Midjourney Is Slower and More Expensive Than It Has to Be
Most Midjourney V7 users still run every prompt at full quality, paying full GPU time even for the rough exploration stage where 90 percent of attempts get thrown away. They also start a brand new character every time, which is why a hero who appears in five marketing visuals ends up looking like five different people across the campaign.
Both problems have a fix that has been sitting inside Midjourney V7 the whole time. The fix is two features used together: Draft Mode for fast cheap exploration, and Omni Reference for character consistency once you have decided on a direction. Used in sequence, these two features cut your generation costs roughly in half and produce far more cohesive image sets.
This guide walks through what each feature does, how they combine into a single workflow, and where each one breaks down.
What Is Midjourney Draft Mode?
Draft Mode is a Midjourney V7 generation setting that produces images at roughly 10x the normal speed and roughly half the standard GPU cost. The trade-off is reduced fidelity. Drafts are clearly lower quality than standard outputs, but they are good enough to evaluate composition, style direction, and prompt logic before you commit to a final render. According to Midjourney's official documentation, Draft Mode is designed for iteration, not deliverables.
You activate Draft Mode in two ways. The first is to open the Create page in the Midjourney web app and click the Draft Mode toggle in the Imagine bar. Every prompt you type then runs in draft. The second is to append `--draft` to any individual prompt, which forces that single generation into draft mode even when the toggle is off.
The hidden third feature inside Draft Mode is Conversational Mode. Once a draft is generated, you can talk to Midjourney in plain language. Ask it to make the lighting warmer, swap a cat for an owl, or change the time of day. Midjourney rewrites the prompt and reruns the draft automatically.
What Is Omni Reference?
Omni Reference is Midjourney V7's built-in tool for keeping a character, object, or style consistent across multiple generations using a single reference image. You upload one image, and Midjourney maintains that subject's appearance across new poses, environments, and scenes. According to Midjourney's documentation, Omni Reference replaced the older `--cref` (Character Reference) parameter and is the only consistency tool that works in V7.
To use it from the Midjourney web app, start a new generation with /imagine, click the image attachment icon, choose Omni Reference, and upload your source image. Add your prompt describing the new scene. In Discord, the syntax is `--oref [image URL] --ow [weight]`.
The `--ow` parameter controls how strongly the reference image influences the new generation. Values from 25 to 100 give the AI more freedom. Values from 200 to 400 are the most-used range, balancing fidelity and variation. Values above 600 lock the character tightly but reduce flexibility for new poses and outfits.
Why Use Draft Mode and Omni Reference Together?
Draft Mode and Omni Reference solve different problems at different stages of a project. Draft Mode handles the exploration phase, when you do not yet know what the image should look like. Omni Reference handles the consistency phase, when you have decided on a character and need to reproduce them across a series. Used in sequence, the two features cut roughly half the cost out of a typical campaign workflow.
The order matters. If you run Omni Reference at full quality during exploration, you burn GPU credits chasing prompts that you would have rejected anyway. If you run Draft Mode after you have established your hero character, you lose the consistency Omni Reference was meant to enforce.
The clean rule: draft first to find the look, then switch to Omni Reference at standard quality to roll out the series.
What Does the Combined Workflow Look Like?
The combined Draft-plus-Omni workflow runs in three phases. Phase 1 uses Draft Mode to explore 10 to 20 cheap variations of the subject, lighting, and composition. Phase 2 picks the strongest draft and runs it at standard quality to lock in the hero image. Phase 3 uses that hero image as an Omni Reference source for every subsequent scene in the campaign.
A real example: imagine you are building a 6-image marketing campaign featuring a fictional Hong Kong cafe owner.
---Phase 1 (Draft): generate 15 to 20 drafts exploring different cafe interiors, lighting moods, and character styles. Total cost: roughly the price of 8 standard generations.
---Phase 2 (Lock): pick the strongest draft, click Enhance, and produce a standard-quality version. This becomes your reference image.
