Qu'est-ce que SociaLover ?

SociaLover est une plateforme IA tout-en-un qui permet de creer des visuels publicitaires, des videos UGC et des avatars IA en quelques secondes, sans competence graphique. Elle est optimisee pour TikTok, Instagram, YouTube et Facebook.

SociaLover est-il gratuit ?

Oui, SociaLover propose un plan gratuit pour commencer. Des plans Pro et Agence avec plus de tokens et de fonctionnalites sont disponibles.

Quels modeles IA sont utilises ?

SociaLover integre les meilleurs modeles IA : Flux, DALL-E 3, BFL pour les images, et Veo 3, Sora 2, Kling 2.1, Runway et Wan 2.5 pour les videos.

Pour qui est fait SociaLover ?

SociaLover est concu pour les e-commercants, les agences publicitaires, les social media managers, les createurs de contenu et les influenceurs qui souhaitent produire des publicites professionnelles rapidement et a grande echelle.

Does Flux 1.1 Pro support negative prompts?

No. Flux 1.1 Pro is not a Stable Diffusion checkpoint: there is no negative prompt, no sampling step count and no CFG or guidance scale. What you actually control is the prompt itself, the output geometry, the seed, and a handful of switches such as prompt_upsampling, output_format, raw and safety_tolerance.

Which FLUX variant should I pick for an ad visual?

FLUX 2 Schnell to explore a lot of angles quickly, FLUX.2 for the everyday feed creative in 1K or 2K, FLUX 2 Pro when the brief is detailed and every clause has to land, FLUX 2 Max or FLUX 1.1 Ultra for the finish. Ultra being the variant that exposes the raw switch. FLUX Kontext is a different job entirely: it edits an image you already have. All of them are billed in tokens included in your plan.

Can Flux 1.1 Pro edit an image I already have?

No, it generates from text only. To restyle an existing product photo or fix one area of a render, switch to FLUX Kontext: you give it the source image plus a plain-language instruction, with no mask to draw and no strength slider to tune.

How do I get a higher-resolution image out of Flux?

Not through the prompt. Flux 1.1 Pro has no upscale parameter, and asking for "8K" only acts as a style cue. It does not change the output resolution. Set an exact size with width and height in multiples of 32, then send the finished image through the Image Enhancer step if a close-up needs more definition.

How do I generate controlled variations of the same visual?

Keep the seed and the aspect_ratio fixed and change a single clause of the prompt: the surface, the light direction, the background colour. That isolates one variable, which is what makes an A/B test on creatives readable. For pure exploration, do the opposite: keep the prompt and increment the seed by 1 across twenty renders.

Using Flux 1.1 Pro for Your Ad Visuals: Complete Guide

Flux 1.1 Pro has become one of the go-to models for creating AI ad visuals. Discover how to use it effectively: prompts, settings, use cases and integration into your creative workflow.

SociaLover Team · March 12, 2026 · 4 min read

Flux 1.1 Pro by Black Forest Labs is one of the models marketers reach for most often for ad image generation. It is a text-to-image model exposed through a simple API, strong on skin and material texture, and fast enough to iterate on a creative in a single sitting. Its Ultra variant is available inside SociaLover, billed in tokens included in your plan like every other model in the catalogue.

This guide gives you the real API parameters, prompt formulas, and workflows specific to creating ad visuals with Flux 1.1 Pro. A warning up front: most "Flux tutorials" circulating online recycle Stable Diffusion settings (steps, CFG/guidance scale, negative prompts) that this model simply does not expose.

