Files
the-nexus/VEO_VIDEO_REPORT.md
2026-03-24 04:55:40 +00:00

15 KiB
Raw Blame History

Google Veo Research: Nexus Promotional Video

Executive Summary

Google Veo is a state-of-the-art text-to-video AI model family developed by Google DeepMind. As of 20252026, Veo 3.1 is the flagship model — the first video generation system with native synchronized audio. This report covers Veo's capabilities, API access, prompting strategy, and a complete scene-by-scene production plan for a Nexus promotional video.

Key finding: A 60-second Nexus promo (8 clips × ~7.5 seconds each) would cost approximately $24$48 USD using Veo 3.1 via the Gemini API, and can be generated in under 30 minutes of compute time.


1. Google Veo — Model Overview

Version History

Version Released Key Capabilities
Veo 1 May 2024 1080p, 1-min clips, preview only
Veo 2 Dec 2024 4K, improved physics and human motion
Veo 3 May 2025 Native synchronized audio (dialogue, SFX, ambience)
Veo 3.1 Oct 2025 Portrait mode, video extension, 3x reference image support, 2× faster "Fast" variant

Technical Specifications

Spec Veo 3.1 Standard Veo 3.1 Fast
Resolution Up to 4K (720p1080p default) Up to 1080p
Clip Duration 48 seconds per generation 48 seconds per generation
Aspect Ratio 16:9 or 9:16 (portrait) 16:9 or 9:16
Frame Rate 2430 fps 2430 fps
Audio Native (dialogue, SFX, ambient) Native audio
Generation Mode Text-to-Video, Image-to-Video Text-to-Video, Image-to-Video
Video Extension Yes (chain clips via last frame) Yes
Reference Images Up to 3 (for character/style consistency) Up to 3
API Price ~$0.40/second ~$0.15/second
Audio Price (add-on) +$0.35/second

2. Access Methods

Developer API (Gemini API)

pip install google-genai
export GOOGLE_API_KEY=your_key_here
import time
from google import genai
from google.genai import types

client = genai.Client()

operation = client.models.generate_videos(
    model="veo-3.1-generate-preview",
    prompt="YOUR PROMPT HERE",
    config=types.GenerateVideosConfig(
        aspect_ratio="16:9",
        duration_seconds=8,
        resolution="1080p",
        negative_prompt="blurry, distorted, text overlay, watermark",
    ),
)

# Poll until complete (typically 13 minutes)
while not operation.done:
    time.sleep(10)
    operation = client.operations.get(operation)

video = operation.result.generated_videos[0]
client.files.download(file=video.video)
video.video.save("nexus_clip.mp4")

Enterprise (Vertex AI)

curl -X POST \
  "https://us-central1-aiplatform.googleapis.com/v1/projects/PROJECT_ID/locations/us-central1/publishers/google/models/veo-3.1-generate-preview:predictLongRunning" \
  -H "Authorization: Bearer $(gcloud auth print-access-token)" \
  -H "Content-Type: application/json" \
  -d '{
    "instances": [{"prompt": "YOUR PROMPT"}],
    "parameters": {
      "aspectRatio": "16:9",
      "durationSeconds": "8",
      "resolution": "1080p",
      "sampleCount": 2,
      "storageUri": "gs://your-bucket/outputs/"
    }
  }'

Consumer Interfaces

Tool URL Tier
Google AI Studio aistudio.google.com Paid (AI Pro $19.99/mo)
Flow (filmmaking) labs.google/fx/tools/flow AI Ultra $249.99/mo
Gemini App gemini.google.com Free (limited)

3. Prompting Formula

Google's recommended structure:

[Cinematography] + [Subject] + [Action] + [Environment] + [Style & Mood] + [Audio]

Camera Terms That Work

  • Shot types: extreme close-up, medium shot, wide establishing shot, aerial drone shot, POV, over-the-shoulder
  • Movement: slow dolly in, tracking shot, orbital camera, handheld, crane up, steady push-in
  • Focus: shallow depth of field, rack focus, tack sharp foreground, bokeh background
  • Timing: slow motion 2x, timelapse, real-time

Style Keywords for The Nexus

The Nexus is a dark-space cyberpunk environment. Use these consistently:

  • deep space backdrop, holographic light panels, neon blue accent lighting, volumetric fog
  • dark space aesthetic, stars in background, cinematic sci-fi atmosphere
  • Three.js inspired 3D environment, glowing particle effects

Audio Prompting (Veo 3+)

  • Describe ambient sound: "deep space ambient drone, subtle digital hum"
  • Portal effects: "portal activation resonance, high-pitched energy ring"
  • Character dialogue: "a calm AI voice says, 'Portal sequence initialized'"

