Midjourney Video V1 made both ends of a shot selectable
On July 21, 2025, someone editing architectural renders asked whether Midjourney could generate a transition from the end of one clip to the beginning of another. A reply described the option available to them at the time. They could make the last image of one clip the starting frame of the next, while the animation tool chose where that new clip ended. The thread documents a request for a bridge between render angles and leaves the eventual project outcome open. July 21 architectural-renders thread
Midjourney launched Video V1 on June 18 in UTC, June 19 in Seoul, as a web-only image-to-video model with a starting image and optional motion instructions. The July 25 update added a second image defining the ending frame. It was published at 00:33:24 UTC, or 09:33:24 in Seoul, according to the archived publication ledger. The same update added looping, which uses the starting image as the ending image, and brought video generation to the Discord bot. On the web, the prompt bar provides a separate Ending Frame field. Discord uses --end image_url for a different ending image and --loop to return to the starting image. June launch · July announcement · Midjourney archive · Midjourney Video help
A launch-day Reddit discussion relayed both controls. Its original submission documents awareness of the update. July 25 community thread
What creators did with the two images
On July 26, Tennen described using the new end-frame control for a Heartbroken Girls transformation in which a character’s hair rises and turns gold. The work credits NijiJourney and Midjourney for the artwork and NijiVideo for the July animation. Tennen had also animated an April rendition with Vidu AI, which already offered an ending-frame control. The creator favored the July version’s anime hair movement and the Vidu rendition’s effects. The creator’s account documents an early use of the control on the NijiVideo surface. It leaves the exact underlying model job unspecified. Tennen’s July production note
In September, Hitsuji used Nano Banana to make a photograph-like still of an intact camera on a table and a second still with the components laid out. They assigned the images to Midjourney’s frame slots and generated the motion between them. The author observed the parts changing shape during the clip. A July Reddit showcase drew a related timing criticism from a viewer, who saw abrupt changes near the opening and closing frames and wanted easing. Hitsuji’s camera and transformation demos · July 27 showcase and July 28 replies
A September Reddit creator used their own photographs, keeping white cars as a recurring element, and assigned photos to the start and end slots to bridge shots. The maker then adjusted timing in Premiere Pro. This is a realized version of the cross-shot workflow the architectural-render thread described in July. It illustrates a division of work visible in the creator’s account. Choosing both images gives the editor the intended endpoints, while the generated motion and final pacing still depend on the clip and the edit. White-car sequence and creator’s workflow reply · July architectural-renders thread