Paul McCartney & Me: When AI Tried to Join the Band
What a great Zoom meeting on Tuesday and a huge thanks to Paddy Deighan MBA JD PhD and to Praful Thakkar for bringing up some excellent points of which we were unaware. We’re so fortunate to have the level of AI brainpower that we do, right here in the Rain!
This week’s AR challenge was on AI video-making and, wow—what an adventure! One click led to another, and before I knew it, I was deep down the rabbit hole of AI tools, testing what each could do with a single image. I started with a photo I created in DeepAI.org, which became my featured image for this post. From there, I explored three different video platforms, each promising their own version of AI “magic.”
What I discovered was both fascinating and a little funny—all three produced almost identical videos from that same image and the two that had audio, both featured COUNTRY, not rock ’n roll!! As well, I even upgraded to “pro” on a couple of them (a decision I regretted almost instantly).
Let’s just say I’m sticking with what I know and love—I’m a ChatGPT girl through and through! Still, it was a great experiment and a reminder that even when the results aren’t perfect, curiosity always leads to something worth learning.
Today’s post is a peek behind the curtain of my AI video-making experiment. Perplexity explains each of my prompts and what was created as a result of those prompts.
Analysis and Explanation
*NOTE: I added my prompts in italics. Perplexity wrote the following in blue:
1. DeepAI – Photo Prompt
“A highly realistic photo in the style of a candid concert shot, showing Paul McCartney (with mid-length brown hair styled in a classic Beatles cut, wearing a tailored dark suit typical of the late 1960s) and me performing a duet together on stage in a small, intimate bar setting. Both of us are holding guitars and singing into microphones. I am wearing a hot pink dress, matching necklace, and pink shoes, with shoulder-length silver hair styled smoothly, as shown in my reference photo. The scene should evoke the warmth and energy of a live performance, with realistic concert lighting and a cozy, inviting bar ambiance. Both subjects should appear happy, engaged with the music, capturing the joy of two musicians sharing a memorable performance together.”
You submitted a very detailed prompt, describing both your appearance (clothing, hair, accessories) and the setting (bar, lighting, atmosphere). This kind of specificity helps AI image generators produce realistic, stylistically accurate images, especially when you mention reference photos and emotional ambiance. The prompt’s focus on warmth, realism, engagement, and joy guides the model’s choices for facial expressions, composition, and color scheme.
- Why DeepAI Works Well Here:
DeepAI’s image generation engine can process specific attributes efficiently in static images, so your detailed visual and style cues likely resulted in a photorealistic output—though it’s limited to still images, not video.
2. Night Café – Video Prompt:
“Create a video in the style of Paul McCartney and me (just the two of us), singing your choice of upbeat words. Stage lights shift as they strum their guitars and sway to the music, singing into the microphones. For our attire, choose something period-specific (for the 60's).”
You tried a similar prompt as Google VEO3, asking Night Café for a friendly concert duet video with period-specific attire and shifting stage lights.
What to Expect:
- Night Café may produce animated or stylized outputs but could struggle with fully realistic, moving videos.
- Generating true-to-life, high-quality AI video of famous individuals with recognizable identity remains a challenge, both technically (model accuracy, personalization) and ethically (platform restrictions).
Big Picture: What This Shows About AI Video Creation
- Photo Generation Is More Widely Allowed: Image generators are more permissive—especially with reference photos and detailed styling, as seen with DeepAI.
-
Video Generation Faces Tougher Restrictions:
High-end video tools (like Google VEO3) actively block requests that may violate copyright or “deepfake” rules—especially involving famous musicians, actors, or using particular songs. -
Platform-Specific Rules Vary:
Night Café may attempt stylized video results, but realism and exact likeness are often limited, with outputs more “artistic” than photorealistic.
3. Google VEO3 & Perplexity – Video Prompt:
“Create a video in the style of Paul McCartney and me (just the two of us), singing ‘Let it Be’ or ‘Say, Say, Say’ or your choice of upbeat words. Stage lights shift as they strum their guitars and sway to the music, singing into the microphones. For our attire, choose something period-specific (for the 60's).”