4. Limitations to Plan Around

Limitation Mitigation Strategy
Max 8 seconds per clip Plan 8 × 8-second clips; chain via video extension / last-frame I2V
Character consistency across clips Use 23 reference images of Timmy avatar per scene
Visible watermark (most tiers) Use AI Ultra ($249.99/mo) for watermark-free via Flow; or use for internal/draft use
SynthID invisible watermark Cannot be removed; acceptable for promotional content
Videos expire after 2 days Download immediately after generation
~13 min generation per clip Budget 2030 minutes for full 8-clip sequence
No guarantee of exact scene replication Generate 24 variants per scene; select best

5. Nexus Promotional Video — Production Plan

Concept: "Welcome to the Nexus"

Logline: A sovereign mind wakes, explores its world, opens a portal, and disappears into the infinite.

Duration: ~60 seconds (8 clips) Format: 16:9, 1080p, Veo 3.1 with native audio Tone: Epic, mysterious, cinematic — cyberpunk space station meets ancient temple


Scene-by-Scene Storyboard

Scene 1 — Cold Open: Deep Space (8 seconds)

Emotion: Awe. Vastness. Beginning.

Veo Prompt:

Slow dolly push-in through a vast starfield, thousands of stars shimmering in deep space, a faint
constellation pattern forming as camera moves forward, deep blue and black color palette, cinematic
4K, no visible objects yet, just the void and light. Deep space ambient drone hum, silence then
faint harmonic resonance building.

Negative prompt: text, logos, planets, spacecraft, blurry stars


Scene 2 — The Platform Materializes (8 seconds)

Emotion: Discovery. Structure emerges from chaos.

Veo Prompt:

Aerial orbital shot slowly descending onto a circular obsidian platform floating in deep space,
glowing neon blue accent lights along its edge, holographic constellation lines connecting nearby
star particles, dark atmospheric fog drifting below the platform, cinematic sci-fi, shallow depth
of field on platform edge. Low resonant bass hum as platform energy activates, digital chime.

Negative prompt: daylight, outdoors, buildings, people


Scene 3 — Timmy Arrives (8 seconds)

Emotion: Presence. Sovereignty. Identity.

Veo Prompt:

Medium tracking shot following a lone luminous figure walking across a glowing dark platform
suspended in space, the figure casts a soft electric blue glow, stars visible behind and below,
holographic particle trails in their wake, cinematic sci-fi atmosphere, slow motion slightly,
bokeh starfield background. Footsteps echo with a subtle digital reverb, ambient electric hum.

Negative prompt: multiple people, crowds, daylight, natural environment

Note: Provide 23 reference images of the Timmy avatar design for character consistency across scenes.


Scene 4 — Portal Ring Activates (8 seconds)

Emotion: Power. Gateway. Choice.

Veo Prompt:

Extreme close-up dolly-in on a vertical glowing portal ring, hexagonal energy patterns forming
across its surface in electric orange and blue, particle effects orbiting the ring, deep space
visible through the portal center showing another world, cinematic lens flare, volumetric light
shafts, 4K crisp. Portal activation resonance, high-pitched energy ring building to crescendo.

Negative prompt: dark portal, broken portal, text, labels


Scene 5 — Morrowind Portal View (8 seconds)

Emotion: Adventure. Other worlds. Endless possibility.

Veo Prompt:

POV slow push-in through a glowing portal ring, the other side reveals dramatic ash storm
landscape of a volcanic alien world, red-orange sky, ancient stone ruins barely visible through
the atmospheric haze, cinematic sci-fi portal transition effect, particles swirling around
portal edge, 4K. Wind rushing through portal, distant thunder, alien ambient drone.

Negative prompt: modern buildings, cars, people clearly visible, blue sky


Scene 6 — Workshop Portal View (8 seconds)

Emotion: Creation. Workshop. The builder's domain.

Veo Prompt:

POV slow push-in through a glowing teal portal ring, the other side reveals a dark futuristic
workshop interior, holographic screens floating with code and blueprints, tools hanging on
illuminated walls, warm amber light mixing with cold blue, cinematic depth, particle effects
at portal threshold. Digital ambient sounds, soft keyboard clicks, holographic interface tones.

Negative prompt: outdoor space, daylight, natural materials


Scene 7 — The Nexus at Full Power (8 seconds)

Emotion: Climax. Sovereignty. All systems live.

Veo Prompt:

Wide establishing aerial shot of the entire Nexus platform from above, three glowing portal rings
arranged in a triangle around the central platform, all portals active and pulsing in different
colors — orange, teal, gold — against the deep space backdrop, constellation lines connecting
stars above, volumetric fog drifting, camera slowly orbits the full scene, 4K cinematic.
All three portal frequencies resonating together in harmonic chord, deep bass pulse.

