The narrator provides a tutorial on how to create a free AI talking avatar using various online tools. Key steps include:
Generate an avatar image prompt using Leonardo Ai
Uncrop the image using an uncrop AI tool
Create a talking head video using Talking Heads website
Add your own custom audio or text-to-speech
Enhance video quality using vmake video enhancer
Remove watermarks using video editing software
The narrator states this method allows creating free AI talking videos for now, but the websites may start charging in the future.
The narrator analyzed Niko Omilana's videos and found he uses a three-step formula to beat the YouTube algorithm:
1. Hook viewers in the first 10 seconds with a "cold open" that shows intense action related to the video topic. This exceeds expectations set by title and thumbnail.
2. Use "mini storylines" every 5 minutes to continually re-engage viewers. Each storyline lasts around 4 minutes building up to a payoff, keeping viewers hooked.
3. Create effective thumbnails by showing rather than telling, leaving unanswered questions, and keeping it simple with one main focal point.
By following this formula, Niko is able to get high watch time and clicks to beat the algorithm.
Here is a condensed summary of the main points from the video:
The narrator shares various tools to help improve AI art created with Midjourney, including:
GFP-GAN - A free online tool to fix faces in images using AI.
Adobe Photoshop - Recommended for editing, adding text, cutting out images, and touching up artwork. Alternative options are Luminar Neo and Canva.
The Adobe Express Background Remover is a free tool to remove backgrounds from images.
To learn how to create better prompts, the narrator suggests referencing sites like Free Flow, Twitter under the #promptshare hashtag, and Midjourney's own community feed.
A face swapper plugin called Insightface for Discord swaps faces in images using AI.
Vectorizer.ai converts images into vector graphics.
For upscaling images, Topaz Gigapixel AI is recommended over Midjourney's built-in upscaler.
The narrator also shares tips on selling AI art on sites like Wirestock, Redbubble, and Displate.
Using prompt generators can help improve prompting skills.
The narrator shares 9 lessons he wishes he knew when starting on YouTube:
1. No one cares about your videos at first, not friends/family or the audience. Make videos that provide value to the audience.
2. Most beginners should take an "archaeologist" approach - try different video topics until you find what works rather than overplanning upfront.
3. Consider supply/demand - crowded niches require unfair advantages to stand out. Find a niche with less competition.
4. Think in systems to streamline and leverage your efforts across videos versus treating each video independently.
5. Optimize titles and thumbnails first since that determines if people will click. Match to your target audience.
6. Hook viewers in the first 30 seconds before dropoff.
7. Gear doesn't matter much initially. Focus on content. Production value depends on the niche and audience.
8. Find an "unfair advantage" that makes you stand out when growth stalls.
9. Outsource editing as soon as possible to free up time for strategic tasks.
The narrator explains how to use AI to turn your voice or any instrument into a different sounding instrument. This is done by training an AI model on samples of an instrument in a DAW, exporting them as a WAV file, and uploading to Musicfy to train a custom model.
It takes 1-2 hours to train the model. The $10/month Musicfy subscription allows 1 custom model. More variety in notes, articulations, etc. improves the realism. The resulting AI instrument can be used with singing, beatboxing, other instruments.
Examples are shown turning a flute into a violin, and many other transformations. Limitations are it's not real-time yet. However the narrator sees great creative potential, especially for those unable to play real instruments, as long as AI enhances rather than replaces human creativity.