In this episode of How I AI, I was thrilled to chat with Anish Acharya, a general partner at Andreessen Horowitz specializing in consumer investing and AI-native products. We moved beyond typical B2B AI discussions and dove into creative and personal AI projects. Anish demonstrated how anyone can harness AI's power, even for fun, artistic endeavors.
We explored three distinct workflows that illustrate AI's versatility and potential to unlock creativity and productivity.
Workflow 1: Creating AI-Generated Music Videos
Anish's first project was generating AI music videos, beginning with a Tiny Desk Concert featuring Kurt Cobain.
Step 1: Image Generation with GPT-4o
Anish used GPT-4o to generate the initial image: "generate an image of Kurt Cobain playing a Tiny Desk concetrt".
He refined the prompt iteratively, removing the guitar for an acapella performance to better fit the Tiny Desk aesthetic.
Step 2: Animating Still Images with Hedra
The generated image was then imported into Hedra, a tool that animates still images and adds custom audio with lip-sync.
Step 3: Audio Extraction and Syncing with Adobe Audition
Anish used 4K Video Downloader (acknowledging its slightly sketchy nature) to download audio of Nirvana’s Unplugged concert from YouTube. Using Adobe Audition, he extracted and synced the audio with the animated video in Hedra.
Step 4: Vocal Extraction with Demucs
For an acapella version, Anish employed Demucs: demucs two-stems vocals /path/to/audio.mp3 to separate vocals from instrumentation.
Step 5: Final Video Assembly with Kapwing
Finally, Anish used Kapwing to assemble multiple short video clips generated with Veo 3, refining prompts in GPT-4o to achieve the desired 90s grunge aesthetic.
The final result was a surprisingly high-quality, evocative music video.
Workflow 2: Building a Book Catalog App with Gemini Flash
Anish's second workflow showcased the power of Gemini Flash for building a practical application: a book cataloging app.
Step 1: App Creation in Google AI Studio
Within Google AI Studio, Anish created an app using the prompt: "Create an app that takes a video of a person flipping through their book collection and extracts the author and title of every book shown."
Step 2: Video Analysis and Data Extraction
The app processes a video, extracting frames showing individual books. Gemini Flash's vision capabilities then identify and extract author and title information from each frame.
Step 3: App Deployment
The app was deployed using Cloud Run, making it accessible via a shareable link.
Workflow 3: Personal Finance Analysis with Comet
Anish highlighted Comet, a browser extension from Perplexity, demonstrating its use for personal finance insights.
Browse with Comet
Comet's AI browsing capabilities allow it to interact with websites, which can unlock all sorts of research and personal productivity workflows that might have been harder to execute before.
Anish uses Comet to access his Robinhood account and request analysis of his portfolio performance and investment recommendations. This is a workflow he could have done manually, but he gets insights faster and deeper letting the agent do a little bit of work.
Conclusion
Anish's demos were a great way to explore AI's potential beyond enterprise applications and vibe coding. I hope it inspires some of you to leverage AI for creative projects and some personal practical tasks.
As someone who loves the intersection of art and technology, this was a really fun episode for me.
Thank you to our sponsors!
This episode is brought to you by:
- Notion: The best AI tools for work.
- Lenny’s List on Maven: Hands-on AI education curated by Lenny and Claire.