The FLUX line-up available inside SociaLover. Picking the variant is the first decision, and it is a decision about what the model does (text to image or image editing, draft or finish) not about a price list: every variant is billed in tokens included in your plan.
Variant	Type	What it is for
FLUX 2 Schnell	Text to image	Drafts and volume exploration, fast enough to render twenty angles of the same idea
FLUX.2	Text to image	The everyday workhorse for ad visuals at feed resolution, in 1K and 2K
FLUX 2 Pro	Text to image	Detailed briefs where every clause of the prompt has to land
FLUX Kontext	Image editing	Editing an image you already have, from a plain-language instruction
FLUX 2 Max	Text to image	The high tier of the FLUX 2 family
FLUX 1.1 Ultra	Text to image	The Ultra variant discussed here, the one that exposes the raw switch

Explore · FLUX 2 Schnell

Drafts and volume exploration: twenty angles of the same idea before you commit to one

Produce · FLUX.2 · FLUX 2 Pro

Feed resolution in 1K and 2K; Pro for briefs where every clause of the prompt has to land

Finish · FLUX 2 Max · FLUX 1.1 Ultra

The high tier of the family. Ultra is the variant this guide is about · the one that exposes the raw switch

Edit · FLUX Kontext

The odd one out: it starts from an image you already have, plus a plain-language instruction · no mask, no strength slider

The same six variants as the table above, regrouped by the job you would hire each one for. It is the practical way to read the line-up: five of them generate from text and differ by how much of a detailed brief they hold, while FLUX Kontext does something else entirely · it edits rather than generates.

Why Flux 1.1 Pro is different from other models

The ratings below are the SociaLover team's editorial assessment on advertising use cases: not an independent benchmark, and not a reproducible measurement. Treat them as a starting point and run your own brief through each model.

Criterion	Flux 1.1 Pro	Midjourney V8	GPT Image 2
Face realism	★★★★★	★★★★	★★★
Product texture rendering	★★★★★	★★★★	★★★
Instruction following (prompt)	★★★★★	★★★★	★★★★★
Text rendering in the image	★★★★	★★★	★★★★
Stylistic diversity	★★★★	★★★★★	★★★
API access / integration	★★★★★	★★★	★★★★★
Generation speed	★★★★★	★★★	★★★★

One brief, four vendors

FLUX.2 · Nano Banana 2 · GPT Image 2 · Seedream 4.5

Black Forest Labs, Google, OpenAI and ByteDance all sit behind image models available in Studio Image and Creative Designer. No single one of them is the platform's engine: the table above rates FLUX 1.1 Pro highest on faces and textures and GPT Image 2 level with it on instruction following, which is exactly why the same brief is worth sending to several before you pick a render.

The structure of the perfect Flux 1.1 Pro prompt

Flux 1.1 Pro follows instructions very precisely, but a poorly structured prompt produces mediocre results. Here's the optimal structure:

Recommended structure:

[MAIN SUBJECT] + precise descriptive details

[ENVIRONMENT / SETTING] + mood

[LIGHTING] + light source and quality

[PHOTOGRAPHIC STYLE] + visual references

[FINISH CUES] + quality keywords (style signals, not real settings)

Full example:

"A luxury skincare serum bottle [SUBJECT], placed on white marble surface with water droplets [SETTING], soft diffused light from upper left casting gentle shadows [LIGHTING], high-end commercial product photography style, shot with 85mm f/2.8 macro lens [STYLE], ultra-detailed, advertising campaign finish [FINISH]"

Flux 1.1 Pro settings: the parameters the API actually exposes

Flux 1.1 Pro is not a Stable Diffusion checkpoint. There is no sampling step count, no CFG / guidance scale, and no negative prompt: what you control is the prompt itself, the output geometry, the seed, and a handful of switches. Here is what each one does, and how to set it for advertising.

Parameters the API actually exposes

— aspect_ratio: 1:1 or 4:5 for Instagram and Meta feeds, 9:16 for Stories and Reels
— width / height: an exact pixel size, in multiples of 32; pick either this or aspect_ratio, never both
— seed: fixed to get variations of the same subject, empty for a genuinely new image
— prompt_upsampling, the API rewrites and enriches your prompt before generating
— output_format: PNG to composite or overlay text afterwards, JPEG to upload straight to the ad platform
— raw: on the Ultra variant, a less processed, more photographic rendering
— safety_tolerance: moderation strictness; the lower the value, the stricter the filter

