Q-Star AI already operates multiple AI generation workflows across video production, app generation, agent orchestration, and automated media production. We are preparing to consolidate these workloads on Google Cloud.
Q-Star AI is building a multi-agent AI creation infrastructure for video, apps, and automated creative workflows. AI is not an add-on feature โ it is the core execution layer of every product we ship.
We are currently in active discussions with institutional investors and startup ecosystem partners. We are preparing the Google Cloud Startup Credits application package in parallel so that the infrastructure plan, AI workload forecast, and evidence materials are ready once the investment or partner channel is confirmed.
These are not planned features. They are running in production today.
End-to-end AI video production: brief โ script โ TTS โ render โ subtitle burn โ CDN delivery. Running in production with 435+ MB of generated video assets across 22+ completed productions.
Codex-led multi-agent collaboration system using OpenClaw, Hermes, Claude, Q-Star agents, shared queues, R2/Qiniu handoffs, and independent verification. Runtime evidence for OpenClaw, Hermes, and Claude is on server 113.
Natural language to production-ready iOS + Android apps. APK packaging and store asset generation included. 7 APK artifacts in storage (136 MB total).
AI Music live on WeChat Mini-Program. ClipAI iOS + Android app in App Store review. Consumer-facing interfaces to the same infrastructure.
Our current infrastructure handles early-stage workloads. Google Cloud is the target for production-scale AI generation.
Gemini is planned as the core model in our routing layer. Vertex AI provides the managed infrastructure for planning, routing, text, code, and multimodal generation at scale โ without managing model serving ourselves.
Video rendering, app generation, and agent execution are compute-intensive. Cloud Run and GKE provide the elastic scaling we need to handle burst workloads without over-provisioning.
Cloud Storage replaces fragmented R2 + Qiniu setup with a unified, globally accessible artifact store โ generated videos, APKs, images, audio, and prompt archives in one place.
Cloud Monitoring and BigQuery give us the reliability dashboards, cost forecasting, and workload analytics we need to operate at scale and demonstrate AI/ML usage to stakeholders.
AI/ML workloads are expected to represent more than 50% of Q-Star AI's cloud usage.
| Workload | AI/ML Relevance | Google Cloud Target |
|---|---|---|
| Agent planning & orchestration | LLM task decomposition, routing, tool selection | Vertex AI / Gemini API |
| Video script generation | LLM narrative writing, scene planning | Vertex AI / Gemini API |
| TTS voice synthesis | Neural TTS, 6 voice profiles | Cloud Run, GPU workers |
| Video rendering & assembly | ffmpeg pipeline, subtitle burn, BGM mix | GKE, GPU-enabled compute |
| App UI/code generation | LLM code generation, screenshot-to-code | Gemini, Cloud Run |
| APK build & packaging | Automated build pipeline, asset generation | Cloud Run, Cloud Storage |
| Model routing & failover | Multi-model orchestration, cost optimization | Vertex AI, Monitoring |
| Usage analytics | Workload forecasting, cost reporting | BigQuery |
Gemini is planned as the core model in our routing layer. Vertex AI is the target for managed, scalable AI inference.
Gemini handles top-level agent planning, task decomposition, and multi-step reasoning across all four products.
Script writing for Drama Engine, UI/code generation for AppForge, and structured output generation for AgentOS.
Image understanding, screenshot-to-code conversion, and multimodal prompt processing for app and video generation.
Vertex AI as primary inference target. External model routing for non-Google models (Claude, GPT, MiMo) where specialized capabilities are needed.
Vertex AI usage metrics fed into BigQuery for workload forecasting, cost attribution, and AI/ML spend reporting.
Claude via Vertex AI Model Garden / partner model credits, where available, as part of the $10K partner model credits included in the AI Tier.
Projections are based on current internal workflows, deployed product demos, generated media artifacts, APK build artifacts, and planned early-access rollout.
| Metric | Current Baseline | 6-Month | 12-Month | Evidence Basis |
|---|---|---|---|---|
| Daily AI generations | ~300 | 20,000 | 80,000 | Workflow logs, generated assets in R2 |
| Monthly token usage | ~60M | 2B | 8B | Model routing logs, current API usage |
| Video rendering minutes/mo | 1,200 | 80,000 | 300,000 | 22+ render jobs, MP4 outputs in R2 |
| Async jobs/day | 1,000 | 120,000 | 500,000 | Queue logs, render plans, task traces |
| Generated storage | ~0.5 TB | 12 TB | 50 TB | R2 assets: 435MB videos + 136MB APKs |
| GPU inference hours/mo | 120 | 3,000 | 10,000 | TTS, video render, media processing jobs |
Phased migration from current infrastructure to Google Cloud, starting from credits activation.
| Phase | Timeline | Scope |
|---|---|---|
| Phase 1 | Before activation | Prepare GCP billing account, project structure, IAM, and budget alerts using company domain email |
| Phase 2 | Month 1 | Integrate Vertex AI / Gemini API for agent planning and model routing workloads |
| Phase 3 | Month 1โ2 | Deploy API services and agent worker containers on Cloud Run; migrate lightweight orchestration |
| Phase 4 | Month 2โ3 | Move generated media (videos, APKs, images, audio) and prompt archives to Cloud Storage |
| Phase 5 | Month 3โ4 | Introduce Pub/Sub for async task dispatch, render queues, retries, and workload fan-out |
| Phase 6 | Month 4โ6 | Add BigQuery analytics dashboards and Cloud Monitoring for reliability, cost, and workload reporting |
| Phase 7 | Month 6+ | Scale containerized generation workers on GKE with GPU-enabled compute for video and media processing |
Real deployed systems, real generated outputs, real infrastructure. Full inventory on the Evidence page.
Q-Star AI platform overview โ products, infrastructure, and Google Cloud migration plan.
Watch Video โAll 5 demo videos are rendered and public. V4 AppForge and V5 Google Cloud Migration are now complete.
7 QClip APK builds (136 MB total) proving real app generation and packaging pipeline.
View Evidence โdrama.q-star.ink โ Drama Engine live demo. AI Music live on WeChat Mini-Program.
Drama Engine โInvestor brief, architecture doc, usage projection, and evidence pack โ full application package.
View Documents โComplete inventory of all evidence assets in Cloudflare R2, with public status and Google Cloud relevance.
View Inventory โGoogle Cloud AI Tier application readiness status.
Full evidence package, architecture documentation, and usage projections are available. Contact us for the complete application materials.