With more and more Gen AI video tools coming out, I wanted to give the most popular ones a test to see how they stack up. Here are the tools I tested:
- Gen-3 Alpha by Runway
- Kling Video 1.5 by Kling AI
- Dream Machine by Luma AI
- Pika 1.5 by Pika Labs
- Hailuo AI Video-01 by MiniMax
Of course, we also have OpenAI’s Sora and Meta’s Movie Gen, however these haven’t been made publicly available quite yet. Here’s a collage of the final results (background music was created using Suno and sound effects using ElevenLabs).
AI Video Generation – The Three Tests
I ran each model through three different scenarios:
- Text to Video Prompt: Starting from scratch with just the following text description: In Paris, Godzilla towers over the skyline, roaring as he crushes buildings beneath his feet. The Eiffel Tower, once a proud symbol of the city, bends and cracks under his enormous weight. As he sweeps his tail through the air, cars and trees are tossed like toys, and the streets around the Seine river are flooded with panic.
- Image to Video Prompt: Using a static image generated using Midjourney as a starting point (see screenshot below).
- Image to Video with a Gen AI Self-Portrait: Feeding in a generative AI image of myself as a Mad Max inspired desert warrior to see how well the models handle personalization and facial coherence. If you’re interested in learning how the base image was created, check out this guide:
Gen AI Video Tools – Evaluation Criteria Used
I evaluated each model based on several key dimensions:
- Ease of Use – how simple the interface is for most people
- Speed – how quickly the tool generates outputs
- Cost – for paid plans, the general affordability of each tool
- Quality (Text to Video) – how well it produces video from text prompts
- Quality (Image to Video) – how well images work to produce videos
- Customization & Control – additional abilities like providing first & last frame, aspect ratios, negative prompts, camera control, etc.
- Free Tiers and Paid Plans – is it possible to try these tools for free and how much do monthly and yearly paid options cost.
Now, let’s get to the juicy part!
Runway Gen-3 Alpha
I’ve been using Runway for a while now and Gen-3 Alpha does a tremendous job of converting images (and videos) to videos…it’s not quite that great at just text to video generation though. There’s also the ability to specify camera motion, aspect ratio, apply presets and extend video generations. You can create either 5 sec or 10 sec videos in HD (not full HD) at present.
Kling AI
Kuaishou’s Kling AI was initially only available in China, but since summer 2024 has gone worldwide and has recently launched Kling v1.5. Kling offers a simple somewhat cluttered interface, with the ability to provide text and image prompts, change a slider to prioritize creativity vs relevance, aspect ratios, number of video generations and lengths of 5 or 10 sec. The older version also offers the ability to specify camera movement, while 1.5 doesn’t have this enabled quite yet. Video output quality is decent, however the speed of generation quite slow.
Luma Dream Machine
Luma AI’s Dream Machine was also launched over the summer, initially with just text to video capabilities, but soon with image to video too. I was impressed with the speed of generation and the interface is quite simple and uncluttered too. It did well with both text and image to video, and with my own image it did a decent job of preserving character consistency for much of the video. You can generate 5 sec clips with the option to extend them. With image prompts you can also provide start and end frames.
Pika Labs v1.5
Pika Labs made v1 available at the end of 2023, mainly through their Discord interface and very recently launched Pika v1.5. Along with the latest version of Pika came the ability to upload images and apply “Pikaeffects” that inflate, melt, explode, squish, crush and cakeify your subject…this, of course, went instantly viral. While these Pika effects work (generation is slow), the model doesn’t seem to be able to produce other text or image to video outputs yet…all you get is a ‘experiencing high demand’ message that never goes away.
HailuoAI Video-01 (MiniMax)
HailuoAI video-01 is the latest entrant on this list, launched by Chinese company MiniMax in September…outputs from this Gen AI video tool have been doing the rounds on social media ever since it launched. The tool offers an unlimited use trial for 3 days, so you can get a feel for how well it works. The interface is pretty simple and you can provide text or image prompts, with the ability to enhance prompts. Video generation speed and quality is pretty good.
The Verdict
After putting these models through their paces, here’s how they ranked overall – as you can see, Runway’s Gen-3 Alpha, Luma’s Dream Machine and MiniMax’s Hailuo Video-01 models are neck and neck, but with slight differences. Kling does decently and Pika 1.5 gets a very low score primarily because the model can’t seem to handle any generations beyond Pikaeffects at the moment (this will likely get resolved in time).
Recommendations:
- Runway Gen 3 Alpha (Best Choice For Professionals): I use this the most and I’d recommend the unlimited plan if you intend to churn out a lot of content, or need plenty of re-rolls in order to create just what you need. The image to video outputs are far better than the text to video outputs and some of the customizations and features like lip sync, video extensions, presets, ability to save work sessions and various other tools that Runway offers give it an edge over the others.
- Luma Dream Labs (Best Choice For Hobbyists): Excels in both text and image to video, with quick generations and the simpler pricing model (including 30 free generations per month) make this a great choice for hobbyists and those looking to play around with Gen AI video generation.
- MiniMax’s Hailuo Video-01 (Rising Contender): A great new contender for the crown, with excellent video outputs…I was particularly impressed by the text to video output. Given the tool is quite new, it doesn’t have all the bells and whistles yet, but is likely to be a very strong contender in the months ahead.
- Kling AI Video 1.5 (Slightly Cheaper All Rounder): Great for users who want a balanced suite of options at a lower price, but are willing to put up with longer generation wait times.
- Pika 1.5 by Pika Labs (Not Ready For Prime Time): Probably released too early as it can’t seem to handle text and image to video outputs…the Pikaeffects (which do work well) are loads of fun to play with though!
Final Thoughts
Choosing the right Gen AI video tool depends on your specific needs. If you’re looking for top-notch text to video capabilities, Runway Gen 3 Alpha (and Alpha Turbo) is still probably your best bet, but Luma Dream Labs and Hailuo by MiniMax are strong contenders that will likely keep getting better. Let me know what you think and which tool you prefer the most…and why.
Happy generating!