Day 14 of 30 Day AI Challenge – AI Generated Video, A Double Edged Sword?

Two weeks into the #30DayAIChallenge and we’re going to take a closer look at how AI can play a good (and bad) role in video production and editing. For those just joining us, every day during this challenge, I’m going to try and do something different or better, using currently available Artificial Intelligence tools.

Let’s kick off with some real-world examples and tools that demonstrate the two sides of AI in the realm of video production.

Good AI: Streamlining Video Production and Editing

1. Adobe Sensei: Adobe’s AI-powered platform, Sensei, has been a game-changer for video editors. It accelerates tedious tasks like masking, tracking, and color grading. With features like Auto Reframe, editors can quickly adapt videos to different aspect ratios, saving hours of manual work. Adobe also also relaunched Firefly (currently for static images) which allows users to tap into generative AI and create whatever their imagination holds.

2. AI-driven video summarization: Summarization tools like VidRovr and Summarize.Tech employ AI algorithms to condense long form content into shorter, digestible summaries without losing the essence of the content. This saves editors a significant amount of time and effort.
3. AI-generated music: Platforms like Amper Music (acquired by ShutterStock) use AI to compose original music tracks tailored to specific video lengths, styles, and moods. These tools provide creators with unique and royalty-free soundtracks, eliminating the need to rely on stock music libraries.

Bad AI: Ethical Concerns and Misuse

1. Deepfakes: Deepfake technology, which uses AI to manipulate video and audio content, has raised serious ethical concerns. By replacing a person’s face or voice with someone else’s, bad actors can create misleading or harmful content. While the movie industry has used deep fakes to their (positive) advantage, we’ve already seen the negative uses play out in the Adult content industry and also in political propaganda. While deepfake detection tools like Microsoft’s Video Authenticator are being developed, the rapid advancement of deepfake technology poses a continuous challenge.

2. Privacy invasion: AI-driven video analytics tools can analyze faces, emotions, and demographic information in video content. While these insights can be valuable for marketing and advertising purposes, they raise privacy concerns, as people may not be aware their data is being collected and analyzed.
3. Bias and discrimination: AI algorithms can inadvertently perpetuate bias and discrimination if they’re trained on biased datasets. For example, AI-generated music tools may favor specific genres or styles based on the training data, which could lead to unintentional cultural bias and homogenization of creative output.

The Latest Craze: Text to Video

Just as DALL-E, Stable Diffusion, Midjourney and other tools have made text to images a popular craze at the moment, an emerging trend is text to video. I ran a quick search to see what’s out there and what it can do:

• Synthesia: Generate videos using 70+ AI avatars in 65+ languages, with various templates and a free media library. I used the free demo tool to create a short video, which looks pretty good, though you can still make out the voice is not natural/human.

The above is a screen recording so the audio is pretty faint… you can also check out the original video here.

• Elai.io: Generate videos at scale from text or blog posts, with over 60 languages and 25+ avatars to choose from. For this test I put in a blog link and chose a template…then let the system do its thing.

Elai.io spits out multiple slides, where you can make adjustments to the script, images, voice, etc. Once you’ve fine-tuned everything, you can render your final video. Notice, though, that the system doesn’t animate the characters when creating video.

• Synthesys: Produce lip-syncing AI videos with 74 real “Humatars” and a voicebank of 254 unique styles in 66+ languages.
• Make a Video Studio (by Meta): Make-A-Video harnesses text-to-image technology to create unique, imaginative videos from just a few words or lines of text. There’s no way to currently get on the waitlist or try a demo, though, so I couldn’t test its capabilities.

• Google’s Text to Video: Phenaki + Imagen, while not available yet to the public looks pretty promising- check out the video demo below.

• InVideo: Easily create marketing and explainer videos from input text using professionally designed templates.
• Pictory: Create and edit high-quality videos based on articles or scripts, with automatic captioning and summarization. To test this I chose the URL option as well, with the same blog post I used with Elai.io – the summary (StoryBoard) generated needed a LOT more work, though I noticed Pictory inserts stock video footage (vs stills) into the different ‘scenes’, which makes the video much more dynamic.

• Synths Video: Transform blog posts into high-quality videos with natural-looking lip movements and 40+ human avatars.
• Veed.io: Generate videos with customizable text, fonts, colors, and themes, plus one-click transcription.
• Deepbrain AI: Quickly create AI-generated videos using basic text and a simple three-step process.

I provided a simple ChatGPT prompt (something along the lines of ‘how AI will change the world’) and pretty much let the system do the rest. Here’s what it produced (I made no edits to the script):

Deepbrain probably produced the most impressive video (see example above) – the interface is clean, the script to the point and you have loads of options for narrators. If you didn’t know any better, it could almost pass as a normal video narrated by a flesh-and-blood human being. Pretty impressive…and I suspect this will get a lot better (and creepier) as the tech matures.

In Conclusion

As you can probably see from the pros, cons and examples above, the potential of AI in video production is immense, but it’s crucial that we remain vigilant about its ethical implications. By striking a balance between harnessing its power and addressing potential pitfalls, we can ensure that the tech is mostly a force for good.

That’s it for today’s deep dive into the world of AI + Video. What are your thoughts on the matter? Any other interesting tools out there? Share your insights in the comments below!

Till tomorrow…