Franklin AI News Brief

Midjourney tests a Thinking rerun for prompt accuracy and typography

Key Takeaways

  • The Alpha test lets users rerun an image with additional thinking and send feedback on where it helps.
  • A new way to revisit an image Midjourney is testing Thinking Mode on its Alpha website.
  • In its October 8 announcement, the company describes a rerun option that applies additional thinking to an existing image job.
  • Users can open a job’s lightbox and select Rerun (Thinking) in the lower-right corner.
  • The entry point matters: this experiment starts with an image a person already generated, which gives them a concrete result to compare with the new version.

A new way to revisit an image

Midjourney is testing Thinking Mode on its Alpha website. In its October 8 announcement, the company describes a rerun option that applies additional thinking to an existing image job.

Users can open a job’s lightbox and select Rerun (Thinking) in the lower-right corner. The entry point matters: this experiment starts with an image a person already generated, which gives them a concrete result to compare with the new version.

Compare the parts that matter to the prompt

Midjourney says its tests show improvements in prompt accuracy, typography and coherence. It is asking users to report cases where the option helps and where it introduces problems. Those are the company’s early observations, rather than an independently measured guarantee for every prompt.

For a useful comparison, keep the original image and note the exact requirement it missed. A poster might need a readable phrase; a scene might need the right number of objects or a specific relationship between them. Compare those details in both versions instead of judging only which image looks more attractive.

An experiment with an open future

The company is considering a broader thinking mode or a way to add more thinking after a job. It has not presented that possibility as a confirmed general release in this announcement.

People trying the Alpha feature should record examples that make the difference easy to inspect. Include the prompt, the original result and the rerun, with a short explanation of what improved or became worse. That feedback is more informative than a general vote for a favorite image. It can also help an individual decide when another pass is worth trying and when revising the prompt would better express the intended result.

Our read

Franklin AI Take

Midjourney is testing Thinking Mode on its Alpha website. The most useful test is a side-by-side comparison against one explicit prompt requirement. Keep examples where it fails as well as where it succeeds, so an attractive rerun does not conceal a missed instruction.