AtomWords logo
Prompt Engineering · Arts Entertainment

AtomWords

A free visual dictionary of AI prompt keywords: each entry pairs one word with one real generated image or clip, so you can see exactly what a keyword does to your picture before you write the prompt.

Active GDPR compliant Free plan Free No public API 16+ Verified by Guidaio
Overview

What is AtomWords?

AtomWords is a visual dictionary of AI prompt keywords, built on a single rule: one word, one image. Where most prompt galleries set a wall of full prompts beside a wall of pictures — subjects, styles, parameters and render settings tangled into one block — each AtomWords entry isolates a single visual variable, so you can finally see what that word does on its own.

The problem it addresses is stated plainly by its founder, a multidisciplinary designer working under the name musefeng: you can look at a finished image, read the whole prompt behind it, and still not know which word did the visual work. Was it cinematic? Liminal? Brutalist? AtomWords answers by demonstration rather than definition.

The catalogue is organised along four orthogonal axes: Subject (what is depicted), Style (the aesthetic treatment), Technique (camera, lighting and render parameters) and Motion (movement and animation cues). Eight keyword categories sit beneath Style and Technique — Lighting, Composition, Camera, Aesthetic, Medium, Material, Color and Mood — and seventeen subjects carry the entries, the largest being Object with 149, People with 113, Interior with 88 and Architecture with 61 as of 11 August 2026.

Open an entry and you get the keyword, its category and subject, a one-line definition and two applied tips under the heading AtomWords Insights, the complete prompt directive with a copy button, the engine and version that produced the example — Midjourney v8.2 on the liminal entry — clickable visual tags, a favourites button and a report link. A separate video section pairs camera-movement and motion keywords with real generated clips, aimed at Sora, Veo, Kling, Runway, Seedance, Hailuo, Midjourney, Pika and Luma. A reverse dictionary called Explore lets you start from an everyday description — golden hour, terrazzo, motion blur — and work back to the exact keyword.

It is a one-person project, launched on 8 June 2026 and growing daily, and its author says so openly: the catalogue is still small. Browsing, copying and downloading are free, commercial use included.

What it does

  • Look up any AI art keyword and see the exact image it produces
  • Copy a complete, ready-to-paste prompt in one click
  • Browse camera-movement and motion keywords paired with real generated clips
  • Work backwards from a plain description to the keyword behind the look
  • Download sample images and videos for free, commercial use included
  • Explore keywords by subject, style, technique or generation engine
  • Save keywords to a personal favourites list
Audience

When to use AtomWords / When not to

A quick filter to help you decide if AtomWords is the right fit.

When to use AtomWords

  • Designers and art directors who write image prompts every day and need to know which word carries which visual effect
  • Video creators and cinematographers hunting for precise camera-movement vocabulary that carries across Sora, Veo, Kling or Runway
  • Marketing and social teams producing brand or ad visuals on no budget, with commercial rights already cleared
  • Prompt engineers building a reusable, model-agnostic keyword vocabulary instead of copying long prompt blocks
  • Curious browsers who want to widen their visual references and stumble on words they would never have thought to search for

When not to use AtomWords

  • Anyone expecting the site to generate images or video: AtomWords is a reference, not a generator
  • Teams needing shared workspaces, seat management or any form of collaboration, since the account is strictly individual
  • Developers looking for an API, a mobile app or a browser extension, none of which exist
  • Anyone who needs the content in a language other than English, the only language the site is published in
  • Businesses planning to bulk-extract, resell or train models on the catalogue, all explicitly forbidden by the terms
Get started

How to use AtomWords

A typical end-to-end flow, from setup to results.

