How to Generate AI Images on magicdoor.ai

Direct answer: To generate an AI image on magicdoor.ai, open the image workspace, describe what you want, choose an image model, and generate. If you already have an image, upload or paste it into the same prompt-first workflow and describe the edit. You do not need to switch to a separate editing mode.

This guide covers the practical decisions that affect the result and cost: which model to start with, what to put in the prompt, when to edit instead of regenerate, and when to upscale.

Fact-checked September 16, 2026. The image model names, editing capabilities, reference limits, and prices below were checked against the current magicdoor.ai image registry. The video handoff later in this guide was checked against the current video registry and quote logic.

Generate your first image

1. Open the magicdoor.ai image workspace.
2. Write a concrete description of the subject, setting, style, composition, and mood.
3. Choose a model based on the job and the displayed cost.
4. Review the result before generating another version.
5. If most of the image already works, upload or paste it and request a targeted edit instead of starting over.
6. Upscale only after the content is approved.

You can also ask a chat model to generate an image. magicdoor.ai can route image-generation requests automatically, which is useful when a conversation has already helped you develop the brief.

Choose the right image model

The active selector contains 8 image models. The prices below are the current per-image rates on magicdoor.ai:

Start with the lowest-cost model that fits the task, not simply the cheapest entry in the table. Recraft Upscaler is less expensive than every generator because it does not create a new image. Recraft V4.1 generates images but does not accept image edits.

For more detailed tradeoffs, use the image model comparison.

Write a prompt that answers the important questions

A useful image prompt usually specifies:

• Subject: What must appear in the image?
• Setting: Where is it, and what belongs in the background?
• Style: Photograph, editorial illustration, minimal graphic, or another clear direction?
• Composition: Close-up or wide, centered or off-center, and where should empty space remain?
• Lighting and mood: Soft daylight, dramatic contrast, muted colors, or another deliberate choice?
• Constraints: Which colors, objects, text, or product details must stay unchanged?

Instead of “a coffee cup,” try: “A ceramic coffee cup on a pale wood desk, clean product photograph, soft morning window light, centered composition, plain background, no extra objects.”

The goal is not maximum prompt length. It is enough information to judge whether the output followed the brief.

Use prompt enhancement before spending on retries

AI-powered image prompt enhancement uses Claude to add useful detail to a rough request. Review the enhanced version and make sure it preserves the non-negotiable parts of your idea before generating.

Prompt enhancement is most useful when the concept is still vague. If an existing image is already close, describe the specific change and use an editing-capable model instead.

Generate, edit, or upscale?

Use this decision rule:

• Generate when the subject, composition, or overall direction is still undecided.
• Edit when the base image is worth keeping and a specific object, color, background, or region needs to change.
• Upscale when the content is final and only resolution needs improvement.

For an edit, upload or paste the source image and write the change directly, such as “replace the gray background with warm white; keep the product shape, label, and shadows unchanged.” This modeless flow also supports style-reference and other image-input intentions, so the prompt should say what you want the uploaded image to do.

For a deeper editing workflow, read how to generate, edit, and upscale without overpaying.

A cost-aware workflow

1. Clarify the brief in chat or with prompt enhancement.
2. Explore with Seedream 5 Pro at $0.045 or Google Nano Banana 2 at $0.067 when either fits the job.
3. Switch models only when the task changes or the current model misses a specific requirement.
4. Preserve a good base image with a targeted edit instead of automatically regenerating it.
5. Use Recraft Upscaler at $0.006 only after the image is approved.

Your actual total depends on how many generations and edits you run. Live cost monitoring keeps that spend visible. magicdoor.ai has no rate limits or cooldowns, but usage is not unlimited: each image is charged to your balance at its model rate.

The base subscription is $6/month and includes $1 in credits. Further image usage comes from usage-based top-ups, so occasional and bursty image work does not require a separate flat-rate image subscription. For the full pricing mechanics, see pay-as-you-go AI image generation.

Common mistakes

• Paying for premium generations before the visual direction is settled.
• Regenerating the full image when a targeted edit would preserve what works.
• Treating Recraft Upscaler as a generator or general editor.
• Trying to edit with Recraft V4.1, which does not support image input.
• Upscaling before the last content edit.
• Leaving important constraints implicit, then paying to correct avoidable changes.

Turn an approved image into a video

Once the still is finished and saved, move to the modeless video workspace when the next task is motion:

1. Choose one approved saved image or upload it as the optional starting image.
2. Describe subject, environmental, and camera motion instead of redesigning the still.
3. Choose Wan 3, Kling 3.0, or FLUX 3 and 720p or 1080p. Wan 3 and Kling 3.0 follow the starting image framing, so prepare it at the target 16:9, 9:16, or 1:1 ratio; FLUX 3 keeps the framing selector available.
4. Review the full quote and start with a five-second test. Each new creative attempt has its own quote; you pay only for a saved video, while provider or delivery failures release the reservation.
5. Inspect product, logo, text, face, and crop fidelity. A starting image guides the opening frame but does not guarantee exact preservation.

Only active or trialing subscribers can generate videos. Usage comes from included credit or account balance; up to three jobs can reserve funds concurrently, and finished videos remain private until deleted. Anonymous visitors are redirected to sign in.

Read the image-to-video guide for the general workflow, how to turn a product photo into a video for product-specific checks, or the AI video prompts guide for motion briefs.

Create a video

FAQ

Which image model should I start with?

Start with the lowest-cost model and tier that match the job. GPT Image 2.5 Flare and Sunburst start at $0.012 on Low. Flux 2 Pro costs $0.03 and supports generation and editing. Seedream 5 Pro costs $0.045 and supports up to 10 reference images. If you already have a finished image that only needs more resolution, use Recraft Upscaler at $0.006 instead of generating again.

Can I edit images after generating them?

Yes. Upload or paste the image, describe the change, and choose an editing-capable model. Google Nano Banana 2, Google Nano Banana Pro (2K), GPT Image 2.5 Flare, GPT Image 2.5 Sunburst, Seedream 5 Pro, and Flux 2 Pro support editing.

How much does AI image generation cost on magicdoor.ai?

Active generation models cost from $0.012 to $0.50 per image. Recraft Upscaler costs $0.006 per image but only upscales an existing image. Each retry or edit is another charged run, so the total depends on the workflow.

Do I need a separate edit mode?

No. Upload or paste an image into the existing image workspace and describe what you want in the prompt. The same modeless flow can handle edits, style references, and other image-input intentions.

Are there daily image caps or cooldowns?

No. magicdoor.ai has no rate limits or cooldowns. You pay for usage from your balance at the displayed model rate.

Ready to make the first draft? Open the image workspace, or compare the current image models before choosing one.

Sources

Accessed September 16, 2026.

• The current magicdoor.ai image model registry and image cost logic for selector membership, editing support, reference limits, and customer prices.
• The current magicdoor.ai video model registry and quote function for the saved-image handoff, exposed controls, access rules, and customer prices.
• magicdoor.ai pricing for subscription and included-credit details.