Stable Diffusion settings that do not exist here

— Sampling step count · there is none to set
— CFG / guidance scale, not exposed by this model
— Negative prompts, you cannot list what you do not want
— The (element:1.5) weighting syntax, it has no effect at all
— An upscale parameter, there is none; enhance the image afterwards instead
— Asking for "8K" in the prompt to raise the output resolution, it only acts as a style cue
— A mask to draw or a strength slider: that belongs to FLUX Kontext, which needs neither

For e-commerce product visuals:

aspect_ratio

1:1 or 4:5

Square or portrait format for Instagram and Meta feeds. Avoid 16:9 for feed placements.

prompt_upsampling

false

When enabled, the API rewrites and enriches your prompt before generating. Useful on a two-line brief, counterproductive on a product shot where every word of your prompt is deliberate.

seed

Fixed for variations

Reuse the same seed with a slightly edited prompt to get variations of the same subject with micro-differences in background or lighting. Leave it empty for a genuinely new image.

output_format

jpeg / png

PNG for anything you will composite or overlay text on afterwards, JPEG for a file you upload straight to the ad platform.

For faces and avatars:

aspect_ratio

9:16 or 1:1

Vertical portrait for Stories/Reels. Square for the feed and thumbnails.

width / height

Multiples of 32

Use them instead of aspect_ratio when you need an exact pixel size, the model expects multiples of 32. Passing both a ratio and a size is redundant: pick one.

raw

true

Available on the Ultra variant. It produces a less processed, more photographic rendering. Closer to an unretouched camera file, which reads as far more authentic on a face.

safety_tolerance

Low for ad accounts

Moderation strictness on the prompt and the output. The lower the value, the stricter the filter. A sensible default when the creative has to stay inside Meta and TikTok ad policies.

Note on upscaling: Flux 1.1 Pro has no upscale parameter. If a close-up face needs more definition, generate it, then send the result through the Image Enhancer step. Asking the model for "8K" in the prompt does not change the output resolution.

Flux 1.1 Pro prompt library by ad category

Beauty / Cosmetics

Beginner

Luxury beauty product photography, glass skincare bottle with gold cap, soft pink marble background, morning light streaming from window, water droplets on bottle, premium commercial photography, pastel color palette, editorial Vogue beauty style

Expected result: Clean background, premium feel, ideal for Instagram beauty ads

Fashion / Apparel

Intermediate

Fashion editorial photography, clothing item displayed as flatlay on light grey linen surface, natural daylight from above, subtle texture visible on fabric, minimalist composition, clean white negative space, high-end fashion brand catalog style, Canon EOS R5 shot

Expected result: Realistic fabric rendering, magazine-style composition, suited to Meta fashion ads

Food / Beverage

Intermediate

Professional food photography, [product] centered on dark slate surface, steam rising gently, fresh ingredients scattered naturally around, dramatic moody lighting from side, shallow depth of field, Michelin star restaurant aesthetic, appetizing warm color grade

Expected result: Gourmet atmosphere, realistic food textures, high CTR for food ads

Lifestyle / Fitness

Advanced

Lifestyle fitness photography, athletic person using product in modern home gym, golden hour light through large windows, motion blur on movement, energetic composition, clean athletic body, authentic sweat, Nike / Adidas campaign aesthetic, shot on 35mm film

Expected result: Realistic usage context, strong emotion, suited to Facebook performance ads

Tech / Electronics

Beginner

Clean tech product photography, device on pure white background, subtle gradient shadow below, sharp product edges, premium minimal design visible, professional studio lighting, Apple.com product page aesthetic, hyper-detailed screen reflection, 8K commercial quality

Expected result: Precise screen and material rendering, Apple/Samsung style, suited to Google and Meta ads

Woman applying a serum in a bright bathroom with daylight from a window

Sneakers on a gradient background with dramatic side lighting and a subtle shadow

Advanced techniques: editing and iterating with Flux

Where each tool takes over

Text

Structured prompt, 5 blocks

Generate

FLUX 1.1 Ultra · raw switch

Edit

FLUX Kontext · no mask

Resolution

Image Enhancer

Three different jobs, three different tools. Flux 1.1 Pro only generates from text, so a fix on an existing render moves to FLUX Kontext · source image plus a plain-language instruction, no mask and no strength slider. Definition is the third job: there is no upscale parameter on this model, and asking for “8K” in the prompt is only a style cue, so a close-up that needs more detail goes through the Image Enhancer step afterwards.