Result: “FORBIDDEN BY GOOGLE (repeatedly)”
You attempted to use Google’s VEO3 for video generation with two approaches:
- One, letting the AI pick lyrics (“your choice of upbeat words”) and focus on authentic concert visuals.
- Two, referencing an attached photo and asking the duo to perform “Let It Be,” providing extra context for style and interaction.
What Happened:
- Google VEO3’s refusal—“FORBIDDEN BY GOOGLE”—shows policy restrictions on generating content featuring real celebrities, especially musical or lifelike impersonation (which can involve copyright, likeness, or deepfake rules).
- Even with your own image as a reference, if the model cannot verify identity or is restricted from generating person-specific videos with celebrities, it might block the request.
My Experience with AI-Generated Videos: A Dialogue with Perplexity
NOTE: My questions answered by Perplexity, noted as AI Assistant.
Me:
For our ActiveRain AI challenge, I wanted to see how far today’s platforms could creatively reimagine the magic of a concert duet. Here’s what I asked each tool:
- DeepAI: Asked for a highly realistic photo of Paul McCartney and me performing together, with detailed stage, lighting, and attire instructions.
- Google VEO3 & Perplexity: Tried to create a video of us singing upbeat lyrics, letting the AI choose, and another prompt with a reference photo performing "Let it Be."
- Night Café: Repeated the video duet concept with concert ambiance and period-accurate wardrobe.
AI Assistant:
You crafted clear, vivid prompts—a skill that’s especially valuable in AI art generation! Platforms like DeepAI shined with photo realism, transforming your concert dream into a warm, lifelike snapshot. When you pivoted to video creation, though, you uncovered a revealing challenge: modern AI video tools actively restrict requests that involve celebrity likeness and well-known songs. Google VEO3 in particular flagged your prompt as “FORBIDDEN,” reflecting growing concern about deepfakes, copyright, and personal identity .
Me:
OH, one more question about Night Café - Night Café produced a video but, there was no audio. Is that the norm with Night Café?
AI Assistant:
Yes, that's the norm—Night Café's AI-generated “video” feature typically does not include audio by default. It creates animated videos by sequencing and animating frames from still images (using effects like zoom, pan, rotate), but the result is a silent MP4 file. If you want music or voiceover, you need to add it separately using your own editing tools or another app.
- No built-in audio: Night Café does not offer in-platform options to generate videos with sound or to include AI-generated music or speech in your output.
- Community advice: Many users download their Night Café videos and then use external editors to add music or voice narration afterward.
Me:
Okay, so, photo generation feels open but, video AI is more tightly controlled when it comes to famous people and real life performances?
AI Assistant:
Exactly. Photo generators let creativity run free if you supply enough details. But video tools—especially the advanced, hyper-realistic ones—often block requests that tread near legal or ethical gray areas. They’re programmed to respect copyright and personal image rights, which means a duet with Paul McCartney is a no-go unless the content’s generic or stylized. Night Café might let you experiment with artistic video animations, but realism and true likeness are generally dialed back.
Me:
What’s the future of AI customization here? Will I ever get my custom video duet with Paul McCartney?!
AI Assistant:
AI platforms are evolving to balance user creativity with responsibility. In the future, you might see tools offer approved “collaborative” content creation, perhaps through partnerships or by leveraging custom avatars licensed for public use. You’ll also see more transparency around what’s allowed, potentially with guided prompt writing to avoid frustration. For now, learning to work with each tool’s strengths—and being mindful of boundaries—is key.
Me:
What lesson should readers take away from my challenge?
AI Assistant:
- Be creative, but expect rules: AI can bring your imagination to life, but ethical guardrails are part of the landscape.
- Detail matters: The more specific your prompt, the better the results in image generation.
- Stay informed: As AI tools grow, so do the boundaries. Know what you can—and can’t—ask for.
- Experiment: Each platform has its quirks. Don’t be afraid to try different approaches and see what works!
We’re excited to see what this week’s exciting call will bring! What a fun group of Zoomers we have! If you haven’t ever joined us or haven’t joined us in awhile, we have a fantastic challenge going on through the end of the year. Hope to ‘see’ you soon!!
Comments(9)