2026-03-24 04:55:40 +00:00
1 changed files with 408 additions and 0 deletions
--- a/VEO_VIDEO_REPORT.md
+++ b/VEO_VIDEO_REPORT.md
@@ -0,0 +1,408 @@
+# 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 2025–2026, 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 (720p–1080p default) | Up to 1080p |
+| Clip Duration | 4–8 seconds per generation | 4–8 seconds per generation |
+| Aspect Ratio | 16:9 or 9:16 (portrait) | 16:9 or 9:16 |
+| Frame Rate | 24–30 fps | 24–30 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)
+
+```bash
+pip install google-genai
+export GOOGLE_API_KEY=your_key_here
+```
+
+```python
+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 1–3 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)
+
+```bash
+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 2–3 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 |
+| ~1–3 min generation per clip | Budget 20–30 minutes for full 8-clip sequence |
+| No guarantee of exact scene replication | Generate 2–4 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 2–3 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 2–3 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 | ~$10–92/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:
+
+```python
+#!/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)
+
+---
+
+## 9. Recommended Next Steps
+
+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 ($3–4 total using Fast model)
+3. **Refine prompts** — Iterate on 2–3 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
+
+- Google DeepMind Veo: https://deepmind.google/models/veo/
+- Veo 3 on Gemini API Docs: https://ai.google.dev/gemini-api/docs/video
+- Veo 3.1 on Vertex AI Docs: https://cloud.google.com/vertex-ai/generative-ai/docs/models/veo/
+- Vertex AI Pricing: https://cloud.google.com/vertex-ai/generative-ai/pricing
+- Google Labs Flow: https://labs.google/fx/tools/flow
+- Veo Prompting Guide: https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1
+- Case study (90% cost reduction): https://business.google.com/uk/think/ai-excellence/veo-3-uk-case-study-ai-video/