Move your team from Flux to Midjourney.
Each person gets a short interview, then a personalized step-by-step guide showing exactly where their saved work, prompts, and projects land in Midjourney — and the few things that need a manual step, walked through. You send one invite; nobody becomes the help desk, and nothing in Flux is changed or cancelled. Want to see where you’re paying for both first? Run the free audit.
Both tools can extend an image beyond its original borders and erase unwanted objects. Moving from Flux to Midjourney for outpainting: open the image in the midjourney.com Editor and use Zoom Out to expand the canvas, or use Pan to add content directionally (left, right, up, down). For object removal, paint the area with the Erase brush and let the inpaint step reconstruct the background. Moving from Midjourney to Flux: call the Flux Tools outpainting endpoint via the BFL REST API, passing the source image; no text prompt is required as the model infers continuation from existing content. For object removal, use the coherent reconstruction endpoint, passing a mask that marks the element to erase.
- Warning: Midjourney's outpaint (Zoom Out) and Pan tools are web-only and cannot be driven programmatically; Flux outpainting is API-only, making it the better choice for batch pipelines.
- Warning: Flux outpainting and object removal sit in the Flux Tools suite and are billed separately from text-to-image generation; check BFL pricing for the specific endpoint.
- Warning: Midjourney Zoom Out produces a new full-image composite and can diverge stylistically from the original at large zoom increments; Flux outpainting infers continuation without needing a guiding prompt, but may still drift at very large extensions.
Both platforms allow users to supply one or more reference images so the generated output preserves a specific character's appearance across prompts. Moving from Flux to Midjourney: take the reference image you were passing to the Flux API and supply it via --cref <image-URL> on V6/Niji-6, or --oref <image-URL> on V7/V8.1. Use --cw (0-100) to dial how strictly Midjourney copies face and clothing. Combine with a normal text prompt to change scene or pose while keeping the character recognizable. Moving from Midjourney to Flux: pass the same reference image(s) as base64 or URLs in the FLUX.2 API request alongside the text prompt; FLUX.2 accepts up to 10 reference images simultaneously, enabling multi-character or character-plus-style-reference in a single call.
- Warning: Midjourney --cref is documented for V6/Niji-6; in V7 and V8.1 it is replaced by --oref (Omni Reference), and --oref is not supported in V8.1 either — check the version compatibility chart before using these parameters.
- Warning: Flux's multi-reference approach is API-only; there is no GUI that exposes uploading multiple reference images interactively the way Midjourney's web Editor does for a single reference.
- Warning: Character consistency quality varies by reference image clarity on both platforms; cluttered backgrounds or group shots reduce accuracy in both Flux and Midjourney.
Both platforms let users mask a region of an existing image and regenerate that region with new content. Moving from Flux to Midjourney: upload the source image to midjourney.com, open the Editor, use the Erase brush to paint the region you want replaced, optionally write a Remix prompt describing the new content, then generate — Midjourney's Vary Region rerenders only that masked area. Midjourney recommends masking 20-50% of the image for best coherence. Moving from Midjourney to Flux: use the FLUX Tools API endpoint for in-context editing, passing the source image and a text instruction describing what to change; FLUX.2 and FLUX.1 Kontext both support this. Mask data can be supplied as a base64-encoded PNG or image URL in the request body.
- Warning: Midjourney's inpaint editor (Vary Region) is available only on the web app, not Discord; Flux inpainting is API-only with no native GUI beyond the BFL playground.
- Warning: Flux FLUX.2 accepts up to 10 reference images in one request for multi-source edits; Midjourney's inpaint operates on a single base image at a time.
- Warning: Midjourney inpaint consumes GPU hours at standard rates; Flux charges per image based on model tier and output dimensions, so high-volume inpainting cost will differ.
Both tools accept a plain text prompt and return high-quality generated images, making workflow migration mostly a matter of interface and pricing model. Moving from Flux to Midjourney: switch from API calls or the BFL playground to midjourney.com (web) or the /imagine Discord command; reformulate prompts — Midjourney responds well to descriptive noun phrases and style adjectives appended with parameters like --ar 16:9 and --stylize 400, whereas Flux handles more verbose, sentence-form prompts. Moving from Midjourney to Flux: sign up for a BFL API key, select the appropriate model tier (FLUX.2 for balanced quality/speed, FLUX.2 Max for top quality, FLUX.2 Klein for sub-second), and call the REST endpoint or use the BFL playground. Prompt phrasing carries over with minimal rewriting. Output dimensions are configured as width/height API parameters in Flux vs. --ar in Midjourney.
- Warning: Midjourney requires an active paid subscription (no permanent free tier as of 2026); Flux charges pay-as-you-go per image with no subscription, so cost structures differ significantly at low and high volumes.
- Warning: Midjourney V8.1 defaults to 1024x1024px square; Flux 1.1 Pro Ultra can generate up to 4MP with flexible aspect ratios, offering higher resolution on demand.
- Warning: Midjourney always returns four image variants per job; Flux returns a single image per API call, so batch generation on Midjourney may need multiple API calls to replicate.
Both platforms offer ways to anchor outputs to a specific aesthetic, but the mechanism differs significantly. Flux fine-tuning involves training on custom datasets using the downloaded Open Weights model, running training in your own infrastructure, and deploying the resulting weights. This is the right path for teams that need a locked brand style reproducible programmatically at scale. Migrating that workflow to Midjourney: use Style Reference (--sref <URL or code>) to apply the color palette, texture, and lighting of a reference image to new prompts, or build a Moodboard for project-level style anchoring. For personal taste persistence, enable Personalization (--p) after rating ~200 images. Moving from Midjourney style controls to Flux fine-tuning: collect the reference images that best capture your aesthetic, assemble a training dataset, acquire an Open Weights license (Builder tier and above include fine-tuning rights), train on downloaded weights, and deploy via the BFL API.
- Warning: Flux fine-tuning requires engineering capacity to run training infrastructure; Midjourney style controls are point-and-click with no training step, making the two approaches differ in setup complexity and reproducibility.
- Warning: Midjourney Personalization requires rating approximately 200 images to build a profile and is user-scoped, not shareable as a team brand asset; Flux fine-tuned weights are portable and sharable across deployments.
- Warning: Midjourney --sref is version-sensitive: --sv 7 (released ~April 2026) costs 4x standard rates; special modes like --hd and --sref together in V8 Alpha also cost 4x standard rates.
- Warning: Flux Open Weights fine-tuning is capped per license tier (e.g. Builder: 10K images/month on the resulting model); high-volume commercial deployment requires Platform, Professional, or Enterprise tiers.