Midjourney v7 vs FLUX 2 Prompts: Dialect Comparison

Midjourney v7 prompts and FLUX 2 prompts describe the same scene in different dialects. Midjourney wants a subject-first line plus trailing parameters: --ar, --style raw, --stylize, --sref, --v 7. FLUX 2 wants photographic sentences, materials, and lens cues in natural language, with aspect and size set in the UI or API. Paste one dialect into the other box and you waste tokens or get the wrong look. This guide compares the two dialects as of mid-2026, shows the same brief rewritten for each model, lists cross-paste mistakes, and gives a soft path through PromptMake /image when you start from a reference photo.

Who this dialect comparison is for

You already generate images on both stacks, or you plan to. You need a clear map of how midjourney v7 prompts differ from FLUX 2 prose so you stop rewriting the same brief twice from scratch. Art directors who approve mood in Midjourney and ship catalog stills on FLUX need both dialects. Product teams that live in FLUX APIs and only open Midjourney for concept art need the reverse translation.

This page stays on two models. Sister posts rank prompting difficulty across three generators or map weight syntax. Stay here when your job is dialect: what words go in the text box, what controls live outside the text, and how to convert one working string into the other without losing the brief.

Skip this guide if you only ever run one host and never cross-paste. Lean in if your week mixes Discord or Midjourney web with a FLUX.2 Pro, Max, or Dev endpoint.

What "dialect" means for image prompts

Dialect is the shape the model expects. Midjourney v7 treats early nouns as heroes and reads trailing double-dash flags as controls. FLUX 2 treats the full sentence as instruction and ignores Midjourney flags. Both models improved natural-language reading through 2025 and mid-2026. Both still punish cross-paste.

A dialect mismatch looks like this. You paste --style raw --stylize 80 --ar 4:5 into a FLUX field. FLUX may ignore the flags or treat fragments as literal text. You paste a long FLUX paragraph into Midjourney without --style raw and get a painterly plate when the brief asked for catalog photo. The scene nouns were fine. The dialect was wrong.

Think of Midjourney as art direction plus a control strip. Think of FLUX as a shot list written in complete sentences. You write both. You keep two templates for the same product line.

Midjourney v7 prompts: the parameter dialect

Midjourney v7 prompts work best when you lead with the subject, add environment and light in short clauses, name the medium, then append parameters. The house aesthetic still pushes polish unless you dial it down. That bias helps concept art and editorial mood. It fights literal product geometry until you add --style raw and keep stylize low.

v7 reads sentences better than old tag soup. You still win when the hero noun sits early and one register owns the prompt. Mix "whimsical storybook" with catalog photography in one string and Midjourney splits the difference in ways clients reject. Pick illustrative or photographic before you touch parameters.

Reference modes do work Midjourney prose cannot. --sref locks palette and grade. --cref or --oref locks identity or object. --sw, --cw, and --ow set how hard those references push. Learn the control strip once; reuse it across a campaign while scene text changes.

Core midjourney v7 prompts structure

Default shape: [Subject + action] + [place] + [light] + [camera or medium] + [parameters]. Put the person, product, or hero object first. Keep one medium word: editorial photography, watercolor concept art, product catalog shot. End with flags.

Photoreal baseline as of mid-2026: --style raw --stylize 50–150 --chaos 5–15 --v 7 plus an --ar that matches the deliverable. Exploratory art can drop raw, raise stylize, and bump chaos for grid variety. Draft Mode (where your plan exposes it) is for composition tests; standard render for delivery.

Example editorial portrait: "Woman in her 40s by a rain-streaked window, soft overcast daylight, muted wool coat, 85mm shallow depth, editorial fashion photography --ar 4:5 --style raw --stylize 100 --chaos 8 --v 7". The nouns carry the brief. The flags keep Midjourney from painting over the camera language.

Midjourney controls that have no FLUX twin

--ar sets frame shape inside Midjourney. FLUX sets width and height outside the prompt. --stylize and --chaos tune house taste and grid spread; FLUX has no equal pair of flags in the text box. --sref / --cref / --oref are Midjourney-native reference paths. FLUX hosts that support image edit or multi-reference do similar jobs through UI slots, not trailing --sref syntax.

--no excludes concepts on Midjourney. FLUX stacks as of mid-2026 prefer positive scene language: "empty white seamless" instead of a negative list. When you translate MJ to FLUX, rewrite exclusions into positive nouns.

Keep --v 7 explicit when you compare old prompts or share strings with teammates. Version drift causes false "the model got worse" reports when someone reran a v6 string.

FLUX 2 prompts: the prose dialect

FLUX 2 (Black Forest Labs family builds such as Pro, Max, Flex, and Dev where your host exposes them) rewards literal, grammatical scene writing. Vague mood words produce accurate-but-generic stills. Specific materials, light direction, lens feel, and spatial placement move pixels. The model follows. It does not invent Midjourney-style cinematic polish unless you ask.

