Midjourney and DALL-E for Recruiting: A Practical Image Workflow
Recruiting is a visual business long before a candidate reads a job description.
A strong image can carry a role, a team or a culture faster than a paragraph, and text-to-image tools have made producing those images cheap and quick. This guide explains what Midjourney and DALL-E each are, how to operate them, where they fit in a hiring funnel, and where they will let you down. It closes with a review gate you can run before anything reaches a candidate.
What Midjourney and DALL-E each are
> Midjourney is a text-to-image generator that runs inside Discord and returns high-quality visuals within minutes. Its ease of use depends on how precisely you write the prompt; a few attempts are usually enough to reach a usable result. Its public channels also function as a learning space, because you can watch how other people phrase prompts and what those prompts produce.
DALL-E is OpenAI's image model. It builds new images or edits existing ones from a written description, including operations such as removing a background, and it is a creative offshoot of the ChatGPT architecture. For recruiters, the practical output is the same kind of asset: social visuals, job-ad artwork, and branding material that reads as more engaging than stock photography.
Neither tool is a design department. Both convert language into pixels, so the quality ceiling is set by how well you describe the scene, the audience and the constraints.
Operating each tool: the actual steps
The two tools are driven differently, and that difference matters more than any quality claim.
Midjourney, in Discord
- Join the Midjourney Discord server.
- Open a channel intended for new users.
- Start a prompt with the
/imaginecommand. - Submit the prompt; four variations come back.
- Use the U buttons to upscale a chosen variation and the V buttons to request further variations of it.
- Use the redo option to iterate, then save the image you are satisfied with.
DALL-E, in the browser
- Open the DALL-E interface.
- Type your prompt into the input box on the left of the screen.
- Select image size and aspect ratio.
- Generate.
- Review the results and download the one you want.
Aspect ratio is worth deciding before you write the prompt, not after: a wide banner for a job board and a square tile for a social feed are different compositions, and re-prompting for each is faster than cropping a scene that was never framed for it.
Five places these images earn their place
Social posts. Describe the scene you need — a modern office, a team that reflects the diversity of your candidate pool — and let the model render it. A prompt that works as a starting point: *"Two colleagues sketching a product roadmap on a glass wall, warm afternoon light, candid documentary style, shallow depth of field."*
Job advertisements. Feed in the role and the culture, and you get artwork that makes a posting stand out in a crowded feed. Starting prompt: *"A single engineer reviewing code on a laptop at a standing desk, neon accent lighting, editorial poster composition with clear space for a headline."*
Employer branding. Visuals that show how the company works and what it values can make an organisation more attractive to applicants and help hold on to people already there. Starting prompt: *"Depict a lively, upbeat, team-driven workplace that conveys the company's commitment to collaboration, fresh thinking and belonging."*
Candidate experience. Career-site imagery, interview-scheduling screens and recruiting emails all benefit from deliberate visuals, and the cumulative effect is a better impression of the employer. Starting prompt: *"A small group of employees chatting with visitors at a casual open-house table with coffee cups and name badges, friendly natural light, photojournalistic feel."*
Recruitment marketing collateral. Brochures, decks and video assets can be built around generated imagery rather than licensed stock. Starting prompt: *"A flat-lay of a printed hiring brochure on a desk beside a laptop and notebook, bold geometric cover design, top-down studio lighting."*
Each of these prompts is a template, not a formula. Swap the role, the setting and the values clause to match the vacancy, and keep the level of specificity high.
Four limits to plan around
Context blindness. The models work from the words you supply and cannot read the situation around them. An image can therefore be technically clean and still irrelevant to the role or the audience.
Bias carried in from training data. Both models learn from large image-and-text collections that may contain stereotypes, and those can surface in output. If the training material skews toward one group, the generated set will too. Many HR teams now run AI-image detection checks on visual content before publishing, specifically to catch skewed or misleading depictions of workplace diversity.
Bounded originality. The output is constrained by both the training data and the prompt, so results are not reliably novel or imaginative.
Coarse control. You cannot dictate fine detail or composition. You describe, generate, and select — you do not art-direct pixel by pixel.
A review gate before anything goes live
Run every generated image through these checks before it reaches a candidate-facing surface.
- Role relevance. Does the scene actually depict the job, the team or the environment the candidate would join?
- Representation check. Who appears, in what roles, and does the composition repeat a stereotype you would not accept in a photograph?
- Authenticity. Would a candidate who joins feel misled by the image? Generated offices are not your office.
- Rights and disclosure. Confirm your organisation's position on labelling AI-generated visuals and on the tool's own usage terms.
- Composition fit. Is the aspect ratio and crop correct for the placement, or was it generated for a different frame?
- Human sign-off. One named reviewer approves the asset before publication.
A worked example: take the office-scene prompt above, generate four variations in Midjourney, upscale the strongest, then ask the representation question before it goes on a careers page. If the four variations all show the same narrow demographic, the prompt needs a more explicit description of the team — and the result still needs the detector check.
What this changes for a recruiter's day
Image generation does not replace judgement; it moves the effort from production to selection and review. The recruiter's job becomes writing a precise brief, generating enough options to choose from, and applying a consistent standard before publishing. Used that way, these tools shorten the path from an idea to a finished asset without handing over the decisions that shape how an employer is perceived.
| Consideration | Midjourney | DALL-E |
|---|---|---|
| Where you work | Inside Discord | OpenAI's website |
| How you start | /imagine command in a new-user channel | Prompt box on the left of the screen |
| How you iterate | Four variations, then U to upscale or V for more variations of one | Adjust prompt, size and aspect ratio, then regenerate |
| Output control | Redo and re-prompt | Size and aspect ratio selected before generating |
| Best fit | Teams already working in Discord who want rapid variation browsing | Teams wanting a browser workflow with explicit size and ratio settings |
Access routes and interface layouts are as described at the time of writing and may change; confirm current availability before standardising a workflow.
Frequently asked questions
Is Midjourney better than DALL-E?
They differ mainly in access and cost, not in a clear quality ranking. Midjourney runs through Discord, which some users find convenient; DALL-E 2 is reached through OpenAI's website. On cost, OpenAI has offered monthly free credits that allow continued free use of DALL-E 2, while Midjourney has offered 25 free images before charging. Which one suits you depends on your preference and budget, and trying both is the practical way to decide.
How do I get better results from AI art?
Write prompts that are specific and detailed, using precise language and examples of what you want to see. Then experiment with different models and parameters to find the combination that fits your needs.
Will AI replace recruiters?
Full replacement is unlikely, but the role will change substantially. AI can take on repetitive tasks, analyse large data sets and improve efficiency; recruiters remain central to relationship building, reading company culture and making nuanced hiring calls. The realistic outcome is AI supporting human expertise and freeing time for the more strategic and personal parts of the job.
Can I use a generated office image as if it were our real workplace?
No. Generated scenes are illustrations, not photographs of your premises, and presenting them as documentary images misleads candidates. Use them as atmosphere or concept visuals, keep the accompanying copy honest about what the workplace is like, and follow your organisation's disclosure policy for AI-generated content.