  1. Open atomwords.com in a browser: no installation, no payment and no account are needed to read
  2. Search a keyword in the header search bar, or pick an axis from Dictionary, Subjects, Collections, Style, Technique or Motion
  3. If you only know the look you want, use the Explore page to work backwards from an everyday description such as golden hour or motion blur
  4. Open an entry to see its generated image or clip next to the AtomWords Insights note: a one-line definition plus two applied tips
  5. Check the engine and version credited on the entry so you know what produced the example
  6. Copy the full prompt directive with the Copy Prompt button
  7. Paste it into your own generator and adapt the trailing parameters, which follow Midjourney conventions
  8. Combine a few precise keywords — subject, style, technique, motion — rather than writing one long paragraph
  9. Download the sample image or clip, or reuse it as an img2img reference to guide your own generation
  10. Create a free account by email or Google sign-in to save favourites, keeping in mind the daily download limits that reset at 00:00 UTC
Quick read

Pros & Cons

Pros

  • Completely free today, commercial use included, with no attribution required
  • Each entry isolates one visual variable, which no gallery of full prompts allows
  • Every keyword is demonstrated by a real generated image or clip rather than an abstract definition
  • The vocabulary is model-agnostic and carries across image and video generators alike
  • The engine and its version are credited on each entry, which makes the example reproducible
  • Four ways in, including a reverse dictionary that starts from plain description
  • Privacy practice well above the average for a project this size: named sub-processors, a retention table, cookieless analytics and no model training on user data

Cons

  • The catalogue is still small and uneven, as the Operator says himself: Character and Mythical Creature hold five entries each
  • Only one of the announced industry collections is live, Business & Marketing, with 49 prompts
  • The free ride is explicitly temporary: both the terms and the privacy policy announce paid subscriptions, and a pricing page already exists
  • A single person, with no identified company and no published postal address, stands behind the service
  • No EU representative under Article 27 and no Data Processing Agreement, with Chinese governing law and a USD 50 liability cap
  • No API, no mobile app, no browser extension and no team or collaboration features
  • English only, with daily download quotas that are never quantified publicly
Pricing

Pricing & Plans

AtomWords is entirely free at the time of review. A permanent free tier is available and no paid plan exists: the terms of service state that this is the free-tier launch phase with no paid subscriptions yet, and the privacy policy confirms that no payment is processed and no payment data collected. Consequently no amount, currency or billing unit is published. A pricing page does exist, titled Pricing | AtomWords and promising simple, transparent pricing for creators, but it renders client-side and displays no plan; it is linked from neither the navigation nor the footer. Both legal documents undertake to be updated, with prior notice to users, before any paid subscription is introduced.

Free tier — browsing, searching, copying prompts and downloading watermarked sample media; all digital, no shipping; reasonable daily per-account download limits that reset at 00
  • 00 UTC
Plan 3
  • No paid plan is published at the time of review
Special offers — No targeted discount for students, jobseekers or non-profits: during the launch phase the whole service is free to everyone, commercial use included
Prices and plans listed above may evolve. Always check the official pricing page before subscribing.
Trust & Privacy

Data, GDPR & hosting

A consolidated view of how AtomWords handles your data.

GDPR overview

Concrete, and unusually specific for a one-person project. The privacy policy grants access, correction, deletion, export, restriction, objection and withdrawal of consent to every user wherever they are — GDPR, CCPA/CPRA and PIPL are all named — with a 30-day response commitment through privacy@atomwords.com. Application data sits in the United States while the Operator works from mainland China; for EU and UK users, registration is presented as explicit consent to that transfer under Article 49(1)(a) and as necessary for the contract under Article 49(1)(b). Breach notification to the supervisory authority within 72 hours is promised where the GDPR requires it, and the minimum age is 16. Two gaps stand out: no EU representative is designated under Article 27 although the Operator is established outside the EU, and no Data Processing Agreement is offered anywhere on the site.

Who owns the data?

