Bing Image Creator Setup: 12 Steps, 15 Free a Day [2026]

Microsoft’s free image generator has quietly become one of the most-used AI tools on the web, built straight into Bing and Copilot with no subscription wall in front of it. Yet most people still type a prompt, click “Create,” and never touch the model picker, the editing tools, or the daily creation limits that shape how far their free quota actually goes. This tutorial walks through the current version of Bing Image Creator step by step: how sign-in works, which models you can choose between, how the boost and fast-creation system works, how to edit an uploaded photo, and what to do when a prompt gets blocked. Every setting described here was pulled directly from Microsoft’s live help documentation at bing.com/images/create/help as of October 2026, not from a 2023 screenshot that no longer matches the product.

·

Google · Preferred Sources

Don't miss new tech stories on Google

Add Tech Insider once in the Google app and our stories appear in your news suggestions.

Add Now

What Bing Image Creator Is in October 2026

Bing Image Creator is Microsoft’s free, browser-based AI image generator, reachable at bing.com/images/create and mirrored inside Copilot at copilot.microsoft.com/images/create. Both URLs load the same tool under the hood, so whether you land on it through Bing search, the Edge sidebar, or the Copilot app, you’re working with the same model picker and the same daily creation limits.

The product has changed shape more than once since it launched. The original 2023 version ran on OpenAI’s DALL-E 3, a model Microsoft’s own Azure documentation confirms was retired for new deployments on March 4, 2026. Bing Image Creator’s current help page no longer mentions DALL-E at all. Instead, it lists two selectable models: MAI-Image-2e for text-to-image generation, and MAI-Image-2.5-Flash, which handles both generation and any editing you do to an uploaded photo. “MAI” is Microsoft’s own in-house AI model family, developed by Microsoft AI rather than licensed from OpenAI, and it’s the same lineage behind the separate MAI-Image-2.5 Pro model Microsoft offers to Azure developers. The consumer tool and the developer API are related but not identical products, and that distinction matters once you get to the API section below.

Where the Help Documentation Actually Lives

Every specific number in this tutorial — the daily creation limits, the model names, the data-retention windows — comes from Microsoft’s own live help page at bing.com/images/create/help, which is the numbered FAQ panel Microsoft ships directly inside the product. It’s worth bookmarking that page separately from this tutorial, because Microsoft updates it in place whenever the underlying model or limits change, without necessarily issuing a press release. A tutorial frozen at a single point in time will drift out of date faster than that help page does.

Where Bing Image Creator Sits in the Broader Market

For context on where this fits against the rest of the image-generation market, Microsoft’s own tool competes directly with OpenAI’s ChatGPT-integrated image models and Google’s Gemini-based generators, both of which charge for their fastest tiers. Bing Image Creator’s pitch is simple: it is free for anyone with a personal Microsoft account, with no credit card and no trial period that quietly expires. Other companies have taken a different route to the same “make it effortless” goal — xAI, for instance, built voice-driven prompting directly into its own Grok Imagine feature rather than relying on typed text alone. Microsoft hasn’t gone that direction with Bing Image Creator, but the comparison is a useful reminder that “free” and “fast” are two separate axes competitors are optimizing on, and Bing Image Creator currently wins mainly on the first one.

Prerequisites and Account Requirements

Before generating anything, confirm you meet these requirements. They are stricter than most people assume, and the most common support complaint (“why can’t I access this at work”) traces back to the account type.

• Account type: A personal Microsoft Account (MSA) — the same login used for Outlook.com, Xbox, or a personal OneDrive. Bing Image Creator explicitly does not support Microsoft Entra ID accounts, which is the identity system behind most work and school logins.
• Browser: Any current version of Microsoft Edge, Google Chrome, Safari, or Firefox. No extension or plugin is required.
• Device: Desktop or laptop via the website, or the Bing mobile app (iOS/Android) via its built-in Image Creator miniapp.
• Region: Availability and some features vary by country; a small number of markets have additional age or consent prompts before biometric-related features (like photo editing) unlock.
• Cost: $0 for standard use. Optional Microsoft Rewards points can be redeemed for extra fast-generation boosts, but nothing is required to use the tool at all.

If you plan to call an image-generation model programmatically rather than through the web UI, you’ll need a separate Azure subscription and an Azure OpenAI resource — that path is covered in Step 10.

