← All episodes

I tested Seedance 2.0. Wow.

| 7 products mentioned
Sirio guest
Watch on YouTube ai video generation video editing automation ai influencers content production ad localization avatar creation prompt engineering

Sirio demonstrates why Sora 2.0 is the superior AI video generation and editing tool, showcasing practical use cases like virtual try-ons, international ad localization, character replacement, and AI influencer creation that builders can monetize immediately. The episode focuses on how to think about multi-input video generation—combining images, videos, and audio in a single prompt—to solve real production problems faster and cheaper than traditional workflows, with detailed prompts and tactics included.

Key takeaways
  • Sora 2.0's multi-input capability (combining up to 2 images, 2 videos, and audio files in one generation) functions as both a video generator and editor, enabling use cases like background/character replacement that previously required expensive post-production work.
  • Use Claude 3.5 Opus for prompt optimization over GPT—it better understands vision model requirements; however, Sora 2.0 requires highly specific, detailed prompts (unlike Cling 3, which favors simplicity) to achieve high-quality output with consistent character identity and motion.
  • For AI influencer and avatar creation, control emotions through muscle movement descriptions rather than abstract emotional labels (e.g., describe "eyebrow raise with lip corner twitch" instead of "sad") to generate realistic lip-syncing and talking-head videos that don't look AI-generated.
  • Video extension—stretching 3-second clips to 15+ seconds while maintaining narrative continuity—is now possible, solving a major pain point for ads and filmmaking where you need additional footage without reshooting.
  • International ad localization can be done in under 60 seconds by replacing actors and translating voiceover in the same generation pass, enabling A/B testing across demographics at a fraction of traditional production costs.
  • Start with strong source reference images—Sora 2.0 uses them to understand your aesthetic taste and produce more concrete, tangible outputs; reference images are as critical to AI video as they are to traditional painting or design.

Recommendations (3)

SeaDance uses

"Seedance 2.0 officially launched today. It's probably the first AI model that allows for multi-input generation. The quality of Seedance 2 based on my testing so far is unmatched."

Sirio · ▶ 2:21

Claude
Claude uses

"You can definitely use an LLM to optimize. I think that Claude does a phenomenal job, especially the 4.6 version. Claude does understand prompt engineering for vision models a bit better."

Sirio · ▶ 6:37

GPT
GPT uses

"I've used GPT before, but I do think that Claude does understand prompt engineering for vision models a bit better."

Sirio · ▶ 6:45

Mentioned (4)

Midjourney
Midjourney "It's almost like Midjourney Pro whereby the use cases are unlimited." ▶ 4:54
Kling "You can do something very similar with Kling 3. Kling 3, if you're simple, straightforward, you'r..." ▶ 5:16
Google Veo "We could not do this before. Google V3.1 kind of tried, but look at this." ▶ 19:05
Adobe
Adobe "I still believe that Adobe is relevant especially for creative professionals who want way more co..." ▶ 30:00