AtomWords, described in the privacy policy as the sole Operator, is the data controller for everything collected through atomwords.com. Under section 7 of the terms, the atomic words, prompt directives, commentary, taxonomy, sample media, name and logo belong to the Operator or are licensed from upstream AI providers. Whatever you generate with those prompts, however, is yours: the Operator explicitly claims no ownership of your outputs, subject to your own AI service's terms. Account data — email, display name, favourites, saved sessions, any connected Google sign-in and any API keys — is erased permanently and immediately from the database when you close your account from the settings page.

Reuse rights

Reuse is unusually open. The AtomWords License grants a non-exclusive, worldwide, free licence to copy, modify and use the atomic words, full prompts and sample media, including for commercial purposes and without attribution. No permission is needed; a backlink to atomwords.com is appreciated but never required. Sample images may also be fed back as img2img reference images to guide your own generations, and anything you produce that way belongs to you. Four uses stay off-limits: scraping or compiling the catalogue to replicate or compete with AtomWords, using the content to train or fine-tune an AI model, reselling the prompts as a product, and selling the sample media unaltered as stock. Selling work you generated using the prompts is expressly allowed. On the personal-data side the picture is deliberately narrow: the Operator states it does not sell your data, share it with ad networks, profile you for behavioural targeting or use anything to train AI models, and that no user activity is transmitted to any AI generation service.

Data retention & training

Retention summary
Active account data — email, display name, favourites and settings — is kept for as long as the account exists. Delete the account and everything linked to it is erased permanently and immediately; the operation is irreversible, although a minimal record of the deletion, including the email address, may sit in server logs briefly before being purged. Download history, keyed to an anonymised user id or IP address for fair-use limits, is pruned automatically after 90 days. Search queries are stored as anonymous strings tied to neither account nor IP address and kept to improve search. Aggregate per-entry counts for views, favourites and downloads stay anonymous totals for as long as the content exists. Security and operational logs are kept for a limited period, then purged. The policy carrying these rules is dated 10 June 2026.
Trains on customer data
No
Subprocessors disclosed
Yes
GDPR contact

Hosting summary

Application data is stored in the United States with Hostinger, while the Operator administers the service from mainland China — a split the privacy policy states plainly. Six sub-processors are named: Hostinger International Ltd. for hosting and the database in the United States, Cloudflare, Inc. for CDN, DDoS protection and Turnstile at the global edge, Cloudflare R2 for media storage at the global edge, Resend for transactional email in the United States, Google LLC for OAuth in the United States where you choose Google sign-in, and Umami Software, Inc. for cookieless analytics in the United States and the EU. DNS resolves to 172.67.133.21, announced by Cloudflare's AS13335 anycast network, so that address is a CDN node rather than the origin server. The security measures listed are bcrypt at 12 rounds, HTTPS with HSTS, a strict content security policy, rate limiting and Cloudflare Turnstile. Data may be disclosed on valid legal process, to protect rights and safety, or in a transfer of the service, with notice to the user.

Hosting countries
🇺🇸 United States
Watch-outs

Things to keep in mind

Risks and trade-offs to weigh before adopting AtomWords.

  • Free access is presented as a launch phase: the terms announce paid subscriptions, so a workflow built on free downloads today may cost money tomorrow
  • No EU representative is designated under Article 27 of the GDPR even though the Operator is established in mainland China and serves European users
  • The transfer of data to the United States rests on Article 49(1)(a) consent, a derogation rather than a durable transfer mechanism
  • Chinese law governs the contract and total liability is capped at USD 50, which leaves very little recourse in a professional dispute
  • No Data Processing Agreement is offered, so corporate use needs a legal review before it is signed off
  • The Operator warns that generative output is probabilistic and that sample media may incidentally resemble copyrighted works despite a deny-list and human review
  • Leaning on a ready-made keyword catalogue is comfortable but narrowing: the visual vocabulary you end up with is someone else's, and the habit of testing your own words quietly fades
Setup

Setup & Integrations

Technical difficulty

Essentially none. AtomWords opens in a browser with no installation, no configuration and no payment, and the whole catalogue can be read without an account. Creating one is optional and takes an email address and a display name — a nickname is fine — or a Google sign-in. The only mildly technical step is pasting a copied prompt into your own generator: the trailing parameters shown in the examples follow Midjourney conventions and need adjusting by hand for another engine.