Step 1: Sign In to Bing Image Creator

Bing Image Creator will not generate anything until you’re signed in with a personal Microsoft account. Microsoft’s help documentation lays out the exact flow:

1. Go to bing.com/images/create, or open the Bing mobile app and navigate to the Image Creator miniapp.
2. Click “Sign in to Create” in the top-right corner of the page (on mobile, this is the “Sign in & Create” button on the right side of the screen).
3. Enter your existing Microsoft Account credentials, or select “Create one” if you don’t have an MSA yet. Creating a new MSA takes under two minutes and only needs an email address and a password.
4. If you reach Image Creator through Copilot Search or Bing Search rather than the direct URL, you’ll be prompted to sign in before your first generation request goes through.

Once signed in, the prompt box, model selector, and your remaining boost count all appear on the same screen — there’s no separate “dashboard” to find.

Step 2: Choose Between MAI-Image-2e and MAI-Image-2.5-Flash

Before typing a prompt, check the model selector sitting above the prompt box. Bing Image Creator currently offers two choices, and picking the wrong one for the job wastes a creation:

Model	Best for	Supports image upload/editing?	Counts as
MAI-Image-2e	Pure text-to-image generation	No	One creation per prompt, even if multiple images are returned
MAI-Image-2.5-Flash	Text-to-image generation and editing an uploaded photo	Yes	One creation per prompt or edit action

The practical rule from Microsoft’s own documentation: if you upload a photo to edit, Bing Image Creator forces you onto MAI-Image-2.5-Flash automatically — MAI-Image-2e is not available once an image is attached to the request. If you’re only typing text prompts with no upload, either model works, and it’s worth testing both on the same prompt since they render detail and lighting differently.

Step 3: Write a Prompt That Actually Works

Microsoft’s help documentation is specific about why vague prompts underperform: “Creating an image with Bing Image Creator works differently from searching for an image on Bing. For the best results, be highly descriptive and imaginative.” The same documentation lays out a five-part structure for a usable prompt:

• Be specific — swap a bare noun like “astronaut” for a fully described subject.
• Use adjectives — describe color, texture, and material so the model has something concrete to render.
• Include action — state what the subject is doing, not just what it is.
• Set the scene — name the environment or background.
• Specify style and mood — terms like “vintage,” “bright,” “dark,” or “photorealistic” steer the overall look.

Applying that structure to a real prompt looks like this:

Weak prompt:
"astronaut in space"

Strong prompt (following Microsoft's five-part structure):
"A realistic photo of an astronaut in a reflective white
spacesuit, helmet visor hiding their face, floating in front
of a star-filled galaxy with soft blue and purple nebula
light, dramatic film-stock lighting, bright saturated colors,
camera angle from slightly below looking up."

Also worth knowing before you spend a creation: Bing Image Creator caches results. If your prompt closely matches one recently submitted by you or another user, the tool may return a previously generated “cached image” instead of rendering a brand-new one, to keep response times fast. If a result looks suspiciously fast or identical to something you’ve seen before, that’s why.

Step 4: Set Your Aspect Ratio and Hit Create

Type your prompt into the box, choose an aspect ratio from the options presented in the interface (standard square and widescreen/portrait ratios are both available), and click “Create.” A single click can return multiple image variations on the same prompt, and Microsoft’s billing logic counts that entire batch as one creation against your daily limit, not one creation per image returned.

Generation typically finishes in under 30 seconds on the fast tier described in Step 5. If it’s taking noticeably longer, you’ve likely dropped to standard speed because your daily fast-creation allowance ran out.

Step 5: Understand Your Daily Creation Limits

This is the part most tutorials get wrong because the limits have changed more than once since 2023. Here’s what Microsoft’s live help page states as the current policy:

Limit type	Current allowance	Resets
Fast creations per day	15	Every 24 hours
Standard (slower) creations	Unlimited	N/A — no daily cap
Total prompts per 24 hours (includes “Edit image” clicks)	200	Every 24 hours
Extra fast-creation boosts	Redeemable with Microsoft Rewards points	Applied immediately on redemption

In plain terms: every account gets 15 fast image creations per day at no cost, and those replenish automatically the next day. Once that allowance is used up, Bing Image Creator doesn’t cut you off — it drops you to standard speed, which is slower but has no daily ceiling. You can keep creating all day at standard speed for free. If you want to stay on fast speed past your 15 free creations, you redeem Microsoft Rewards points for additional “boosts,” which is Microsoft’s term for a fast-creation credit. Rewards points come from everyday Bing searches and other Microsoft Rewards activities — there’s no way to buy boosts with cash directly through Image Creator.

