Is Midjourney, the "great demon king" of image generation, now getting into video generation too?!
The above demonstrates a video effect.
You can see the running movements, character, and spatial transitions are very smooth.
The scene below, where someone is digging into a cake, is not only realistic but also has a reflection on the spoon, which is very detailed.
This news caused a stir, with Reddit upvotes reaching 2.5k.
It also sparked heated discussions among netizens.
Some said, "This is the first time I thought it was a manually shot video," and "It's almost indistinguishable from reality."
Not only is the video model performing well, but Midjourney's image model V7 is also continuously being updated.
Amazing results, and you decide the price
Let's look at more effect demonstrations.
Multi-character movements and perspective changes are also very smooth.
The cat's movements and the human hand's movements both have a strong sense of physical realism.
Here comes the dog skateboarding~
Car drifting is also no problem.
The cat's manicure is indeed very fine, but even more detailed are the textures on the hand; there are even fingerprints on the fingers (although some fingerprints are missing).
However, as that netizen mentioned above, some aspects are still not quite rational.
For example, in this scene of folding a blanket, although the wrinkles caused by the hand's force are considered, giving a certain physical realism, it later looks like the blanket shrunk back by itself...
And this one, why does it feel like climbing a meaningless flight of stairs, and the flower in the woman's right hand suddenly floated to her left hand, just so her right hand could grab the railing.
Overall, Midjourney's video generation model performs very well in terms of physical realism, texture details, and movement smoothness.
However, if you've seen the effects of Veo 3 before, wouldn't you feel something is amiss with Midjourney's videos—
No audio functionality.
Yes, netizens also noticed this.
In the case of playing the violin, Midjourney's version only has music added in post-production.
Whereas Veo 3 can generate the sound of the violin.
Thus, some questioned whether Midjourney's entry into this field now is a bit late?
However, just two days ago, Midjourney held a company meeting where they showcased some video generation demos and mentioned "animated images," which seems to be a distinguishing feature from other video generation models.
In fact, compared to realistic styles, anime style is what Midjourney is better at.
Currently, Midjourney's video model has not been officially released and is undergoing final improvements.
The team is calling for active participation in video rating to help the model learn the combinations of movements and compositions people like to see in videos.
Furthermore, Midjourney expressed its sincerity by stating that they hope everyone provides suggestions so that the pricing can meet everyone's needs.
It must be said, this move is very sincere.
Midjourney V7 supports voice-to-image generation
In addition to the video model, the image generation model Midjourney V7 is also continuously being updated.
Since March this year, Midjourney has been continuously calling for users to actively participate in image rating to finalize V7.
In April, Midjourney released V7 alpha.
There are two versions: Relax and Turbo modes.
Below are some example images, where you can see that V7 generates highly realistic hand textures.
V7's flagship feature is "Draft Mode."
When using this function, the prompt bar will change to "Conversation Mode."
For example, tell it to replace a cat with a falcon or turn it into night, and it will automatically process the prompt and start a new job.
Click "Draft Mode" and then the microphone button to enable "Voice Mode" – you can think aloud and let images flow like a dream in the generation area.
In other words, you can generate images by speaking, and multiple images can be generated for you to choose from~
Draft Mode halves generation cost and increases image rendering speed by 10 times.
Currently, the team has distinguished "Draft Mode" from "Conversation Mode," allowing you to freely choose whether to use these functions separately or in combination.
The team also launched V7 Fast Mode, which updates the acceleration function.
This means model optimization will take 40 seconds in Fast Mode and only 18 seconds in Turbo Mode.
Through continuous efforts by the team, Midjourney V7 image generation speed has increased by about 40%.
Fast Mode job rendering time decreased from 36 seconds to 22 seconds.
Turbo job rendering time decreased from 13 seconds to 9 seconds.
With image model V7 continuously updated and a video model coming soon, Midjourney is indeed the king of visuals!