how to make unlimited video ads with Sora 2 (without the watermark or an invite code)

Key Concepts

• Sora 2: OpenAI's advanced text-to-video generation model, capable of creating highly realistic and complex video scenes from text prompts and reference images.
• Arc Ads AI: A platform that integrates with Sora 2 Pro, enabling users to generate AI videos for advertising purposes.
• UGC (User-Generated Content) Style Ads: Advertisements designed to mimic authentic content created by everyday users, often featuring influencers or relatable scenarios.
• Custom GPT (in ChatGPT Store): A specialized version of ChatGPT, like the "UGC Video Script Generator Mr. Page Social," tailored for specific tasks such as writing viral ad scripts.
• Reference Image: An initial image provided to Sora 2 to guide the visual style, product appearance, or scene setting of the generated video.
• Nano Banana / Google Gemini / Cadream: AI image generation tools used to create custom reference images or backgrounds for Sora 2.
• Cap Cut: A video editing software used for assembling, cutting, adding music, B-roll, and captions to AI-generated video clips.
• B-roll: Supplementary footage used to intercut with the main narrative or dialogue, often showing product shots or environmental scenes.
• Content Blocks: Restrictions or limitations within AI models that prevent the generation of certain types of content, often due to suggestive or sensitive themes.
• AI Avatar: The AI-generated character or person featured in the video.
• V3 (AI Video Model): An alternative AI video generation model (likely referring to Synthesys X or similar) mentioned for its ability to maintain consistent characters throughout longer videos, contrasting with Sora 2's current limitations.

---

Introduction to Sora 2 and AI Video Ads

The video introduces Sora 2 as a groundbreaking AI video generation tool capable of producing "insane looking AI videos for your ads" that are "the best AI generated video ads I've ever seen. Hands down." It highlights Sora 2's ability to create content ranging from full-blown South Park episodes to reimagined TV shows and movies. A key challenge until recently was the inability to export these videos without a watermark, which has now been addressed. The presenter emphasizes the ease of use, stating that Sora 2 "removes the barrier to creating AI videos and AI like scripting all this stuff like to zero."

Showcase of Sora 2 Capabilities with Specific Examples

The presenter demonstrates Sora 2's power through several ad examples, often using a single-sentence prompt and a reference image:

• Vantic V8 Apex Vacuum: A 3D, polished ad was generated using a Nano Banana image of the vacuum and the prompt "create an animated ad for this vacuum." The resulting video showcased features like "powerful cleaning in the palm of your hand," "dual HEPA filters trap 99.9% of dust," and "30,000 pascals of suction."
• Gruin Superfoods Greens Gummies: An influencer-style ad was created from a product screenshot and the prompt "influencer talks about how much she loves the gummies while in the gym." Sora 2 automatically scraped product information (e.g., "20+ organic vitamins and minerals") from the screenshot and wrote the script, demonstrating its ability to infer product details.
• iPhone 16 Pro (mistaken for 17): An "Apple style ad" was generated using a screenshot of the iPhone. The ad featured details like "machine from Titanium," "strong, light, and now in three new finishes," and a "new 48 megapixel main camera with the most advanced optics we've ever built," powered by a "Crazy fast A18 Pro" chip. The presenter noted that generating multiple variations could provide enough clips for a cohesive ad.
• Fujifilm XH2S Camera: A "fast-paced cinematic high-quality commercial" was created from a starting frame of the camera body. The ad highlighted features like "built for speed, for clarity, for the shot that lasts a lifetime," a "stack 26 megapixel X-Trans CMOS 5 sensor," and "40 frames per second blackout, free 6K" video.
• E.L.F. Waterproof Liquid Liner: A POV (Point of View) video of an influencer applying eyeliner was generated with the prompt "POV video of an influencer talking about E.L.F. cosmetics while putting on eyeliner." The ad mentioned the product's affordability ("six bucks") and performance ("no skipping, no crumble, and it's dry already").
• Breeze Soda Alternative: An influencer ad was created for a soda alternative, with the prompt "An influencer talking about how much they love their new soda alternative uh alternative pronounced Breeze." The prompt specifically included the pronunciation to ensure correct dialogue, and Sora 2 again pulled product details (e.g., "lemon elderflower," "micro do mushrooms") without explicit input.

Step-by-Step Process for Generating AI Video Ads

The presenter outlines a clear methodology for creating AI video ads:

1. Script Generation (ChatGPT Store):

   • Access the ChatGPT Store and search for the custom GPT "UGC Video Script Generator Mr. Page Social."
   • Click "help me write a viral script" and answer questions about the product (e.g., Liquid Death), target audience (e.g., "people looking to look cooler while drinking water," "enjoy concerts, bars"), dream outcome (e.g., "alcohol-free alternative that doesn't taste bad, looks cool"), pain points, and customer reviews.
   • The GPT generates a 30-second UGC selfie script with quotes and actor actions.

