Stop Generating Images In ChatGPT — Use This Instead
The Playground Guide

Stop generating images in ChatGPT. Use this instead.

Most people open chatgpt.com, type a prompt, and get one image. They spend the next 20 minutes regenerating, fighting the size, asking for a square version, then a portrait, then watching it ignore the request entirely. There is a better way. It costs cents per image. It lives at platform.openai.com, and almost nobody outside developer circles is using it.

This guide shows you what the OpenAI Playground actually is, why it beats the ChatGPT web UI for any serious image work, and gives you 5 detailed prompts you can paste in today for the assets you keep needing.

What is the Playground?

The OpenAI Playground is the developer-facing interface for every model OpenAI ships. Same models, same engine, no chat clutter. The Images section is a stripped-back workspace where every setting that's hidden inside ChatGPT is exposed as a dial you can turn.

You don't need to be a developer to use it. You don't need to write code. You just need a $5 credit on your OpenAI account and a browser.

Why this beats the ChatGPT web UI

01

Generate up to 10 images per prompt

ChatGPT gives you one image. Maybe two if it's in a generous mood. The Playground lets you set n=10 and get ten variations from a single prompt. Pick the best one. Bin the rest. This alone changes everything about how you iterate.

02

Pick your exact model

ChatGPT picks for you and never tells you. The Playground gives you the full lineup: gpt-image-2 (newest, sharpest, up to 4K), gpt-image-1.5, gpt-image-1, and gpt-image-1-mini for cheap drafts. Match the model to the job.

03

Lock the exact size and ratio

Square, portrait, landscape, or a custom dimension up to 3840px wide on gpt-image-2. Click and select. No more arguing with ChatGPT about why it gave you a square when you asked for a portrait.

04

Control quality vs cost

Three quality tiers: low, medium, high. Draft at low for pennies. Final at high. ChatGPT runs everything at one setting and bills you a subscription regardless of what you actually use.

05

Transparent backgrounds

Set background=transparent and the model outputs a PNG with no background at all. Logos, product cutouts, stickers, icons. ChatGPT cannot do this.

06

Output format and compression

PNG, JPEG, or WebP. Set the compression level. Useful when you're generating assets meant for web and need file sizes that don't tank page speed.

07

Saved prompt history

Every prompt you've ever run, listed and re-runnable. Click to send the same prompt to a different model. No more digging through chat threads.

08

Cost transparency

You see exactly what each image costs before you generate. A low-quality 1024x1024 on gpt-image-1 is roughly $0.01. Even high-quality images sit in cents. Compare that to a Midjourney subscription.

Side by side

Images per prompt
ChatGPT: 1
Playground: up to 10
Model choice
Hidden
4 models, you pick
Size control
Vague
Exact, up to 4K
Transparent BG
Not supported
Yes
Cost per image
Subscription
~$0.01–$0.17

Setup in 3 minutes

1
Go to platform.openai.com and sign in with your OpenAI account. Same login as ChatGPT.
2
Add a payment method and load $5 of credit. That's enough for several hundred images at low quality, or 30+ at high.
3
For gpt-image models, you may need to verify your organization. It's an ID check that takes two minutes. Find it under Settings → Organization → Verify.
4
Open platform.openai.com/playground/images and you're in.

The settings that matter

model
gpt-image-2 for finals, gpt-image-1-mini for cheap drafts. Switch with one dropdown click.
size
1024x1024 (square), 1024x1536 (portrait — Instagram, story), 1536x1024 (landscape — thumbnail), or custom on gpt-image-2.
quality
low / medium / high. Drives both cost and sharpness. Use low for exploration.
n
Number of images per generation. Go high (8–10) when exploring. Drop to 1 for final renders.
background
opaque or transparent. Transparent is essential for logos and cutouts.
output_format
PNG, JPEG, or WebP. WebP is your smallest file size for web use.
moderation
auto or low. Set to low for less aggressive filtering on brand and lifestyle work.
★ Pro Tip

Start every project with quality=low, n=10. Browse the grid, pick the strongest composition, then re-run that exact prompt at quality=high with n=1. You'll spend cents instead of dollars on exploration.

5 ready-made prompts

Paste these into the prompt box. Swap in your specifics where bracketed. These cover the five image types people ask for most.

01 / HEADSHOT

Professional headshot

Model: gpt-image-2 · Size: 1024x1536 · Quality: high · n: 4

Editorial professional headshot of [describe person: gender, age range, ethnicity, hair, distinguishing features], wearing a [smart casual outfit — e.g. charcoal crew-neck sweater, white oxford shirt]. Studio lighting with a soft key light from the upper left and a subtle rim light on the right shoulder. Warm neutral background, gentle gradient from sand to off-white, slightly out of focus. Sharp focus on the eyes, natural skin texture preserved, no over-smoothing. Confident closed-mouth smile, head turned 10 degrees to the left, shoulders relaxed. Shot on a medium-format camera with an 85mm lens at f/2.8. Photorealistic, magazine quality, vertical 4:5 framing with the subject's eyes on the upper third line.

