La nueva forma de crear anuncios (que no apestan) con IA

Summary of technological concepts & tutorial/guide takeaways

• New workflow for AI ads (cloning proven ads)

  • The creator proposes generating marketing videos by cloning existing “proven” ads/trailers and replacing the protagonist with the viewer’s face/brand/product/voice.
  • The key idea: take an existing video reference, analyze it, then recreate it with changes while keeping the same structure and timing.

• Hixfield launch: “video analysis”

  • The standout feature is Hixfield’s Video Analysis, which goes beyond simply watching:
    • Extracts video concepts and subtitles
    • Identifies shot durations (timing/fidelity per shot)
    • Produces the basis for accurate recreation with fidelity
  • Previously, this required many separate tools; now the system unifies the models in one place.

• “Single interface” / multi-model integration via MCP

  • The video shows connecting Hixfield MSP to an orchestration tool (named Cloud Code, but it also works with similar tools like Codex/Hermes/OpenCloud).
  • Once connected, Hixfield can call multiple AI/video functions without manually opening tools like Capcat/Sedans/Premiere for the full pipeline.
  • Capabilities mentioned inside Hixfield include:
    • Video analyzer
    • Marketing video generator
    • “Soul character training” (subtitle term unclear)
    • Cinematic image-to-video
    • Clip generator
    • Virality prediction
    • Other video generation features

• How the system works (end-to-end pipeline)

  1. Connect: From the orchestration/coding interface, choose “Connect to Hixfield MSP” (via a secure approval flow).
  2. Provide reference video: Paste a YouTube link (no need to download).

  3. Prompt: Example instructions like:

     “Recreate this ad, keep shot durations, replace protagonist with me, change message to my product.”

  4. Video analysis runs: Determines shots, characters, timing, etc.

  5. Generation:
     • Creates images, then scenes/clips
     • Generates music/audio
     • Composes/renders the final ad video
  6. Optional post-step: Later, extract individual clips and edit externally (e.g., in Premiere). The system can also output a complete one-shot ad.

• Practical walkthrough: recreate a commercial

  • The creator demonstrates recreating a known anti-piracy ad structure (mom/daughter scenario).
  • They change the topic from pirated movies to pirated online courses, using the viewer’s own photos/character.
  • The output keeps character consistency and shot consistency (as claimed), including dialogue beats and pacing.

• Model comparison and pricing analysis (credits)

  • The creator benchmarks generation models (mainly Sedans / Sedans 2.0 vs Cling—spelling may vary).
  • Observed cost behavior:
    • Example claim: 15 seconds / 1080p costs significantly less on Cling than on Sedans 2.0.
    • They claim Cling can be ~5x cheaper at comparable resolution.
  • Approximate ad cost examples mentioned:
    • One ad generation described as ~$6.50 (based on narration: “130 divided by 20 credits = $6.50”).
    • Another run described as ~$1.50 for a 10-second case on the cheaper model.
  • Conclusion: choose the best model based on output needs (quality, cost, FPS settings), not just “cheapest.”

• Marketing strategy advice (volume testing)

  • The creator reframes ad creation as high-volume experimentation:
    • “Create 50 ads with different angles”
    • Upload to platforms (mentioning Meta campaigns)
    • Remove poor performers quickly and keep winners
  • This supports faster iteration and reduces dependence on expensive manual production.
  • They compare costs: traditional agency/production ads can be thousands of dollars, while AI generation is positioned as a coffee-price alternative.

• Resources promoted in the video

  • A free complete prompt is said to be provided in the video description to replicate the ad style.
  • They also promote an “Imperio” community as the place for workflows, playbooks, training, and achievements.

Main speakers / sources

• Main speaker: the YouTuber (referred to earlier as Ben in the ad example). The narrator appears to be the creator and later references “Benja” during the price breakdown.
• Product/source tools mentioned:
  • Hixfield MSP / Hixfield Video Analysis
  • Integration via Cloud Code (also mentions Codex, Hermes, OpenCloud)
• Platforms referenced:
  • Meta campaigns
  • YouTube (used as the reference-video source)