2. Video Generation (Arc Ads AI with Sora 2 Pro):

   • Sign up and log in to arcads.ai.
   • Create a new project and navigate to the "describe a video" section.
   • Select "Sora 2 Pro" from the video generation options.
   • Copy sections of the generated script from ChatGPT.
   • Paste the script section into Arc Ads, adding context like "influencer says..." and "while holding the can."
   • Include a reference image (e.g., a Liquid Death can from Amazon listing or a custom-generated image).
   • Click "generate the video." This process is repeated scene-by-scene, with each video potentially featuring a different UGC actor due to Sora 2's 12-second character consistency limitation.

Advanced Tips and Techniques

The presenter shares crucial insights for optimizing Sora 2 video generation:

• Controlling Background Music: To ensure consistency across multiple clips, explicitly add "no background music" to the prompt when generating videos, as background music added by Sora 2 can be difficult to remove or match.
• Improving Reference Images:
  • Avoid using generic product images (e.g., from Amazon listings) as they can result in "fake" or "unhinged" looking products with unnatural lighting.
  • Use actual images or video frames of people holding the product for more realistic lighting and integration.
  • Utilize AI image generators like Nano Banana, Google Gemini, or Cadream to create custom reference images of the product in desired settings (e.g., "on a kitchen counter," "on a bar top at a concert," "in the fridge next to a bunch of beer").
  • To match the vibe of an existing scene, take a screenshot, use an AI image tool to "remove the actor from this scene" to get just the background, and then use that background as a starting frame for Sora 2.
• Generating B-roll: Sora 2 can generate B-roll footage by providing a reference image (e.g., a Liquid Death can in a fridge) and a prompt like "Slow pan towards the liquid death can. No dialogue or music."

Editing and Post-Production in Cap Cut

The presenter demonstrates how to assemble and refine the generated clips into a cohesive ad using Cap Cut:

• Timeline Assembly: Arrange clips in the timeline, making cuts to define start and end points.
• Multiple Takes: Generate multiple variations of each script line with different actors to allow for swapping clips when something "doesn't look or feel right" or to hide AI glitches (e.g., a floating can) with quick cuts.
• Background Music: Add a consistent background music track (e.g., "metal music" for Liquid Death) over the entire ad to enhance cohesion.
• B-roll Integration: Use generated B-roll as an ending scene, potentially with text overlays (e.g., "water but make it scary") or a voiceover.
• Captions: Easily generate captions in Cap Cut based on the entire script and choose from various styles to make them "pop."

Case Study: Recreating a Liquid Death Ad

The presenter undertakes a challenge to recreate a viral Liquid Death ad ("Small Ones" campaign) entirely with AI:

• Original Ad Analysis: The original ad features multiple people humorously discussing their preference for "small ones" (referring to Liquid Death's smaller cans).
• Recreation Process:
  • The script of the original ad was broken into 12-second sections.
  • For each section, a prompt was written for Arc Ads/Sora 2 (e.g., "woman holding her phone sitting on her bed saying jokingly, I like small ones, not like micro ones... she changes the distance between her thumb and her finger to demonstrate the size difference. No background audio.").
  • Reference images were used, sometimes generated by removing the original actor from a screenshot of the scene to capture the background vibe.
• Result: The AI-recreated ad, while having some "AI tells" and requiring minor edits (e.g., cropping out a tiny phone), successfully captured the essence and narrative of the original with "very little effort." A "happy acoustic background" was added to match the original ad's vibe.
• Limitations Encountered: Sora 2's "content blocks" sometimes required regenerating videos due to the suggestive nature of the script. Issues like blurry elements or inconsistent object sizes (e.g., a tiny phone) were fixed with cropping or by regenerating the prompt.

Sora 2 Limitations and Trade-offs

The presenter discusses the practical limitations and trade-offs when using Sora 2 compared to other AI video models like V3:

• Character Consistency: Sora 2 is limited to approximately 12 seconds of consistent character appearance per generation. Unlike V3, which allows for consistent AI avatars throughout an entire ad, Sora 2 requires generating multiple clips with different "AI avatars" for longer narratives.
• Reference Image Usage: Sora 2 does not allow reference images with a person's face, necessitating the use of product-only or background-only reference images.
• Avatar Manipulation: Users have limited control over manipulating the appearance of the AI avatar generated by Sora 2 due to policy restrictions.
• Video Cut-offs: Generated Sora 2 videos with dialogue often "cut off" abruptly at the end, requiring manual editing to create smooth transitions.
• Trade-off: While V3 offers better character consistency for narrative ads, Sora 2's AI avatars "look and sound better," making it suitable for UGC-style ads with multiple actors or for recreating narrative videos by piecing together different clips.

Conclusion: The Future of AI Ads

The presenter concludes by emphasizing the rapid advancement of AI video generation: "AI ads are getting scary good." The technology has progressed beyond a "cute little thing" to something "real" and impactful. The ease and speed of creating high-quality, engaging video ads with tools like Sora 2 and Arc Ads AI, even for complex narratives or UGC styles, are unprecedented. The cost for the demonstrated Liquid Death ad was estimated at around $50 in credits, taking only 30 minutes to make. The presenter encourages viewers to experiment with the technology, highlighting its accessibility and potential for transforming advertising.