Separately, there’s a hard ceiling of 200 total prompts in any 24-hour window, and that count includes every time you click “Edit image” on an existing result — not just brand-new generations. Hit that number and you wait for the daily reset regardless of boosts or Rewards points.

How Microsoft Rewards Points Turn Into Boosts

Microsoft Rewards is a separate, free loyalty program tied to the same Microsoft Account you use to sign in to Bing Image Creator. Any Rewards points you’ve already accumulated from everyday Bing searches, Microsoft Edge usage, or other Rewards-eligible activity can be redeemed directly inside the Image Creator interface for extra fast-creation boosts once your 15 daily allowance runs out. If your Rewards balance is too low to cover a boost, the tool shows an explicit error telling you to keep earning points through Bing search or other daily Rewards activities before trying again — it won’t let the redemption go through partially.

There’s no option to pay cash for a boost directly through Image Creator. The only on-ramp is the Rewards points balance tied to your account, which makes the economics of “free” slightly different depending on how actively you already use Bing and Edge day to day — a heavy Bing search user effectively banks boosts passively, while someone signing up purely to generate images starts from zero Rewards points and relies entirely on the 15 free fast creations plus unlimited standard speed.

Step 6: Upload and Edit an Existing Photo

Beyond text-to-image, Bing Image Creator can edit a photo you already have — swapping a background, changing a hairstyle, or applying a new style to an existing shot. To do this:

1. Click the upload/attach icon next to the prompt box and select an image from your device.
2. The model selector automatically switches to MAI-Image-2.5-Flash — this is the only model that accepts an uploaded image.
3. Describe the edit you want in the prompt box (e.g., “give this person a new hairstyle” or “replace the background with a sunset over mountains”).
4. Click “Create” (or “Edit image” on an existing result) to generate the edited version. Each edit counts toward both your daily creation allowance and your 200-prompt ceiling.

Microsoft’s documentation is explicit about what happens to uploaded photos. Any image you upload is processed only to fulfill your request — it is not used to train models or personalize your experience, and Microsoft states it will not attempt to identify faces in uploaded images. Images containing detected faces are deleted within 30 days of upload; images without a detected face may be retained for up to 18 months for general service improvement. If you’re uploading a photo of another person, you need their consent first — Microsoft’s Code of Conduct puts that responsibility on the uploader, not the platform.

In some regions, before any photo-editing feature activates, you’ll see a consent prompt because the edit may involve processing biometric data like facial features. That consent, once given, covers all of your Bing Image Creator sessions that involve an uploaded image — you won’t be asked again on every single edit.

Step 7: Add or Remove the AI Watermark

Bing Image Creator supports an optional visible watermark that flags an image as AI-generated to anyone who sees it. The setting lives in the Image Creator interface as a toggle labeled “Include a visible watermark on images I create.” It’s off by default on most accounts, applies only to images created during your current session once enabled, and can be switched back off at any time. All images carry an invisible Content Credentials (C2PA) metadata tag regardless of whether you turn the visible watermark on — that metadata persists even if the visible mark doesn’t.

Step 8: Download, Share, or Discard Your Images

Once an image generates, hover or tap it to reveal the download icon, which saves it locally at the resolution Bing Image Creator rendered it at. There’s a direct “Share” option for posting to social platforms without downloading first. If a result misses the mark, there’s no penalty for discarding it beyond the creation it already cost — Bing Image Creator doesn’t store a “trash” or undo history, so once you navigate away from a result you haven’t downloaded, treat it as gone.

Step 9: Work in Other Languages

Bing Image Creator’s prompt box accepts input in more than 100 languages, according to Microsoft’s own help documentation, which points users to Microsoft Translator for the full supported-language list. You don’t need to translate your prompt to English first — the model processes the prompt in its original language, which matters if you’re generating culturally specific scenes where an English paraphrase would lose nuance.

