Cuts AI team maintenance time by roughly 50% with Cloud Build
Improves TTS API response time by 25% on Cloud Run
Increases voice cloning scaling speed by 800% on Cloud Run
Becomes the first in Vietnam to offer large-scale voice cloning on Cloud Run
Drives 20% monthly user growth with reduced service latency on Cloud Run
VBEE AI delivers human-like, localized AI voices for Vietnam, scaling Cloud Run to drive 20% monthly user growth.
VBEE AI delivers human-like, localized AI voices for Vietnam, scaling Cloud Run to drive 20% monthly user growth.
When overnight success creates a scaling bottleneck
Our voice cloning demand exploded overnight. Our static infrastructure couldn't scale for the surge, and it created a frustrating service experience for our customers.
Kỳ Nguyễn Hoàng
R&D Lead, VBEE AI
Global text-to-speech platforms often fail to capture the complex nuances of the Vietnamese language. VBEE AI was founded to solve this, offering content creators, educators, and media companies a library of high-quality, localized AI voices that carry natural emotion and distinct regional accents.
In early 2025, the company launched a large-scale voice cloning solution, which became an immediate hit. But VBEE AI's existing system, built on static virtual machines, could not scale to meet the surge. As high-volume training requests created a massive bottleneck, the engineering team was forced to manually provision new servers. Customers were left waiting for days to receive their personalized voice clone. Even standard text-to-speech requests were impacted, with latency stretching to several minutes.
To deliver instant, high-quality voice services, VBEE needed a scalable Graphic Processing Unit (GPU) solution that could handle intensive, on-demand AI workloads without manual intervention.
Delivering real-time voice cloning with a serverless AI pipeline
VBEE AI selected Google Cloud to solve its scaling challenge. Specifically, the serverless GPU capabilities of Cloud Run allowed the team to run intensive AI workloads on demand without managing servers. The platform also offers seamless integration with MLOps tools in line with VBEE's automation requirements.
To deliver its end-to-end voice cloning services, VBEE designed a new, sophisticated and flexible pipeline:
- When a user uploads or records voice samples through the VBEE portal, files are saved in Cloud Storage.
- A pre-processing job on Cloud Run automatically cleans and segments this data.
- This triggers a fine-tuning job on a Cloud Run GPU instance, training a new model on the user's unique vocal characteristics.
- The newly trained voice model is pushed to Artifact Registry.
- Cloud Build then automatically deploys this model to the production inference service.
- The user can then type text, and the inference service performs real-time text-to-speech, delivering audio in their new cloned voice.
This fully automated pipeline allows VBEE AI to handle the entire process, from initial upload to real-time inference, without manual intervention.
Google Cloud offered seamless integration with tools like Cloud Build, Artifact Registry, and Cloud Monitoring, providing the scalability we needed without the infrastructure management.
Kỳ Nguyễn Hoàng
R&D Lead, VBEE AI
To accelerate the implementation, VBEE AI worked with Google Cloud partner Cloud Ace. The partner helped design and validate the scalable, container-based architecture on Cloud Run (GPU). Cloudways also assisted in configuring the CI/CD pipelines, setting up monitoring dashboards in Cloud Monitoring, and providing critical billing consultation for Committed Use Discounts (CUDs).
Winning the market with 20% monthly user growth
With the shift to a serverless architecture, the voice cloning service's scaling speed increased by 800%, while the AI engineering team's maintenance time dropped by roughly 50%. Not constrained by infrastructure bottlenecks, the text-to-speech API response time improved by 25%. This means that users don't have to wait minutes, but can now hear their audio outputs almost in real time.
Halving the maintenance overhead freed the team to focus more time on research, experimenting with larger models to produce more natural-sounding speech. This superior product experience has allowed VBEE AI to become the first AI voice company in Vietnam to officially launch large-scale voice cloning – and is now driving a 20% monthly growth in its active user base.
Google Cloud helped us turn innovation into real customer impact. Now we can act faster, smarter, and at any scale.
Thi Thu Trang Nguyen
Founder and CTO, VBEE AI
As part of the next stage of growth, VBEE AI is now focused on becoming a one-stop shop for all voice needs, offering, alongside voice cloning and text-to-speech, new AI models for transcription and translation. The company is also planning on expanding to Southeast Asia, counting on Google Cloud to provide the scalable infrastructure to support this vision.
A leader in humanlike Conversational AI in Vietnam, VBEE AI is on a mission to automate interactions with lifelike, emotional AI voices to deliver superior and more efficient user experiences.
Industries: Technology, Startup
Location: Vietnam
Products: Cloud Run, Cloud Storage, Cloud Build, Artifact Registry, Cloud Monitoring
About Google Cloud partner - Cloud Ace
Cloud Ace is a Google Cloud Partner serving the Asia Pacific with consultation on Google Cloud deployment, architecture, migration services, data/AI, and system optimization. It has over 100 engineers across Japan, Singapore, Vietnam, Thailand, Taiwan, and Indonesia.
