Quick Answer
Midjourney costs $10/mo minimum with no free tier, while local FLUX and Stable Diffusion are free and unlimited. The deciding detail for sellers: Leonardo’s free tier grants Leonardo rights to use and distribute your images, whereas local FLUX output is yours outright.
Midjourney costs $10 a month minimum with no free tier. Running FLUX or Stable Diffusion locally is free and unlimited. The detail that decides this for anyone selling art: Leonardo's free tier grants Leonardo the right to use and distribute your images, while local FLUX output is yours outright.
Disclosure: some links below are affiliate links. If you sign up through them we may earn a commission at no extra cost to you. It never changes our picks. See our affiliate disclosure.
Key takeaways
- Midjourney has no free tier. Basic is $10/mo, Standard $30, Pro $60, Mega $120, with 20% off annually.
- Local generation is free and unlimited, and FLUX.1 dev's licence explicitly permits selling the images despite the model being non-commercial.
- Leonardo's free tier gives Leonardo rights to use, reproduce, modify and distribute your images, and lets other users access your public assets.
- Ideogram Free $0, Plus $15/mo, Pro $42/mo. Adobe Express runs $9 to $39.
- Local needs a graphics card: 4GB video memory for SD 1.5, 8GB for SDXL, 16GB for FLUX.1 dev.
What each costs, verified September 2026
| Tool | Free tier | Entry paid | Commercial rights on free |
|---|---|---|---|
| Local (FLUX/SD) | Unlimited | $0 forever | Yes |
| Midjourney | None | $10/mo | n/a |
| Ideogram | Yes, slow credits | $15/mo | Limited |
| Leonardo | Yes, daily tokens | $12/mo | Yes, but see below |
| Adobe Express | Yes | $9/mo | Not indemnified on free |
Midjourney's annual billing cuts 20%, making Basic $8, Standard $24, Pro $48 and Mega $96 per month. It bills by fast GPU hours: 3.3 on Basic, 15 on Standard, 30 on Pro, 60 on Mega. Ideogram's Pro tier is $42 and Team $20 per user.
The free-tier rights problem
This is the section worth reading before you sell anything, and it is where most comparisons stop short.
Leonardo's own terms state that for free tier users, "Leonardo.Ai holds rights to use, reproduce, modify, and distribute any images you create." You receive a non-exclusive royalty-free licence to use your creations commercially, but other users can also access and use your publicly available assets through platform features.
Read that carefully if you were planning to sell Leonardo free-tier output as exclusive artwork. You are granted usage rights, not exclusivity, and the platform retains broad rights over what you make.
Midjourney's free tier does not exist, so the question is moot, but its paid output falls under terms where subscribers own their assets, with companies over $1 million in revenue required to be on Pro or Mega.
Local generation has no such clause, because there is no platform. The model licence is the only thing governing your output, and for the right model that licence is clean.
The FLUX licensing detail nobody explains
Most articles tell readers that FLUX.1 dev is non-commercial and they cannot sell its images. That is wrong, and it matters because FLUX is the current quality leader.
Black Forest Labs' licence states directly:
"We claim no ownership rights in and to the Outputs... You may use Output for any purpose (including for commercial purposes), except as expressly prohibited herein."
The non-commercial restriction applies to the model, not the images. You cannot build a paid service around FLUX.1 dev itself, and you cannot use its output to train a competing model. Selling prints, using images in client work, or putting them on products is permitted.
FLUX.1 schnell is Apache-2.0, making both model and output fully commercial with zero ambiguity. For anyone who wants no licensing questions at all, that is the answer.
For Stable Diffusion, SD 1.5 and SDXL use CreativeML Open RAIL-M, which permits commercial use royalty-free with restrictions on illegal and harmful content. SD 3.5 uses Stability's Community License, free commercially only while your organisation earns under $1 million in total annual revenue, not merely revenue from AI work.
Where paid genuinely wins
Quality out of the box. Midjourney's aesthetic defaults are excellent with no configuration. Local models need prompt work, model selection and settings tuning to match it.
No hardware requirement. Midjourney runs anywhere. Local generation needs a capable graphics card, and if you do not have one, that is a several-hundred-dollar barrier that dwarfs a year of subscription.
Zero maintenance. No installs, no model downloads, no version conflicts, no CUDA errors.
Speed and convenience. Fast GPU hours mean immediate results. Local generation on modest hardware can be slow.