Step 10: Access Microsoft’s Image Models Programmatically via Azure OpenAI

Bing Image Creator itself has no public developer API — you can’t call it from a script the way you call an OpenAI or Google endpoint. If you need programmatic image generation on Microsoft’s stack, the supported path is Azure OpenAI, and it runs on a different model family than the consumer tool. Microsoft’s Azure AI Foundry documentation confirms that DALL-E 3 was retired for new deployments on March 4, 2026, and that existing DALL-E 3 deployments are now non-functional. New projects are pointed to the GPT-Image model family instead, which supports three output resolutions: 1024×1024, 1024×1536, and 1536×1024.

Setting Up Your Azure Resource

Before writing any code, you need three things provisioned in the Azure portal: an active Azure subscription, an Azure OpenAI resource created in a region that supports image generation, and a GPT-Image model deployment inside that resource through Azure AI Foundry. Microsoft’s own Azure OpenAI quickstart guide walks through resource creation and deployment naming if you haven’t provisioned one before. Once deployed, you’ll need two values for every request: your resource’s endpoint URL and an API key, both visible on the resource’s “Keys and Endpoint” page in the Azure portal.

Making Your First Request

A basic request against an Azure OpenAI image deployment looks like this:

curl -X POST "https://YOUR_RESOURCE.openai.azure.com/openai/deployments/YOUR_DEPLOYMENT/images/generations?api-version=2026-03-01" \
  -H "Content-Type: application/json" \
  -H "api-key: $AZURE_OPENAI_API_KEY" \
  -d '{
    "prompt": "A stenciled street-art style collage of London landmarks",
    "size": "1024x1024",
    "n": 1
  }'

Replace YOUR_RESOURCE with your Azure OpenAI resource name and YOUR_DEPLOYMENT with the name you gave your GPT-Image deployment inside Azure AI Foundry. Treat the consumer tool and the API as separate products built on related but distinct Microsoft model families, not two doors into the same system.

If you’d rather call the API from a Node.js backend than shell out a curl command, the equivalent request looks like this:

const endpoint = "https://YOUR_RESOURCE.openai.azure.com";
const deployment = "YOUR_DEPLOYMENT";
const apiVersion = "2026-03-01";
const apiKey = process.env.AZURE_OPENAI_API_KEY;

async function generateImage(prompt, size = "1024x1024") {
  const url = `${endpoint}/openai/deployments/${deployment}/images/generations?api-version=${apiVersion}`;
  const res = await fetch(url, {
    method: "POST",
    headers: {
      "Content-Type": "application/json",
      "api-key": apiKey,
    },
    body: JSON.stringify({ prompt, size, n: 1 }),
  });
  if (!res.ok) {
    throw new Error(`Azure OpenAI request failed: ${res.status}`);
  }
  const data = await res.json();
  return data.data[0].url;
}

generateImage("A stenciled street-art style collage of London landmarks")
  .then((url) => console.log("Image URL:", url))
  .catch((err) => console.error(err));

What the Response Actually Looks Like

A successful call to the generations endpoint returns a JSON body with a data array containing one entry per requested image. Here’s the shape of a typical response:

{
  "created": 1759996800,
  "data": [
    {
      "url": "https://your-resource.blob.core.windows.net/images/abc123.png",
      "revised_prompt": "A stenciled street-art style collage depicting London landmarks including Big Ben, the Tower Bridge, and a red double-decker bus, rendered in bold spray-paint texture with high contrast."
    }
  ]
}

Two things worth noting in that response: the url field is a temporary, time-limited link to the generated image hosted on Azure storage, so download or re-host it promptly if your application needs to keep it long-term. The revised_prompt field shows the model’s own expanded interpretation of what you asked for, which is useful for debugging why an output didn’t match your intent — if the revision drifted far from your original wording, that’s usually where the mismatch started.