Editing an existing image: switch to FLUX Kontext

Flux 1.1 Pro generates from text. To restyle a product photo you already have, or to fix one area of a render, use FLUX Kontext instead: you give it the source image plus a plain-language instruction, with no mask to draw and no strength slider to tune.

Emphasis through wording, not weight syntax

The (element:1.5) weighting syntax inherited from Stable Diffusion has no effect here. To push an element forward, name it first in the sentence, describe it in more detail than the rest of the scene, and state its place in the composition explicitly, "centered", "filling the frame", "in sharp focus".

Batch generation with incremental seed

Generate 20 variations of the same prompt, incrementing the seed by 1 each time. You get 20 different interpretations of the same concept. Perfect for pinpointing the best angle.

Lock the frame, vary the scene

Keep aspect_ratio and seed fixed and change only one clause of the prompt (the surface, the light direction, the colour of the background). You isolate a single variable, which is what makes an A/B test on creatives actually readable.

Frequently asked questions

Does Flux 1.1 Pro support negative prompts?: No. Flux 1.1 Pro is not a Stable Diffusion checkpoint: there is no negative prompt, no sampling step count and no CFG or guidance scale. What you actually control is the prompt itself, the output geometry, the seed, and a handful of switches such as prompt_upsampling, output_format, raw and safety_tolerance.
Which FLUX variant should I pick for an ad visual?: FLUX 2 Schnell to explore a lot of angles quickly, FLUX.2 for the everyday feed creative in 1K or 2K, FLUX 2 Pro when the brief is detailed and every clause has to land, FLUX 2 Max or FLUX 1.1 Ultra for the finish. Ultra being the variant that exposes the raw switch. FLUX Kontext is a different job entirely: it edits an image you already have. All of them are billed in tokens included in your plan.
How do I push one element forward in a Flux prompt?: Not with weights. The (element:1.5) syntax inherited from Stable Diffusion has no effect on this model. Name the element first in the sentence, describe it in more detail than the rest of the scene, and state its place in the composition explicitly, "centered", "filling the frame", "in sharp focus".
Can Flux 1.1 Pro edit an image I already have?: No, it generates from text only. To restyle an existing product photo or fix one area of a render, switch to FLUX Kontext: you give it the source image plus a plain-language instruction, with no mask to draw and no strength slider to tune.
How do I get a higher-resolution image out of Flux?: Not through the prompt. Flux 1.1 Pro has no upscale parameter, and asking for "8K" only acts as a style cue. It does not change the output resolution. Set an exact size with width and height in multiples of 32, then send the finished image through the Image Enhancer step if a close-up needs more definition.
How do I generate controlled variations of the same visual?: Keep the seed and the aspect_ratio fixed and change a single clause of the prompt: the surface, the light direction, the background colour. That isolates one variable, which is what makes an A/B test on creatives readable. For pure exploration, do the opposite: keep the prompt and increment the seed by 1 across twenty renders.

Conclusion

Flux 1.1 Pro is a solid default for ad visuals: realism, instruction following, and texture rendering are among its strengths. Prompt mastery is the key skill here, precisely because the model gives you so few knobs to turn: the prompt, the geometry, and the seed are almost the whole surface.

In SociaLover, FLUX.2 is one of the image models available in Studio Image and Creative Designer, alongside Nano Banana 2, GPT Image 2, Seedream 4.5, Recraft V4.1 and others. No single model is the platform's engine: the point is to send the same brief to several of them and keep the one that renders your product best.