own infrastructure (the klein tier is Apache 2.0). For organizations that cannot send data outside the company, this option alone is an advantage other tools do not offer. You will, however, need a server and staff to manage it.
⚖️ The weaknesses are clear, too
There is effectively no free tier to try, and the interface and documentation are entirely in English. For placing Korean text on images or accurately depicting Korean props and settings, another tool is faster. It is better suited to one or two people who own it than to rolling out across the whole organization.
💡 In plain EnglishIt’s less a painter of flashy backdrops than a portrait artist who keeps refining fingers and eyes.
In images with people, a single awkward finger makes them unusable for external materials. For recruiting promotions or internal campaigns with many portrait shots, this one difference decides whether you have to redo the work.
Same job, different approach
You need “people at work” shots for recruiting materials
The old way
- Choose photos of international models from a paid stock photo site
- It doesn’t suit our company’s look, so we pick again
- Find a usable shot, pay, and fit the specs
- Next month, we start the same work over from scratch
With Flux
- Describe the people and situation you want in a sentence and generate several drafts
- Filter out shots with awkward hands or expressions
- Regenerate only the chosen shots on a higher tier to finish
- Save the prompts that worked and reuse them next time
Common uses
Promotional shots with people in them
Create images where people are the focus, such as recruiting notices and internal campaigns. Avoid requests to make someone look like a real person.
Reviewing drafts in bulk
Generate several versions of the same scene on a cheaper tier to settle on a direction, then regenerate only your pick on a higher tier.
Photo retouching and partial edits
Use it to edit an existing image, such as swapping only the background or props. Check the rights to any people in the original photo first.
Who it’s for
Best for
- ✓Anyone who often creates promotional materials featuring people
- ✓Anyone who needs to generate images in bulk while tracking per-image costs
- ✓Organizations that can’t send data outside the company and want to run it on their own servers
Better served by other tools
- –People who need Korean text inside the image
- –Anyone who wants to start for free
- –Beginners uncomfortable with an English interface
Can we use it at work?
The questions that adoption leads and security teams ask first.
Please confirm: data retention periods for inputs, whether inputs are used for training, and exact pricing by license tier. Check these against the official terms or with the vendor’s sales team.
10-minute quick start
- 1
The official way to use it is bfl.ai. Create an account at playground.bfl.ai and top up your credits.
- 2
Describe the scene you want in an English sentence. It’s easiest to ask Claude to "turn this into an English image prompt."
- 3
Generate several images on a cheaper tier to settle on a direction, then regenerate only your picks on a higher tier.
- 4
If you will use it repeatedly for work, connect through the API. If your data cannot leave the company, consider running the open weights on your own servers.
Copy-and-paste prompts
Replace only the [bracketed] parts, then paste it in.
A Korean office worker in their 30s smiling and chatting with a colleague in a bright office. Natural light, shallow depth of field, hands that look natural. Photorealistic texture, vertical 3:4.
Close-up of [product name] held in both hands. Knuckles and fingernails clearly visible, with a softly blurred solid-color background. Soft studio lighting, square format.
In a meeting room, a presenter standing in front of a whiteboard explaining, and four attendees seated and listening. Faces clearly visible, natural hand gestures. Indoor lighting, 16:9 landscape.
Keep the person and clothing in the attached photo as they are, and change only the background to a modern office. Match the lighting direction and shadows to the original.
Create four images of [situation], changing only the mood. First, bright and upbeat; second, calm and professional; third, warm tones; fourth, black and white.
Using it with other tools
Linking two or three tools changes your work more than any single tool does.
Portrait shots → carousel post (40 min)
Portrait shots → short video (1 hour)
Before you start
- Results are more reliable if you write your prompt in English instead of entering Korean sentences directly.
- Don’t ask it to resemble a real person or a celebrity.
- Do not expect it to put Korean text inside images. Add the text afterward in an editing tool.
- Each tier has a different price. Generating hundreds of drafts on a pricier tier makes costs climb fast.
Pricing: personal use and work use
Plans that individuals pay for are different from plans that organizations contract for by member.
For individual use
Cheapest tier, for reviewing drafts in bulk
Default pick for work
A tier with higher quality and broader control
Top tier, for final versions
Previous-generation editing model (max is $0.08)
Using it at work
FLUX.2 [klein] self-hosted, based on 10,000 images per month, 1 domain, 10 users
Includes FLUX.2 [dev]; based on 100,000 images per month and 1–3 domains
All models, new releases, volume and domain customization, commercial use
Checked 21 Sep 2026 · Official per-image prices in US dollars. Prices rise with resolution (megapixels), and playground and API prices are the same. Pricing can change without notice, so check the official page before you pay. Black Forest Labs official pricing ↗ · FLUX official docs pricing table ↗
Where Flux falls short
We don’t just list the positives. Here is what fell short when we used it.
- Keeping the same face is hard
Even with the same prompt, the person’s face changes a little each time. To keep one person across several scenes, you have to try repeatedly and attach reference images.
- You can watch your balance go down
Because you prepay for credits, you see your balance drop with every draft. If you expense it to the company, you also need to track usage history separately.
- Choosing a tier is hard
With several tiers (klein, pro, flex, max) and all the documentation in English, you first spend time flipping between pages to work out which tier to use. Choosing is already work.
Tools to compare
FAQ
Last updated 22 Sep 2026 · The ratings on this page come from John Kim, CEO of Nanum Consulting, who has used these tools firsthand in training and real work.
Flux as a core skill across your organization
Send us a message and we’ll reply within 24 hours. We’ll propose a curriculum tailored to your industry and job roles.
Courses covering this fieldAI Tool Training (by Use Case)