Step 11: Build a Simple Batch Generation Script

If you’re generating images for a project with more than a handful of prompts, scripting against Azure OpenAI’s REST endpoint saves far more time than clicking through the Bing Image Creator UI one prompt at a time. Here’s a minimal Python loop that reads prompts from a file and generates an image for each:

import os
import time
import requests

ENDPOINT = "https://YOUR_RESOURCE.openai.azure.com/openai/deployments/YOUR_DEPLOYMENT/images/generations"
API_VERSION = "2026-03-01"
API_KEY = os.environ["AZURE_OPENAI_API_KEY"]

def generate_image(prompt, size="1024x1024"):
    response = requests.post(
        f"{ENDPOINT}?api-version={API_VERSION}",
        headers={"Content-Type": "application/json", "api-key": API_KEY},
        json={"prompt": prompt, "size": size, "n": 1},
        timeout=60,
    )
    response.raise_for_status()
    return response.json()["data"][0]["url"]

with open("prompts.txt") as f:
    prompts = [line.strip() for line in f if line.strip()]

for i, prompt in enumerate(prompts):
    url = generate_image(prompt)
    print(f"[{i+1}/{len(prompts)}] {prompt[:50]}... -> {url}")
    time.sleep(1)  # stay under per-minute rate limits

Add retry logic around rate-limit (HTTP 429) responses before running this against more than a few dozen prompts — Azure OpenAI enforces per-minute request and token quotas that vary by your deployment’s assigned tier.

Step 12: Know What Gets Blocked and Why

Bing Image Creator enforces a Content Policy and a separate Code of Conduct, both linked directly in the tool’s footer. In practice, the restrictions that trip up new users most are around uploaded photos of real people and public figures. Microsoft’s documentation is direct about the baseline rule: “You may only upload images of people with their consent.” Requests that attempt to generate sexualized, violent, or clearly identity-infringing content involving a real, identifiable person are the most common source of outright refusals, and repeated attempts can lead to account restriction rather than just a blocked single prompt.

Copyrighted characters, logos, and trademarked designs are a second common trigger — the model will often refuse or substitute a generic alternative rather than reproduce a protected design closely. If a prompt gets rejected without explanation, the most reliable fix is removing anything that names a real identifiable person, a living public figure, or a specific branded character, and replacing it with a general descriptive phrase instead.

Common Pitfalls When Using Bing Image Creator

These are the mistakes that account for most of the frustration reported around the tool, based on the gap between what users assume and what Microsoft’s documentation actually states:

• Trying to sign in with a work account. Bing Image Creator does not support Microsoft Entra ID (work/school) accounts at all. If your organization’s email is your only Microsoft login, you’ll need a separate personal MSA.
• Assuming a multi-image result costs multiple creations. One click that returns a batch of image variations still counts as a single creation against your daily fast-creation limit.
• Not noticing the speed drop after 15 fast creations. Generation doesn’t stop — it silently switches to the slower standard tier, which some users mistake for the tool being broken.
• Forgetting the 200-prompt ceiling includes edits. Repeatedly clicking “Edit image” on the same result burns through your 24-hour prompt allowance just as fast as new generations do.
• Uploading a photo of someone else without checking consent rules. This is both a policy violation and, depending on the subject and the edit requested, can trigger an account restriction.
• Expecting a public API. There is no developer API for the consumer Bing Image Creator product — Azure OpenAI is the only supported programmatic path, and it uses a different model deployment entirely.
• Assuming the visible watermark setting persists across sessions by default behavior. It applies to the current session once enabled and needs to be checked each time if you need consistent watermarking across a project.
• Re-submitting a near-identical prompt and assuming it’s a fresh generation. The caching system may silently return an earlier cached result to keep response times fast.

Troubleshooting Bing Image Creator

Below are the most common problems reported with the current version of the tool and the fixes that actually address them:

• “Sign in to Create” button does nothing. Clear your browser cache or try an incognito/private window — stale session cookies from a previous Microsoft account are the most common cause.
• Generation suddenly takes 60+ seconds instead of ~30. You’ve exhausted your 15 daily fast creations and dropped to standard speed. Wait for the next day’s reset or redeem Microsoft Rewards points for more boosts.
• “You don’t have enough points” error when redeeming boosts. This means your Microsoft Rewards balance is below the point cost of a boost — earn more through Bing searches or other Rewards activities before retrying.
• Upload button is missing or grayed out. Confirm your model selector is set to MAI-Image-2.5-Flash; MAI-Image-2e does not support image upload and will hide the attach option.
• A prompt returns a blank or error result with no explanation. This usually indicates a Content Policy block rather than a technical failure. Remove references to real identifiable people, public figures, or branded characters and resubmit.
• Images look identical across slightly different prompts. You’re likely hitting the cached-image system. Add more specific, less commonly used descriptive details to force a fresh generation.
• Account flagged or restricted after uploading a photo. This typically follows an uploaded image that violates the consent or prohibited-content rules in the Bing Image Creator Code of Conduct. Review the linked Code of Conduct before appealing.
• Azure OpenAI request returns HTTP 429. You’ve exceeded your deployment’s per-minute rate limit. Add exponential backoff to your script and confirm your deployment’s assigned quota tier in Azure AI Foundry.
• Azure OpenAI request returns a model-not-found error. Your deployment is likely still pointed at a retired DALL-E 3 model. Redeploy using a current GPT-Image model and update your deployment name in the request URL.

