As Nvidia-backed Sora competitor Luma AI launches Dream Machine 1.5 version amid fierce competition in the AI video sector.
According to Luma's founding members, compared to version 1.0 released in June, this upgrade mainly focuses on text-to-video generation and is overall more intelligent. The new version is now fully available for online trials and has received positive feedback from users.
Based on user-shared trial results, Dream Machine 1.5 performs well in several aspects:
- Can generate creative scenes, like a runny egg yolk riding a bicycle
- Excellent visual effects for fast-moving cameras, comparable to cinematic camera movements
- Solved the challenge of adding text to images
- Can achieve complex effects like split screens
- Natural and smooth dynamic effects like flowing clothes
These improvements have led to rapid growth for Luma AI in the past two months. According to Similarweb data, Luma AI's global visits reached 22.6 million in July, ranking first among AI video generation tools.
In actual tests, Dream Machine 1.5 performs well in text-to-video generation:
- Can generate scenes that match prompts, but details like hands still have flaws
- Dynamic effects like smoke and cat ear movements are very realistic
- Text addition function is basically usable, but occasional spelling errors occur
The company behind Dream Machine is Luma AI, with co-founders including a former Apple AR/CV engineer and a Berkeley computer vision researcher. The company completed a $43 million Series B funding in January, led by a16z with participation from NVIDIA and others.
Overall, Dream Machine 1.5 shows good potential in AI video generation but still has room for improvement in detail processing. As technology continues to advance, the application prospects of AI video generation tools are worth looking forward to.