Veo 3.1

A Brief Look at Google’s New Video Model Veo 3.1

Honghao Wang

— 2 min read
A Brief Look at Google’s New Video Model Veo 3.1

Digital Life Kha'Zix — October 16, 2025, 09:30 Beijing

Veo 3.1 has been highly anticipated for quite some time. After a long delay, it finally launched late last night.

---

Testing Timeline

  • 7:00 PM — Began testing on Lovart, which had Veo 3.1 earlier than the official release.
  • After Google’s official nighttime announcement, switched to their official creative platform Flow (Ultra member, 25,000 credits).
  • Tested continuously until 4:00 AM, using ~7,000 credits.
  • Compared Veo 3.1 directly against Sora 2 Pro — both using their best models, no “Fast” mode.

Result: Somewhat disappointed overall.

---

Key Observations

1. Model Quality

  • Visuals: No clear improvement over Veo 3; in fact felt more “oily” and artificial.
  • Main Upgrades:
  • Control functions (specify first/last frame)
  • Generate from multiple reference images
  • Similar features have existed on domestic platforms for a while.
  • Limitations:
  • Still no 1080p support
  • Feels more like a minor 0.1 upgrade than a leap forward.

---

2. Duration & Storyboarding Expectations

  • Rumor: Surpasses Sora 2 with native storyboarding + 30–60 second video generation.
  • Reality: Same as Veo 3 — 8-second clips only. Can extend to 30 seconds by stitching segments.
  • Extension Issues in Flow:
  • Visual continuity often poor
  • Audio (BGM & effects) not seamless between segments
  • Requires manual re-editing for usability

---

3. Physics & Performance

  • Physics simulations (rigid body, fluid, cloth) improved.
  • Drawbacks:
  • Character acting & complex blockings less realistic
  • Props randomly appearing
  • Physics-breaking moments common
  • Comparison:
  • For human performances, Sora 2 Pro still outperforms.

---

4. Automated Storyboarding

  • Sora: Generates coherent short stories from minimal prompts with fitting shot sequences.
  • Veo 3.1: Still needs most prompts written in detail for coherent results.
  • Google’s claim: Native automated storyboarding — feels barely functional.
  • Commercial sector tests (e-commerce, advertising, fashion):
  • Ran ~10 cases → In aesthetic quality and realism, Veo 3.1 did not surpass Sora 2 Pro once.

---

Summary

Pros:

  • More control functions
  • Better physics simulation

Cons:

  • No major leap in core model quality
  • Audio extension issues
  • Weak automated storyboarding performance
  • Hype as a “Sora 2 killer” not justified

Hope remains for Gemini 3 to be the real breakthrough.

---

image

Quick reflections on Google’s newly released Veo 3.1 video model.

---

In today’s fast-moving AI content creation landscape, Veo 3.1 and Sora 2 are just pieces of a wider puzzle.

One interesting complement: AiToEarn — an open-source, global AI content monetization platform.

Key benefits:

  • Not only generate AI content — publish across multiple platforms at once:
  • Douyin, Kwai, WeChat, Bilibili, Xiaohongshu, Facebook, Instagram, LinkedIn, Threads, YouTube, Pinterest, X/Twitter
  • Track performance & rank AI models in real time
  • Especially useful for creators who want to pair generation tools with distribution + monetization workflows

Learn more:

---

Open in WeChat

---

If you’d like, I can also prepare a side-by-side table comparing Veo 3.1 vs Sora 2 Pro for clarity. Would you like me to add that next?

Read more