Midjourney refines the quality of V8.1

Midjourney V8.1 exits alpha on midjourney.com with sharper SREF style references, a faster native 2K HD mode, and the return of the Describe tool.

Midjourney published an update to V8.1, its generative AI image model, on April 30, 2026, focusing on output sharpness and fidelity to stylistic references. The improvement primarily targets SREF (Style References), moodboards, and HD renders, but more broadly affects all generations. Outputs obtained by stacking multiple SREFs gain stability, and an older generation relaunched with the same seed should appear more precise.

V8.1 simultaneously exits its alpha phase and becomes accessible on midjourney.com, in addition to Discord where the model was already deployed. The version had been opened to alpha users on April 14, featuring a return to the V7 aesthetic, deemed less flat than V8.0, a native 2K HD mode three times faster and three times less expensive than the previous generation, the return of image prompts, and a Describe tool allowing the extraction of a prompt from an existing image.

To absorb the server transition related to this opening, Standard Definition mode remains set by default. Switching to HD is done from the settings or by adding `--hd` to the prompt. The team is simultaneously continuing to collect user ratings on midjourney.com/rank-v8-1, with a view to training the future V8.2. Several projects are announced next: models dedicated to image editing and a multi-reference system covering characters, environments, and objects.