AI&Design
All change articles

Midjourney · Announcement · Official source

Midjourney Video V1 brought motion into the image workflow

A creator opened earlier Midjourney images and used Auto Low Motion to turn them into a collection of clips. Some stills needed more direction. The creator described specifying a stationary camera, naming the object that should move, or spelling out what was happening in the scene. Reusing an existing image made it easy to start, and the added motion direction still took iteration. Low Motion Showcase

What shipped in June

Midjourney released Video V1 on the web. A generated image or an uploaded still could serve as the starting frame for a five-second clip. Automatic animation supplied a motion prompt. Manual animation let the creator describe the movement and how the scene should develop. Low and High Motion were separate choices. Midjourney said Low Motion could result in almost no movement and High Motion could produce more visible errors. At launch a job produced four candidate clips, and a chosen clip could be extended up to four times in roughly four-second increments. Midjourney launch announcement

Midjourney’s archived launch entry records 16:35:15 UTC on June 18, which converts to 01:35:15 on June 19 in Seoul. This accounts for the June 19 release record and the June 18 Reddit submission headers. Midjourney archive The June launch offered a starting still and clip extension. Midjourney announced a user-specified ending image, looping, and Discord video generation on July 25. The later V8.2 Edit Model belongs to the image-editing workflow. July video announcement · Midjourney image model versions

What creators made

On June 19, a creator shared a stylized scene that included a walk and a close-up. Their reply describes deciding how the action should continue after seeing those shots. They pointed to Extend Auto to keep the existing prompt and Extend Manual to change it. The creator praised continuity as the scene progressed. A viewer described the visual style shifting within the same shot, from smooth 3D through cel-shaded rendering to 2D. The post gives no full prompt or job history, and the creator said they had yet to test the same character in another scene. Scene extension thread

Two other accounts give context for those observations. On June 18, eStuffeBay posted a clip and credited Midjourney for both the image and the animation. In replies, they reported that clearer motion descriptions could help and that limbs sometimes moved too much. On June 20, Cagnazzo82 reported testing native stills and ChatGPT images from the Sora gallery. They thought faces and hands held up in extensions and found the physics and resolution less convincing. These observations come from individual jobs and give no success rate. Launch-day 2D test · Imported stills and extension discussion

On July 14, outside the June 12–July 3 search window, the creator of MARGHERITA explicitly linked their anime-style short to testing the first Midjourney video model. They credited Midjourney with image and video generation, ElevenLabs with voices, Artlist with music and effects, and DaVinci Resolve with editing and grading. They described using Omni Reference to prepare recurring character images. The creator reported difficulty getting a character to turn in the intended direction and keeping the bike, two characters and a car consistent across shots. The production credits establish a finished multi-tool workflow. Shot settings and source jobs are undisclosed, so the cross-shot problems cannot be assigned to a single stage. MARGHERITA production thread

Evidence limits

The June 18 Reddit launch thread reproduces vendor communication. The other examples use original signed-out creator posts and separate author replies for workflow detail. Reddit’s indexed submission headers supply the absolute dates. The accounts establish reported use for existing stills, same-scene continuation and an assembled short. They leave generation reliability, shot-to-shot consistency, and the relative performance of competing tools unmeasured. Reddit launch announcement