As of mid-2026, public guidance around Flux 2 favors declarative prose, camera and lens language for photoreal work, hex color on objects when brand match matters, and quoted strings for in-image text. Aspect and resolution live in the host or API. Leave Midjourney parameters out of the text field.

FLUX.1.x builds remain common on many UIs. Name the build in your prompt notes so teammates compare like with like. PromptMake /image writes FLUX-shaped natural language you can paste into either generation of hosts; confirm which slot your account bills against before you blame the prompt.

FLUX sentence stack that works

Lead with the subject noun. Add action and place. Layer light quality and direction. Name materials that catch speculars. Close with camera or medium cues. Aim for dense photographic English in the 30 to 80 word band for single-hero stills. Longer prose helps multi-object sets when every object needs a role.

Example catalog mug: "Product photograph of a matte white ceramic coffee mug on a light oak desk, soft diffused daylight from camera left, shallow depth of field, neutral background, commercial catalog style, 85mm lens look." No flags. Aspect set in the UI.

Brand color example: "Matte water bottle centered on white seamless, soft rectangular softbox from above, label ink is #C41E3A, sharp focus throughout, empty scene." Hex binds to the object. Positive empty-scene language replaces --no clutter.

FLUX strengths midjourney v7 prompts seldom cover

In-image typography and layout briefs often land cleaner on FLUX 2 when you quote the exact string and place it in space. Spatial instructions ("subject on the right third, empty left third for headline") stick with fewer improvisations. API and open-weight workflows let you template the same prose across thousands of variants; Midjourney remains stronger for one-shot art direction and --sref campaign look.

FLUX will not rescue a thin prompt with house beauty. "Nice coffee shop" returns a generic interior. Write window direction, fixture materials, and one customer action if those details matter. That literal burden is the price of obedience.

Same brief, two dialects

Use one brief. Write both dialects. Keep a two-column note for every recurring job. The subsections below share nouns and light so you can see where the dialects diverge.

Rule for conversion: Midjourney to FLUX means strip flags, expand camera into prose, rewrite --no into positive scene words, move aspect to UI. FLUX to Midjourney means compress into subject-first clauses, add --ar and photo stack (--style raw, low stylize, --v 7), and attach --sref or --cref when identity or grade must lock across a set.

Trail runner ad

Brief: female trail runner on alpine ridge at golden hour for a sportswear campaign.

Midjourney v7: "Female trail runner on alpine ridge, golden hour rim light, dynamic low angle, sportswear ad photography --ar 3:2 --style raw --stylize 120 --chaos 10 --v 7". Optional: --sref from an approved grade plate with --sw 80.

FLUX 2: "Photoreal sportswear ad of a female trail runner on a rocky alpine ridge at golden hour, warm rim light from camera right, dynamic low angle, red jacket and black shorts, sharp focus on runner, soft distant peaks, 35mm look, empty sky for logo." Set 3:2 in the host. Quote any on-jacket text if the brief needs it.

Ceramic product plate

Brief: matte ceramic mug on oak, soft window light, ecommerce square.

Midjourney v7: "Ceramic mug matte white glaze on warm oak surface, soft 45-degree window light, subtle shadow, catalog photography --ar 1:1 --style raw --stylize 100 --chaos 5 --v 7".

FLUX 2: "Catalog product photograph of a matte white ceramic mug on warm oak, soft 45-degree window light from camera left, subtle contact shadow, clean empty background, sharp focus throughout, square commercial still." Square size in UI fields.

Painterly concept art

Brief: floating islands, rope bridges, sunset god rays, fantasy concept.

Midjourney v7: "Floating islands connected by rope bridges, sunset god rays, painterly fantasy concept art, atmospheric perspective --ar 16:9 --stylize 400 --chaos 25 --v 7". Skip --style raw so the house aesthetic helps.

FLUX 2: "Wide fantasy concept painting of floating stone islands linked by rope bridges at sunset, volumetric god rays through haze, painterly brush texture, atmospheric perspective, rich orange and violet sky, no photoreal camera look." Name medium in prose because FLUX will not invent Midjourney polish from a short mood phrase alone.

Cross-paste mistakes that burn credits

Mistake 1: Midjourney parameter strings pasted into FLUX. Strip --ar, --stylize, --sref, and friends before paste, or regenerate with FLUX selected in an image-to-prompt tool.

Mistake 2: FLUX essays pasted into Midjourney without a photo stack. Add --style raw and low stylize when the brief is photographic. Compress meandering clauses so the subject leads.

Mistake 3: Treating dialect as quality ranking. Midjourney can look "better" on a vague line because it art-directs. FLUX can look "worse" on the same line because it obeyed a thin brief. Fix the dialect and the specificity, then compare.

Mistake 4: Expecting --no lists to behave on FLUX. Rewrite into positive framing.

Mistake 5: Asking midjourney v7 prompts to carry hex brand codes the way Flux 2 does. Use --sref for grade lock or post-process color; do not assume hex in MJ text binds like FLUX object color.