Variation: swap "warm neutral background" for "deep navy seamless backdrop" for a more corporate feel, or "exposed brick studio wall" for a creative/founder vibe.

02 / BRAND MOCKUP

Brand product mockup

Model: gpt-image-2 · Size: 1024x1024 · Quality: high · n: 6

Top-down flat lay of a [product — e.g. minimalist cream-coloured cosmetics bottle with matte black label reading "BRAND NAME"] on a textured linen surface in [colour — e.g. warm putty]. Surrounded by complementary props: a small olive branch, two smooth river stones, a folded raw-edge linen napkin, and a single dried palm leaf casting a soft natural shadow. Diffused morning window light from the top right, creating long soft shadows. Muted earth-tone palette throughout: bone, stone, sage, and warm taupe. Editorial photography style, sharp product detail, subtle film grain, shot at f/4 on a 50mm lens. Square framing with the product positioned slightly off-centre on the rule of thirds.

Variation: change the surface to "polished black marble with thin gold veins" and props to "single white orchid, gold-rimmed glass" for luxury positioning.

03 / LIFESTYLE

Editorial lifestyle scene

Model: gpt-image-2 · Size: 1536x1024 · Quality: high · n: 4

Candid lifestyle photograph of [describe subject — e.g. a man in his early 30s with a beard, wearing a forest green knit sweater] working on a laptop at a sunlit cafe table. Warm afternoon light streaming in from a window on the left, creating golden highlights on the rim of a half-finished flat white in a stoneware cup. Open notebook with handwritten notes, a closed paperback book, and a small succulent in a ceramic pot on the table. Shallow depth of field with the subject in sharp focus and the cafe interior softly blurred behind. Cinematic colour grading, slightly desaturated, with warm shadows and clean highlights. Shot on a 35mm lens at f/2.0, photojournalism style, no posed feeling. Landscape framing.

Variation: swap "sunlit cafe" for "minimalist home office with floor-to-ceiling windows" or "co-working space with exposed concrete walls" to shift the brand connotation.

04 / SOCIAL ASSET

Instagram carousel cover background

Model: gpt-image-2 · Size: 1024x1536 · Quality: medium · n: 8

Abstract background image for a social media carousel. Topographic contour line texture in [primary brand colour — e.g. burnt sienna] tones, flowing organic curves resembling a topographic map, set against a [background colour — e.g. cream paper texture]. Subtle paper grain throughout. A soft radial glow in the centre-right, slightly warmer than the base. Negative space on the lower third for headline text to be added later. No text, no logos, no faces. Editorial poster aesthetic, minimal, museum-print quality. Portrait 4:5 framing optimised for Instagram. The composition should breathe — restrained, not busy.

Variation: replace topographic lines with "thin geometric grid lines fading from corner to corner" for a more technical brand, or "soft watercolour wash gradients" for a wellness brand.

05 / HERO BANNER

Website hero / blog header image

Model: gpt-image-2 · Size: 1536x1024 · Quality: high · n: 4

Wide cinematic hero image for a website header. [Describe scene — e.g. a glass-walled modern studio space at dusk, warm amber interior lighting glowing against a deep blue twilight sky outside]. Strong sense of depth with a clear foreground, midground, and background. Atmospheric haze in the background to create separation. Cool-to-warm colour gradient across the frame. Composition leaves clean negative space on the left third for headline text overlay. No people, no readable text, no logos. Shot on a wide 24mm lens at f/8 for full sharpness front to back. Photorealistic, high dynamic range, magazine cover quality. Landscape 3:2 framing.

Variation: use this same structure with "an empty stage lit by a single overhead spotlight" for event landing pages, or "an aerial view of a winding coastal road at golden hour" for travel/lifestyle brands.

Three more things that level you up

  • Use image inputs for style transfer. Paste a reference image into the prompt and tell the model to match its style, lighting, or composition. Far more reliable than describing aesthetics in words.
  • Build a prompt library in Notion or Apple Notes. Once a prompt works, save it with a screenshot of the result. You'll stop reinventing the wheel every Monday.
  • Run drafts on gpt-image-1-mini. It's the cheapest model in the family. Use it to test composition. Only upgrade to gpt-image-2 once the prompt is locked.
★ Cost reality check

Generating ten variations of a high-quality portrait headshot on gpt-image-1 costs around 16 cents total. Midjourney's basic plan is $10 per month. Do the maths on which makes sense for your workflow.

That's the playbook.

If you want the rest of my prompt templates — including the ones I use for client work, ad creative, and content production — drop the keyword below in the comments on the Instagram post that sent you here.

IMAGE

Build smart. Save the prompts.
— Mike

Useful links: platform.openai.com · Image Playground · Image generation docs · Billing & credits

Keep Reading

View more
caret-right
© 2026 Automate and Hustle.
beehiivPowered by beehiiv