Image Rights: What You Can Actually Do With the Output

A question that comes up constantly once people start using Bing Image Creator for anything beyond personal fun: who owns the image once it’s generated, and can it go into a commercial project? Microsoft’s Services Agreement grants users broad rights to use the images they generate, including for commercial purposes, but that permission comes with caveats worth taking seriously before you drop a generated image into a client deliverable or a paid product listing.

First, you’re still bound by the same Content Policy and Code of Conduct restrictions at generation time — an image that slipped through moderation but depicts a real person, a trademarked character, or another party’s intellectual property doesn’t become safe to use commercially just because the tool produced it. Second, because the underlying training data and output behavior of MAI-Image-2e and MAI-Image-2.5-Flash aren’t identical to any other generator, there’s no guarantee that a visually similar image hasn’t also been generated by someone else — AI image generators are not guaranteed to produce unique output, and that matters for exclusivity claims in commercial contexts. Third, the invisible Content Credentials metadata embedded in every image (discussed in Step 7) persists through most editing software, so if your downstream use case requires stripping AI-origin metadata for a client who specifically wants to avoid AI disclosure, you’ll need to do that deliberately rather than assuming it’s absent by default.

For anything going into paid advertising, published editorial content, or a commercial product, it’s worth reading Microsoft’s current Services Agreement directly rather than relying on a summary, since usage terms for AI-generated content have shifted more than once across 2025 and 2026 as the underlying models changed.

Advanced Tips for Better Results

Once the basic workflow is routine, a few less-obvious habits noticeably improve output quality and stretch your daily allowance further:

• Test both models on the same prompt. MAI-Image-2e and MAI-Image-2.5-Flash render lighting and texture differently enough that the same prompt can produce visibly different results. Running both costs two fast creations but often surfaces the better option faster than tweaking a single model’s output repeatedly.
• Save standard-speed creation for exploration, fast creation for finals. Since standard speed has no daily cap, use it to iterate on prompt wording, then switch to a fast creation only once you’ve landed on wording you’re confident in.
• Front-load the style and mood terms. Models weight earlier tokens in a prompt more heavily in practice — leading with “photorealistic, dramatic lighting” before the subject description tends to produce more consistent style adherence than appending it at the end.
• Use the edit flow instead of re-prompting from scratch. If a generation is 80% right, clicking “Edit image” and describing the specific change is often more reliable than writing an entirely new prompt and hoping for the same composition.
• Bank Microsoft Rewards points before a high-volume day. If you know you have a project needing more than 15 fast creations, redeeming boosts in advance avoids getting stuck on standard speed mid-project.
• For production pipelines, skip the consumer tool entirely. If you’re generating more than a few dozen images regularly, the 200-prompt daily ceiling and lack of an API make Bing Image Creator a poor fit — move to Azure OpenAI’s GPT-Image deployment from the start.

Complete Working Example: A Daily Social Content Workflow

Here’s a full workflow combining the consumer tool and the API path, suited to a small team that needs a handful of custom images daily without paying for a subscription:

1. Sign in to bing.com/images/create with a personal MSA each morning.
2. Draft 3-5 prompts offline first, following the five-part structure from Step 3, so you’re not burning fast creations mid-thought.
3. Generate each on standard speed first to check composition, then re-run the two or three best on fast speed for the final crop.
4. Toggle the visible watermark on for anything going into a public-facing deliverable where AI disclosure is required by your organization’s policy or by platform rules.
5. Download finals directly from the result grid — no separate export step is needed.
6. If your daily need ever exceeds 15 fast creations plus comfortable standard-speed waiting, that’s the signal to provision an Azure OpenAI GPT-Image deployment and switch the team to the script from Step 11 instead of manually repeating the web workflow.