Deployment

Web app

Supported languages

English
Company

Behind AtomWords

Company name
AtomWords
Founded
INFORMATION_NOT_FOUND
Country of origin
🇨🇳 China
UBO
musefeng
UBO country
🇨🇳 China
Domain registrar country
🇱🇹 Lithuania
Legal contact
Support contact

Social

Official links

Resources

All the official URLs gathered for verification and reference.

FAQ

Frequently asked questions

Is AtomWords free?
Yes. Browsing is free and every published entry is fully free: view the example image or clip, copy the complete prompt and download the media, commercial use included. The terms describe this as the free-tier launch phase, with no paid subscriptions yet.
Can I use the images and videos commercially?
Yes. The AtomWords License grants a non-exclusive, worldwide, free licence to copy, modify and use the atomic words, full prompts and sample media, including for commercial purposes and without attribution. A backlink to atomwords.com is appreciated but never required.
What exactly is an AI art keyword?
A single atomic word or short phrase — Dutch angle, chiaroscuro, cinematic lighting — that controls one visual variable in an image generator. Combining a few precise keywords gives far more predictable results than one long prompt paragraph.
Which AI models do these keywords work with?
The vocabulary is model-agnostic. For images the site names Midjourney, Google Nano Banana, GPT Image, FLUX and Stable Diffusion; for video, Sora, Veo, Kling, Runway, Seedance, Hailuo, Midjourney, Pika and Luma. Each entry credits the engine and version behind its example.
Do I need an account?
Not to read. Searching, browsing, copying prompts and downloading media all work without signing in. An account, opened with an email address and a display name or through Google sign-in, only adds a personal favourites list. Downloads are subject to daily per-account limits that reset at 00:00 UTC.
Is my data used to train AI models?
No. The privacy policy states that data is never used to train AI models, never sold, never shared with ad networks and never used for behavioural profiling, and that no user activity is transmitted to any AI generation service.
Where is my data stored, and how long is it kept?
Application data is stored in the United States with Hostinger, while the Operator administers the service from mainland China. Account data lasts as long as the account, download history is pruned after 90 days, and deleting your account erases everything linked to it permanently and immediately.
What is the minimum age to use AtomWords?
Sixteen, or your jurisdiction's digital-consent age if that is higher. The privacy policy adds that data belonging to under-16s is deleted if the Operator learns it has been collected.
Conclusion

Should you pick AtomWords?

AtomWords does one thing, and does it with unusual clarity: it shows what a single prompt keyword looks like, isolated from everything else competing for the model's attention. That is a real contribution to a field where most reference material is a pile of long prompts next to a pile of images, and where the link between the two is left to guesswork.

The value is immediate and free. No account, no payment and no configuration stand between you and a working prompt, and the licence clears commercial use without attribution — a rare combination. The four browsing axes and the reverse dictionary make the same catalogue searchable in ways that suit different habits, and crediting the engine and version on every entry is the kind of small rigour that makes an example reproducible.

The reservations are mostly about age and scale. The site launched on 8 June 2026 and its author says plainly that it is still small; coverage, not concept, is the limiting factor. One industry collection out of several announced is live. Free access is explicitly a launch phase, with a pricing page already waiting in the wings. And behind it all stands a single person, identified only as musefeng, operating from mainland China with no company, no postal address, no Article 27 representative and no data processing agreement, under Chinese law and a USD 50 liability cap.

For an individual creator none of that is disqualifying: the risk is losing a free reference, not losing data, and the privacy practice here is markedly better than the size of the project would suggest. For an organisation putting the tool into a documented workflow, the legal thinness deserves a look first. Treat AtomWords as an excellent bookmark rather than as infrastructure and it delivers far beyond its price.