Where free genuinely wins
Unlimited volume. Midjourney's Basic tier gives 3.3 fast GPU hours a month. Local generation has no meter at all. For anyone producing hundreds of images, this reverses the economics completely.
Full control. Custom models, LoRAs, fine-tuning, ControlNet, inpainting, exact seeds and settings. Local generation offers a depth of control no hosted service matches.
Privacy. Nothing is uploaded. For client work under NDA or unreleased product imagery, that matters.
Clean rights. With FLUX schnell or SD 1.5, no platform holds any interest in your output.
What hardware you actually need
| Model | Minimum VRAM | Comfortable |
|---|---|---|
| SD 1.5 | 4GB | 8GB |
| SDXL | 8GB | 12GB |
| FLUX.1 dev (quantised) | 7GB | 12GB |
| FLUX.1 dev (standard) | 16GB | 16GB+ |
On Apple Silicon, SDXL wants at least 16GB of unified memory and FLUX in reduced precision is comfortable on 24GB. Intel Macs are effectively out.
Use ComfyUI, Fooocus or InvokeAI. Do not start with AUTOMATIC1111: its last release was February 2025 and it cannot run FLUX, despite still being the default recommendation in most guides.
Where the litigation stands
In Getty Images v Stability AI, decided in the English High Court in November 2025, Getty dropped its main training and output copyright claims mid-trial and its secondary infringement claim was rejected. The court accepted that model weights contain no copies of training images.
In the US, Andersen v. Stability AI was scheduled for trial in September 2026, which is imminent as of writing. Check current status before relying on this.
Generating and selling AI images is legal and widely done today. The unresolved questions concern how models were trained, not what you make with them.
Which should you choose?
Choose local generation if you have a graphics card with 8GB or more, you generate at volume, you need full control or privacy, or you want unambiguous commercial rights without a subscription.
Choose Midjourney if you want the best aesthetic quality with no setup and generate modestly enough that $10 a month beats the hardware and time.
Choose Ideogram if you need accurate text inside images, which remains its distinctive strength.
Be careful with Leonardo's free tier if exclusivity matters to your work. The paid tiers change the terms; the free one grants the platform broad rights.
None of the above if you are starting from a photograph. Every tool here generates an image from text, which is the wrong category when you already have the picture. Colourising or restoring an old photo is a different job, and Palette.fm does that one specifically. So is turning selfies into professional headshots, which Aragon and HeadshotPro handle with a trained likeness rather than a prompt. Local Stable Diffusion can approximate both with enough effort, and for a one-off it rarely repays the setup time.
For setup details, read free AI image generation at home. To compare the free options, see the best free AI image generators.
The verdict
If you have the hardware, local generation is the best value in AI right now: unlimited images, full control, and with FLUX schnell or SD 1.5, cleaner commercial rights than most free tiers offer.
If you do not have a suitable graphics card, Midjourney at $10 a month is excellent and the hardware maths does not favour buying a GPU just to save it.
The one thing worth avoiding is selling work made on a free tier whose terms you have not read.
Related reading: free alternatives to Midjourney and Midjourney versus DALL-E versus Ideogram.
Frequently Asked Questions
Does Midjourney have a free tier?
No. Midjourney has no free plan and no free trial. Basic is $10 a month, Standard $30, Pro $60 and Mega $120, with 20% off annual billing. It bills by fast GPU hours: 3.3 on Basic up to 60 on Mega.
Can you sell images made on Leonardo's free tier?
With caution. Leonardo grants free users a non-exclusive royalty-free licence for commercial use, but its terms also state Leonardo holds rights to use, reproduce, modify and distribute any images you create, and other users can access your public assets. Exclusivity is not guaranteed.
Can you sell FLUX.1 dev images commercially?
Yes. Despite the non-commercial model label, Black Forest Labs' licence states they claim no ownership of outputs and you may use output for any purpose including commercial purposes. The restriction covers the model, not the images. FLUX.1 schnell is Apache-2.0 with no ambiguity.
Is local AI image generation actually free?
Yes, if you have the hardware. There is no subscription or per-image cost. You need a graphics card with at least 4GB of video memory for SD 1.5, 8GB for SDXL and 16GB for standard FLUX.1 dev. Without a suitable card, a subscription is cheaper than buying one.
What does Ideogram cost in 2026?
Ideogram has a free tier with slow credits and one concurrent generation. Plus is $15 a month, Pro $42 and Team $20 per user. Its distinctive strength remains accurate text rendering inside images.