This mirrors the two-tier approach Elon Musk described for a different image tool, where he recommended on X that “by far the best way to use @Grok Imagine is to speak, rather than type. It is a super smooth flow of consciousness if you do that,” pointing to a broader shift across image generators toward faster, less deliberate input methods once a workflow becomes routine. Bing Image Creator hasn’t added voice input as of this writing, but the underlying principle — reduce friction between the idea and the prompt — applies just as well to typing out the five-part structure above until it becomes automatic.

How Bing Image Creator Compares to the Azure Developer Path

Factor	Bing Image Creator (consumer)	Azure OpenAI GPT-Image (developer)
Cost	Free	Pay-per-image via Azure billing
Account needed	Personal Microsoft Account (MSA)	Azure subscription + Azure OpenAI resource
Daily limit	15 fast + unlimited standard + 200 prompts/24h	Governed by deployment rate limits, not a fixed daily count
Models available	MAI-Image-2e, MAI-Image-2.5-Flash	GPT-Image family (DALL-E 3 retired March 4, 2026)
Scriptable/automatable	No public API	Yes, full REST API
Image upload/editing	Yes, via MAI-Image-2.5-Flash	Depends on specific model/endpoint support
Best suited for	Individuals, occasional use, quick social content	Apps, automated pipelines, high-volume generation

Frequently Asked Questions

Is Bing Image Creator actually free?
Yes. There’s no subscription, trial period, or credit card requirement. You get 15 fast creations per day plus unlimited slower standard-speed creations, all at no cost with a personal Microsoft account.

Why can’t I sign in with my work email?
Bing Image Creator only supports personal Microsoft Accounts (MSA). Work and school accounts use Microsoft Entra ID, which the tool explicitly does not support. You’ll need a separate personal MSA.

What happened to DALL-E in Bing Image Creator?
Microsoft’s own Azure documentation confirms DALL-E 3 was retired for new deployments on March 4, 2026. The consumer tool’s current help documentation lists Microsoft’s own MAI-Image-2e and MAI-Image-2.5-Flash models instead, with no remaining mention of DALL-E.

What’s the difference between the two models I can pick?
MAI-Image-2e handles text-to-image generation only. MAI-Image-2.5-Flash handles both generation and editing — it’s the only option once you upload a photo, since MAI-Image-2e doesn’t support image input at all.

Does Bing Image Creator have a public API?
No. The consumer product has no developer API. Developers who need programmatic image generation on Microsoft’s infrastructure should provision an Azure OpenAI resource and deploy a current GPT-Image model instead.

What happens to photos I upload for editing?
Uploaded images are processed only to fulfill your specific request. Microsoft states it does not use uploads to train models or personalize your experience, and will not attempt to identify faces in them. Images with a detected face are deleted within 30 days; images without one may be retained up to 18 months for general service improvement.

Can I use Bing Image Creator in languages other than English?
Yes. Microsoft’s documentation states the prompt box supports over 100 languages, with the full list available through Microsoft Translator.

Why did my prompt get blocked with no explanation?
This is almost always a Content Policy enforcement rather than a bug. The most common triggers are prompts referencing a real identifiable person without consent, sexualized or violent depictions of real people, or close reproduction of a copyrighted character or logo. Remove the specific named reference and resubmit with a general description instead.

How do I get more than 15 fast creations a day without waiting?
Redeem Microsoft Rewards points for extra boosts directly inside the Image Creator interface. There’s no cash purchase option — the only way to add fast creations beyond the daily 15 is spending points earned through Bing searches and other Rewards-eligible activity, or switching to the slower, uncapped standard-speed tier.

Is there a difference between using Bing Image Creator through Bing.com versus Copilot?
No. Both bing.com/images/create and copilot.microsoft.com/images/create load the same underlying tool, the same model selector, and the same daily limits. Which URL you land on usually just depends on whether you arrived through Bing search, the Edge sidebar, or the Copilot app directly.

Related Coverage

• Z-Image Turbo Setup: 6B Params, 14 Steps [2026]
• Claude Opus 5.5 API Setup: 13 Steps, $4/$20 per 1M [2026]
• Claude Sonnet 5.5 API Setup: 13 Steps, $2/$10 per 1M [2026]
• GLM-5.3-Flash Setup: API vs Self-Host in 13 Steps [2026]
• 5 Best Free AI Image Generators in 2026: Tested and Compared