Mistake 6: Mixing three mediums in one string on either model. One register per generation.

Mistake 7: Rewriting the entire prompt when one layer failed. Change light, material, or one parameter. Keep the working half.

Step-by-step: pick a dialect and convert

Run this loop when a new brief lands. Decide the deliverable first: mood board and art direction, or literal catalog and API batch. That choice picks the primary dialect. Convert when a second host must match the approved look.

Keep a shared sheet with columns for brief ID, Midjourney string, FLUX string, aspect, date, and build name. Teams that skip the sheet relearn the same conversion every Monday. Store the sheet next to your brand --sref URLs and FLUX hex notes so art and ops open the same source.

Budget twenty minutes the first time you convert a campaign hero. Later briefs take five minutes when the sheet already holds a twin pair for that product line. The subsections walk host choice, native draft, and conversion without rewriting the brief from memory.

1. Name the job and the host

Write one sentence: who or what, where, light, medium, aspect. Circle primary host. Midjourney for stylized exploration, --sref campaigns, and editorial mood. FLUX 2 for photoreal obedience, typography in frame, and API pipelines. If you need both, pick the host that ships the file clients approve, then translate for the secondary stack.

2. Write native dialect first

Do not translate from a foreign dialect on pass one. Write Midjourney with parameters if MJ is primary. Write FLUX sentences if FLUX is primary. Generate once. Score subject, light, and medium. Fix one layer.

3. Convert only after the primary passes

When you need the other model, run the conversion rules above. Or upload the approved still to PromptMake /image, select Midjourney or FLUX as the write-out target, and edit the draft. Soft path: https://promptmake.net/image. Guests get about 3 image generations per day; free registration raises the image quota. Quotas for /image and /text stay separate.

Soft next step on PromptMake /image

PromptMake /image fits when you hold a reference photo and need midjourney v7 prompts or FLUX prose without hand-building both dialects. Upload a still, pick Recreate Exactly, Change Style, Adjust Lighting, or Create Variation, choose Midjourney or FLUX, copy the draft, then add your parameter strip or UI aspect settings.

Use Recreate when composition must hold. Use Adjust Lighting when only key and fill change. Use Change Style when medium swaps. Use Create Variation when the photo is a mood seed. Soft start: https://promptmake.net/image

Blank-page generators help when you have no still. Weighting syntax for Midjourney and FLUX lives in a separate guide. Three-way difficulty ranking across DALL·E-class, Midjourney, and FLUX lives elsewhere. This article owns the two-dialect map for v7 and Flux 2.

FAQ

What are midjourney v7 prompts?

Midjourney v7 prompts are subject-first scene lines plus trailing parameters such as --ar, --style raw, --stylize, --sref, and --v 7. v7 reads natural language better than older tag lists, yet the control strip still decides photoreal versus house aesthetic. Write photographic jobs with raw and low stylize. Raise stylize and skip raw for painterly concept work.

How do Midjourney v7 prompts differ from FLUX 2 prompts?

Midjourney packs controls into trailing flags and leans on art direction defaults. FLUX 2 packs detail into full sentences and expects you to name light, materials, and lens in prose. Aspect lives in Midjourney via --ar and in FLUX via UI or API fields. Cross-paste without conversion wastes credits on both stacks.

Should I use Midjourney v7 or FLUX 2 for product photos?

FLUX 2 often wins when the brief needs literal geometry, empty seamless, and API batching. Midjourney v7 can ship strong catalog work with --style raw, low stylize, and tight chaos, then --oref or --sref for consistency. Pick by pipeline: Discord or web art direction versus FLUX API production. Many teams concept on Midjourney and finalize on FLUX.

Can I paste the same prompt into both models?

You can paste, but you should convert first. Strip Midjourney flags before FLUX, and add a photo parameter stack before Midjourney photoreal runs. Rewrite negatives into positive scene language for FLUX. Keep two saved strings per brief instead of one shared mush.

Does FLUX 2 support Midjourney parameters like --sref?

No. FLUX ignores Midjourney trailing parameters. Style and identity lock on FLUX hosts use image edit, multi-reference, or LoRA workflows where your stack exposes them. Set aspect outside the prompt and keep FLUX text as photographic sentences.

How do I turn a photo into midjourney v7 prompts or FLUX text?

Upload the photo to an image-to-prompt tool that lets you pick the write-out model. PromptMake /image formats Midjourney and FLUX as targets: https://promptmake.net/image. Edit invented props, fix light calls, then add Midjourney flags or FLUX UI size settings. Guests get a small daily image quota; registration raises it.

Which dialect should beginners learn first in 2026?

Learn the dialect of the host you can open today. If that host is Midjourney, master subject-first lines plus --style raw and --ar. If that host is FLUX 2, master dense photographic sentences and positive framing. Add the second dialect when a brief forces a second host; conversion is faster once one native string already passes.