The team from Atlabs.ai reached out to let me know about their new AI video generation platform. Among many unique features, it can be used to generate a music video by uploading an audio file, image, or video. In the case of providing an audio file, Atlabs will scan the file for style genre, tempo, mood, key scene references, and recommend a general story concept – or you can prompt your own story and characters.
There are other tools too such as script to video, product promotion videos, placing your own image into a music video, and much more. You can do lip syncing, add motion, create avatars, clone product videos, upscale videos, modify videos, and choose from premade video styles or browse what the community is creating.
They were kind enough to let me demo their platform in exchange for a review. If you haven’t already read my reviews of other AI Video Generators check out my post for Revid.ai and Vidmuse.ai here.
I was able to create 3 videos in the duration of my free trial. I chose to test music videos as that is what I have reviewed and created previously so I would have a good baseline comparison. I chose 2 vertical 16:9 tiktok style canvases. 1 “cute” anime style and 2 photo realistic. I started a 4th in a different more classic anime style but ran out of credits before completion.
Speaking of credits, let’s take a look at cost effectiveness. When I started the 4th video 70 credits, and it told me I needed about 310 credits. I could have bought 2 80 credits packs for $10 each for a total of 160 credits, or for slightly more got 250 credits for $27. This makes Atlabs very affordable and good value.
As for the video creation process itself, it is very easy and user friendly even if you do not have an idea exactly for how you want your video to look or feel, Atlabs will help you select a basic flow. I chose narrative for all 3 of my videos. You can also choose a dance style where there is no story and just a dancing lip syncing performative video.
After uploading your audio file, Atlabs will sync the lyrics to display and analyze the tempo and genre and vibe of the video. Atlabs will generate your main character(s) and you can always add extra characters as you’d like. I find it is better to add these characters before moving forward with generation. Adding them after and trying to insert them into scenes costs extra credits and sometimes ignores the prompt to insert them or does so unnaturally. This happened to one of my videos, but the 2nd time when I defined the cast ahead of time, it turned out perfect. Character consistency was very high in all of the videos I made. The photo realistic videos had great skin and hair and clothing texture and natural lighting and movement. The anime video seems kinda blurry to me compared to some competitors, but the character consistency still makes it a great AI video.
Most of the granular editing has to be done on a desktop computer, which is a bit of a let down as I enjoy exploring AI while killing time, doctors offices, passengers in a car on a road trip, late at night lying in bed, etc. I hope they will update their mobile capabilities in the future. Also I did encounter 2 errors, however support was very helpful. One error was fixed within the same day, and the other in about 4 days.
Let’s take a look at the creation process (mix of mobile and desktop screenshots showing the creation process):
Ready to take a look at the final results? Let’s start with “Big Breakfast Boy” – This is the one I decided to edit after my first generation where I just went with Atlabs suggestions. I wanted to make it more like a kpop video by adding some backup dancers. I feel like it would have been better if I had added them prior to the script and scenes and next steps being generated. Here are the before and after.
And here was the original video before I tried to add more characters. Pretty similar really. And both miss the mark a bit for a smooth synchronized kpop boyband dance video.
In comparison here are some other versions in other tools (Sora and Revid to name a few) – that’s why I chose to start with Big Breakfast Boy since I had a good comparison to draw from.
Sora (plus pic play post mobile app):
Revid:
Next, here is an Atlabs example same photorealistic style to my song “Dismissed”. This time I created the entire cast, and I chose my own prompt for the scene and characters in great detail. I am more pleased with how this one turned out.
Lastly, here is the “Cute Anime” video to my song “Gamergirl Backlog”. It is indeed very cute, but a little low-res, no? Especially in comparison to the clarity of the photorealistic videos, or competitor anime styled videos. Still cute though!
Overall I would definitely put Atlabs above Revid given the character consistency, but I still prefer Vidmuse for video quality and finer controls over custom styles and chat interface guided creation. I think Atlabs has an advantage over Vidmuse in terms of pricing and ease of use for beginners allowing them to jump right in. I also like the Atlabs has a strong understanding of the lyrics – Vidmuse does not add captions automatically, and Revid tries, but needs a lot of correcting. Atlabs had about 98% lyric caption accuracy, and pricing that can’t be beat.
Visit https://app.atlabs.ai/subscribe to get started.