---Phase 3 (Roll out): for each remaining scene (cafe interior, street shot, working at laptop, customer interaction, sunset close-up, group photo), use the locked image as Omni Reference with `--ow 200` and write a fresh prompt describing the new scene.
Try This Prompt: Draft, Enhance, Then Roll Out a Character Series
Below is a complete copy-paste prompt set you can run through the Draft-plus-Omni workflow today. The scenario is a young professional protagonist for a tech newsletter cover series.
Phase 1: Run 4 to 8 of these as drafts to find your direction. Toggle Draft Mode ON before pasting.
portrait of a young Hong Kong creative professional in their late twenties, working at a laptop in a modern coffee shop, warm afternoon light, candid editorial photography, shallow depth of field, 16:9
portrait of a young Hong Kong creative professional in their late twenties, walking through Sheung Wan at golden hour, holding a coffee cup, editorial fashion photography, 16:9
portrait of a young Hong Kong creative professional in their late twenties, in conversation at a co-working space, soft window light, documentary editorial, 16:9
Phase 2: Pick your favourite draft. Click Enhance to upgrade to standard quality. Download that final image.
Phase 3: Open a new prompt, attach the Phase 2 image as Omni Reference with weight 200, and generate the remaining scenes:
same person, presenting at a small whiteboard meeting, conference room with city view, editorial photography, --oref [your image URL] --ow 200 --v 7
same person, reading on the MTR train heading home in the evening, soft natural light through windows, candid documentary style, --oref [your image URL] --ow 200 --v 7
Where Does This Workflow Break Down?
The Draft-plus-Omni workflow has three failure modes worth knowing before you commit a campaign to it. Drafts can mislead about how a prompt scales to full quality, Omni Reference costs 2x the GPU time per image compared to a regular V7 generation, and the system produces "recognisably the same person" rather than pixel-identical results.
The honest limitations to plan for:
---Drafts sometimes resolve detail poorly. A composition that looks great in draft can fall apart at standard quality if the prompt is ambiguous about facial features or hand positions. Always enhance the draft to standard quality before committing the look.
---Omni Reference at low weight (--ow 25 to 100) gives Midjourney more creative freedom and can shift facial features. If consistency is the priority, stay in the 150 to 250 range.
---Omni Reference does not produce pixel-perfect matches. If you need an exact, repeatable character (for example, a logo character or a known person's likeness for compliance reasons), Midjourney is not the right tool. Use it for "same character, different scene" not "same character, identical face."
How Much Does This Actually Save?
Running a 6-image campaign with the Draft-plus-Omni workflow typically costs around 60 percent of what the same campaign would cost run entirely at standard quality. The savings come from two places: cheap drafts during exploration (10 to 20 attempts at half cost each) and avoiding rework caused by inconsistent character appearance across the series.
The harder-to-measure saving is time. A typical 6-image consistent series done entirely at standard quality without Omni Reference requires roughly 30 to 50 prompt attempts as you try to match the character across scenes manually. With Omni Reference, the same series usually lands in 8 to 12 attempts.
The output is also more cohesive. Visual cohesion is the part marketers usually trade off for budget. The Draft-plus-Omni workflow lets you keep both.
Make This Your Default Workflow
The next time you open Midjourney V7 for any project involving more than two related images, default to this workflow rather than guessing your way through standard generations. The result is faster, cheaper, and visually more coherent. We know AI's cold edges. We know your real challenges. 28 years with UD, turning technology into a partnership with warmth.
The fastest way to internalise the workflow is to use it on your next real project. Pick something small, like a 4-image LinkedIn post series, and run the full Draft-Enhance-Omni loop end to end. By the third project, you will not remember how you used to work.
Take Your AI Visual Skills to the Next Level
Now that you have the workflow, the next step is building it into your team's daily creative process. We'll walk you through every step, from setting up the Draft-Enhance-Omni pipeline to managing reference libraries and scaling consistent character visuals across multiple campaigns.
