Midjourney Folder Defaults: A Session Checklist
Midjourney AI Team · October 8, 2026 · 6 min read
Midjourney folder defaults help separate working parameters between projects. The October 7 Alpha changelog, published October 8, says folder prompt pills now persist across sessions. This replaces the earlier tab-session limit, but a saved project brief is still useful.
This tutorial uses two example projects and a blank record you can fill from your own checks. The steps are editorial recommendations, not a claim that we tested every browser or device. The source was rechecked on October 13, 2026.
Give each folder a clear output purpose
Start with two projects whose needs are visibly different. One could hold square studies of a ceramic vase for a catalogue. The other could hold portrait botanical images for a printed invitation. The difference makes a wrong setting easier to notice than switching between two nearly identical jobs.
Name folders after the work they contain rather than temporary excitement. “Vase catalogue studies” and “Botanical invitation” tell a colleague what belongs inside. A name such as “new ideas” does not explain the required format or the stage of approval. Keep folder structure shallow enough to inspect quickly.
Before generating, write each project's intended subject, image shape, background treatment and reference roles. The record can be a short text note in your existing project system. Do not put private client details into a prompt unless you have permission and understand the service's handling of that material.
Keep a record separate from the active controls
For each project, record the selected model, intended ratio, active style references and any deliberate non-default setting. Also save the text prompt you actually used. An image alone does not show which part of the result came from text, references or interface settings.
Use a compact starting record:
Project: Vase catalogue studies
Purpose: square product exploration, quiet background
Prompt: a matte ivory ceramic vase on a pale stone surface,
soft side light, simple background, clear outline
Model: record the actual selected model
Ratio: record the selected square ratio
References: record the actual inputs, or none
Other settings: record deliberate changes, or noneThe placeholders are intentional. Fill them from the current interface rather than copying a model or parameter value from an unrelated guide. A durable record describes what you selected, not what you assume the default ought to be.
Check the first folder before testing the second
Open the first project's folder, inspect the prompt bar and choose the intended settings. Clear stale references that do not belong to the task. Run one small check and inspect the resulting image's recorded settings before you start a larger batch.
Now move to the second folder and repeat with its different brief. Do not change the model, ratio and style merely to make the test more complicated. Change only what the second project requires. The purpose is to discover whether the visible state matches your intention, not to maximise the number of variables.
Return to the first folder and compare the active settings with your record. The latest changelog supports persistence across sessions, not only within the current tab. Still check the references and prompt against the task; a remembered setting may belong to an older experiment.
Test boundaries without assuming the outcome
Use this table to record the conditions that matter to your own work. Leave it blank until you perform the checks. Changing browser state may affect more than folder defaults, so save useful prompts and outputs before testing a fresh session.
| Situation | What to inspect | Your observation |
|---|---|---|
| Return to a folder in the same tab | Prompt parameters and references | Record after checking |
| Open another tab | Active folder and visible settings | Record after checking |
| Start a fresh browser session | Defaults, prompt and selected inputs | Record after checking |
| Hand work to a colleague | Their settings against the written brief | Record after checking |
Cross-session folder pills are now documented. The table is a practical verification record, not evidence that every field synchronizes across every browser, device or collaborator. Record the conditions you actually checked before promising that behavior to a teammate.
Treat advanced parameters as deliberate choices
The October changelog mentions pinning --exp as a folder default through advanced settings. It does not mean every project needs that parameter. If you are not testing its effect, leave it out of your exercise and focus on the settings your brief actually requires.
When you do change an advanced value, record it and compare outputs under otherwise similar conditions. A striking result does not prove the setting helped. Ask whether it improved the subject, composition or material treatment you needed. Remove experimental values before handing a stable project to someone who did not participate in the test.
The parameter guide explains broader prompt controls. Keep this folder exercise separate from learning several new parameters at once. Otherwise, it becomes difficult to tell whether a surprising result came from remembered state or an unfamiliar instruction.
Know when a setting is irrelevant to the current action
The same October source says edits keep the base image's aspect ratio, preventing a stray --ar from stretching the result. That is a useful reminder that generation and editing do not always interpret context in the same way. A portrait generation setting is not a reliable instruction to recrop an existing landscape image.
Before an edit, state whether you want to change the contents or the canvas. Then inspect the exported dimensions. If the task requires a different final format, plan a crop or canvas workflow explicitly rather than relying on whatever ratio happens to remain in the prompt bar.
Finish with a handoff another person can use
Save the selected output, exact prompt, active references and short acceptance criteria together. Include a sentence about any setting that did not behave as expected. The next person should be able to understand the goal without reconstructing your browsing history.
At the start of their session, they can compare the visible controls with that record and run a small check. Folder memory then becomes a convenience, while the brief remains the source of intent. For broader asset organisation, see the Alpha workspace guide.
Source: Midjourney Alpha changelog, October 1, published October 2, 2026.