[01:08 Thu,19.June 2025 by Thomas Richter] |
Long awaited, the time had finally come - Midjourney, provider of one of the most popular and best image AIs, has released its first (V1) video AI model. As an image-to-video model, the generation of a clip starts with an image - either from Midjourney or from another source. This can then either be animated automatically (the text prompt for the movement is then generated by AI) or - for more control - manually using your own text prompt that describes exactly what should be animated and how. ;)
To start, the Video Model V1 is initially only available via the web interface, not via Discord. Generating a video costs approximately eight times as much as an image and delivers four videos of five seconds each. According to Midjourney, this is many times cheaper than comparable offers on the market. According to Midjourney, the actual operating costs for the new video model and, consequently, the pricing are difficult to estimate. Therefore, Midjourney wants to observe how the new model is used in the next month and then - if necessary because server capacities become scarce - adjust the prices accordingly. The currently cheapest way to try out Midjouney Video V1 is the Basic subscription from Midjourney at a price of (currently reduced) 8 dollars per month, which should be sufficient for the generation of around 25 video clips. Subscribers of Midjourney&s 60 dollar per month Pro Plan and the 120 dollar Mega Plan can even create unlimited videos in the (known from image generation) slower "Relax" mode. video.twimg.com/amplify_video/1933307710163804160/vid/avc1/3840x2160/F5D4Te50W9FFnVlc.mp4 Limited OptionsThe degree of motion can be determined using the two options "high motion" for scenes in which both the camera and the subject should be moved, or "low motion" for quiet scenes in which the camera hardly moves and the subject moves only slowly. The former has a higher susceptibility to image errors, the latter sometimes results in no movement at all. A generated video can be extended up to four times by 4 seconds each, i.e. a video can be up to 21 seconds long in total. However - typical for the first generation of a video AI - there are still no possibilities for more precise control of the camera movement, character consistency or for editing clips. video.twimg.com/amplify_video/1933991664802000896/vid/avc1/1080x1080/xC2mNGJ1EWcAwbwJ.mp4 In view of the great competition from other video AIs such as Google&s Veo 3, OpenAI&s Sora, Runway&s Gen 4, Pika, Adobe&s Firefly and the Chinese models, Midjounrney seems to want to score points through a low price and - as with its image AI - through a particularly artistic look, not through realism. Future Music: An Open World in Real TimeThe actual goal of Midjourney is set much higher, namely an AI model that is able to simulate an open world in real time, i.e. a dynamic 3D environment with which one can interact live including the objects contained therein. On the way there, a video model (moving images) is the second step after a model that generates images of any world. The next development stages are a 3D model that allows you to move freely through the created world and ultimately a model that enables this in real time. These steps are to be realized in the next year - initially the use will be quite expensive, but should then become affordable. Bild zur Newsmeldung:
 more infos at bei www.midjourney.comdeutsche Version dieser Seite: Midjourney startet eigene Video-KI V1 |
;)
 |
|
|