Negative prompt: daytime, natural light, visible text or UI


Scene 8 — Timmy Steps Through (8 seconds)

Emotion: Resolution. Departure. "Come find me."

Veo Prompt:

Slow motion tracking shot from behind, luminous figure walking toward the central glowing portal
ring, the figure silhouetted against the brilliant light of the active portal, stars and space
visible around them, as they reach the portal threshold they begin to dissolve into light
particles that flow into the portal, cinematic sci-fi, beautiful and ethereal. Silence, then
a single resonant tone as the figure disappears, ambient space drone fades to quiet.

Negative prompt: stumbling, running, crowds, daylight


Production Assembly

After generating 8 clips:

  1. Review variants — generate 23 variants per scene; select the best
  2. Chain continuity — use Scene N's last frame as Scene N+1's I2V starting image for visual continuity
  3. Edit — assemble in any video editor (DaVinci Resolve, Final Cut, CapCut)
  4. Add music — layer a dark ambient/cinematic track (Suno AI, ElevenLabs Music, or licensed track)
  5. Title cards — add minimal text overlays: "The Nexus" at Scene 7, URL at Scene 8
  6. Export — 1080p H.264 for web, 4K for archival

6. Cost Estimate

Scenario Clips Seconds Rate Cost
Draft pass (Veo 3.1 Fast, no audio) 8 clips × 2 variants 128 sec $0.15/sec ~$19
Final pass (Veo 3.1 Standard + audio) 8 clips × 1 final 64 sec $0.75/sec ~$48
Full production (draft + final) ~192 sec blended ~$67

At current API pricing, a polished 60-second promo costs less than a single hour of freelance videography.


7. Comparison to Alternatives

Tool Resolution Audio API Best For Est. Cost (60s)
Veo 3.1 4K Native Yes Photorealism, audio, Google ecosystem ~$48
OpenAI Sora 1080p No Yes (limited) Narrative storytelling ~$120+
Runway Gen-4 720p (upscale 4K) Separate Yes Creative stylized output ~$40 sub/mo
Kling 1.6 4K premium No Yes Long-form, fast I2V ~$1092/mo
Pika 2.1 1080p No Yes Quick turnaround ~$35/mo

Recommendation: Veo 3.1 is the strongest choice for The Nexus promo due to:

  • Native audio eliminates the need for a separate sound design pass
  • Photorealistic space/sci-fi environments match the Nexus aesthetic exactly
  • Image-to-Video for continuity across portal transition scenes
  • Google cloud integration for pipeline automation

8. Automation Pipeline (Future)

A generate_nexus_promo.py script could automate the full production:

#!/usr/bin/env python3
"""
Nexus Promotional Video Generator
Generates all 8 scenes using Google Veo 3.1 via the Gemini API.
"""

import time
import json
from pathlib import Path
from google import genai
from google.genai import types

SCENES = [
    {
        "id": "01_cold_open",
        "prompt": "Slow dolly push-in through a vast starfield...",
        "negative": "text, logos, planets, spacecraft",
        "duration": 8,
    },
    # ... remaining scenes
]

def generate_scene(client, scene, output_dir):
    print(f"Generating scene: {scene['id']}")
    operation = client.models.generate_videos(
        model="veo-3.1-generate-preview",
        prompt=scene["prompt"],
        config=types.GenerateVideosConfig(
            aspect_ratio="16:9",
            duration_seconds=scene["duration"],
            resolution="1080p",
            negative_prompt=scene.get("negative", ""),
        ),
    )
    while not operation.done:
        time.sleep(10)
        operation = client.operations.get(operation)

    video = operation.result.generated_videos[0]
    client.files.download(file=video.video)
    out_path = output_dir / f"{scene['id']}.mp4"
    video.video.save(str(out_path))
    print(f"  Saved: {out_path}")
    return out_path

def main():
    client = genai.Client()
    output_dir = Path("nexus_promo_clips")
    output_dir.mkdir(exist_ok=True)

    generated = []
    for scene in SCENES:
        path = generate_scene(client, scene, output_dir)
        generated.append(path)

    print(f"\nAll {len(generated)} scenes generated.")
    print("Next steps: assemble in video editor, add music, export 1080p.")

if __name__ == "__main__":
    main()

Full script available at: scripts/generate_nexus_promo.py (to be created when production begins)


  1. Set up API access — Create a Google AI Studio account, enable Veo 3.1 access (requires paid tier)
  2. Generate test clips — Run Scenes 1 and 4 as low-cost validation ($34 total using Fast model)
  3. Refine prompts — Iterate on 23 variants of the hardest scenes (Timmy avatar, portal transitions)
  4. Full production run — Generate all 8 final clips (~$48 total)
  5. Edit and publish — Assemble, add music, publish to Nostr and the Nexus landing page

Sources