FLUX Kontext vs Midjourney V7: legacy and current choices

If you specifically need to reproduce an old Kontext or V7 result, keep the legacy version pinned and preserve its prompt, seed, references and settings. Do not assume the current defaults will recreate it.

FLUX.2 vs Midjourney V8.2 at a glance

This is a documented workflow comparison checked September 13, 2026, not a controlled image-quality benchmark. “Better” depends on whether the deliverable is an API-integrated edit, a reusable production system or a hands-on creative exploration.

The direct decision

• Choose FLUX.2 when software integration, repeatable inputs, exact model selection, multiple references or pay-as-you-go API usage drive the decision.
• Choose Midjourney V8.2 when a creator wants to explore visual directions, use the web Editor, combine style controls and iterate within Midjourney's gallery workflow.
• Keep FLUX.1 Kontext only for a legacy integration or when a provider has not exposed FLUX.2. BFL calls Kontext previous-generation.
• Keep Midjourney V7 only for V7-specific behavior such as Omni Reference or an old prompt you need to reproduce. V8.2 replaces Omni Reference with the Edit Model.

What changed since the original comparison

FLUX.2 replaced Kontext as BFL's recommendation

FLUX.2 combines generation and image editing across [klein], [pro], [max], [flex] and [dev]. BFL documents up to eight reference images through the API and ten in the playground for pro, max and flex, while klein supports four. Output can reach 4 megapixels. Kontext remains available, but BFL labels it previous-generation for new projects.

Midjourney V8.2 replaced V7 as the default

Midjourney V8.2 became the default on July 24, 2026. Its Edit Model replaces V7's Omni Reference, Character Reference and legacy Retexture path. The current Edit Model accepts written instructions and up to four reference images; the web Editor also supports masks, inpainting, outpainting, Smart Select and layers.

Editing an existing image

For FLUX.2, assign each input a role: for example, “Use the person from image 1, the jacket from image 2 and the office from image 3. Preserve the person's face and the jacket logo.” This maps naturally to an API pipeline where inputs and model versions are recorded.

For Midjourney V8.2, use the Edit Model for instruction-based editing and the Editor when you need to erase a region, extend the canvas, move a layer or regenerate a masked area. This is a real editing workflow; the old article's claim that Midjourney only reinterprets references is no longer accurate.

Neither workflow guarantees exact identity, logos, product geometry or readable type. Compare the output with the source and finish exact brand assets or legal copy in a conventional editor.

Generating and controlling style

Midjourney's creator workflow combines text prompts with Style References, Moodboards and Personalization. FLUX.2 uses text or structured prompts, multiple references, exact color instructions and variant-specific controls. The distinction is less “editing versus generation” than interface and deployment: both current families can generate and edit.

For a fair comparison, use the same brief and permitted references, then score prompt adherence, subject preservation, text accuracy, edit locality, output dimensions, retry count and total cost per accepted image. Keep rejected attempts; a single favorite image hides variance and workflow cost.

Current pricing and access

BFL's rates are model- and megapixel-dependent. Reference images can add cost, and resolution is rounded according to the pricing rules. Midjourney subscriptions renew automatically and allocate GPU time; Standard, Pro and Mega include Relax image generation, while Stealth is limited to Pro and Mega. Verify checkout and the exact job before budgeting a production batch.

Privacy and commercial-use checks

• Confirm whether prompts, references and outputs are stored or publicly visible in the selected interface.
• Midjourney says V8.x Editor creations can appear on its site unless Stealth is enabled; external-image handling has additional rules documented in the Editor guidance.
• For FLUX open weights, read the exact variant's license. The 4B klein and other variants do not all use the same license.
• Record the model, provider, plan, generation date and terms for client or commercial work.
• Use only reference images you have permission to upload and transform.

Where Magic Hour fits

Use Magic Hour's AI Image Editor when you want browser-based generation and editing without wiring a direct BFL API or managing Midjourney's version controls. Check the selected model shown in the interface and judge the downloaded result. Platform model availability, credit pricing and defaults can differ from BFL's direct service.

Official sources checked

FLUX status, editing and pricing were checked using BFL's FLUX.2 overview, image-editing documentation and pricing page. Midjourney status, editing and plans were checked using its Version page, Edit Model documentation, Editor documentation and plan comparison. Recheck current model defaults and checkout before a consequential purchase.