VIDTENSOR TECHNOLOGIES INC. / VIDTENSOR AI STUDIO

The company behind
responsible video compute.

Company information, service terms, privacy commitments, and safety policies for the VidTensor AI Studio.

About Terms of Service Privacy Policy Acceptable Use Contact

01 / About VidTensor AI

Built for media at machine scale.

VidTensor Technologies Inc. operates VidTensor AI Studio, an engineering-led video generation and rendering platform. The company was formed to make high-quality video synthesis more dependable and accessible by pairing modern model pipelines with resilient distributed compute.

Our engineering roots

Our core engineering team brings together distributed-systems engineers, PyTorch C++ operator and compiler specialists, and computer-vision researchers from Silicon Valley and Singapore. The team focuses on making demanding video workloads observable, restartable, and reproducible—from model execution and tensor kernels to queueing, storage, and delivery.

Why heterogeneous consumer GPUs

We build a high-throughput, stateless video-rendering network across heterogeneous consumer GPU hardware, including RTX 3080 Ti, RTX 4070, and RTX 4090 classes. This fleet strategy lets us schedule independent work across diverse nodes, keep capacity elastic, and recover quickly from interruptions. Jobs are split into bounded 10-minute micro-batches with checkpoints, so a single node failure does not strand an entire production run.

Hardware availability and performance vary by region and workload. GPU model examples describe supported node classes, not a promise that every model is continuously available in every location.

Singapore HQ & Global Distributed Engineering Team

Singapore HQ

One Raffles Quay, #28-12 North Tower, Singapore 048583

US Engineering Hub

548 Market St, Suite 62410, San Francisco, CA 94104, USA

The Global Distributed Engineering Team spans Singapore, San Francisco, Zurich, and Tokyo, with 24/7 site reliability engineering and kernel development coverage. Team locations describe distributed operations and do not change the legal contracting entity or applicable terms.

02 / Legal agreement

Terms of Service

Effective November 10, 2026. These terms govern access to VidTensor AI Studio and related APIs, unless a separately signed enterprise agreement states otherwise.

1. Accounts, credentials & API key security

You must provide accurate account information, protect your credentials, and promptly notify us of suspected unauthorized use. API keys are confidential credentials: do not publish them in browser code, public repositories, or client applications. You are responsible for activity performed using your account unless caused by our breach of these terms. We may suspend compromised credentials to protect customer data and the platform.

2. Render Credits, billing & cancellation

Plans include the stated monthly Render Credits and queue capacity. Credits are consumed for successfully completed output according to the applicable rate schedule; failed or preempted shards are not charged. Unless a signed order form says otherwise, monthly credits expire at the end of the billing period, are not redeemable for cash, and are non-refundable once purchased, except where applicable law requires otherwise. Subscription fees are billed in advance; taxes and approved overages may apply. Cancellation stops future renewal and does not automatically refund prepaid fees.

3. Stateless execution & service levels

Jobs may execute as bounded 10-minute stateless micro-batches on heterogeneous infrastructure. Work can be retried or rescheduled after interruption; availability of a particular GPU or region is not guaranteed. Any SLA, including a 99.95% Batch Completion SLA, applies only to the eligible service, measurement window, exclusions, and remedies explicitly stated in a signed enterprise order form. Dashboard status and marketing availability figures are operational telemetry, not an individual customer guarantee.

4. Intellectual property and customer materials

As between you and VidTensor, you retain all rights in your input media and custom LoRA weights. Subject to your compliance with these terms and third-party rights, you own 100% of the commercial rights in your synthesized output videos. You grant us only the limited, non-exclusive rights needed to process, transmit, secure, and return your inputs and outputs to provide the service. We do not use customer media or custom weights to train foundation models.

5. Acceptable use, service changes & suspension

You must follow the Acceptable Use Policy, applicable laws, and third-party license terms. We may investigate credible abuse, suspend access to prevent imminent harm, and notify you where lawful. We may update service features; material changes to these terms will be posted with a revised effective date, and continued use after that date constitutes acceptance where permitted by law.

6. Disclaimers & limitation of liability

Except for express written commitments, the service is provided “as available” and to the maximum extent permitted by law without implied warranties of merchantability, fitness for a particular purpose, or uninterrupted operation. Neither party is liable for indirect, incidental, special, consequential, exemplary, or punitive damages, or lost profits, revenue, data, or goodwill. To the maximum extent permitted by law, each party's aggregate liability arising from the service is limited to the fees paid or payable for the service during the 12 months preceding the event giving rise to the claim. These limits do not restrict liability that cannot legally be limited, including applicable rights of consumers.

7. Indemnity & third-party claims

You are responsible for having rights and consents for submitted materials and for your use and distribution of outputs. To the extent permitted by law, you will defend and indemnify VidTensor Technologies Inc. against third-party claims arising from your unlawful content, infringement, or material breach. We will provide reasonable notice and cooperation; no settlement may impose an admission or obligation on the indemnified party without its consent.

8. Governing law & general terms

These terms are governed by the laws of Singapore, without regard to conflict-of-law rules, and the courts of Singapore have exclusive jurisdiction, except where mandatory consumer law provides otherwise. If a provision is unenforceable, the remainder remains effective. These terms, the privacy policy, and any signed order form constitute the agreement for the service and supersede prior discussions on the same subject. For legal notices, contact [email protected].

03 / Data stewardship

Privacy Policy

GDPR (EU 2016/679) & CCPA / CPRA aligned privacy practices. Last updated November 10, 2026.

Zero Customer Media Training Guarantee

Customer short-drama masters, voice samples, e-commerce SKU assets, custom LoRA weights, and generated media are never used to train foundation models.

Information we process and why

We process account and billing records to provide the service, authenticate users, prevent fraud, and meet legal obligations. Job metadata—including model, settings, queue state, timestamps, and usage—is used for scheduling, billing, support, reliability, and security. Customer media is processed only to execute the requested workload and return results. We do not sell personal information or use customer media for advertising profiles.

Ephemeral Stateless Worker Memory Hygiene

Each pt_aten_worker node is designed to clear GPU memory (VRAM Zeroing) and its container temporary directory immediately after a 10-minute micro-batch completes or terminates. The worker does not retain customer plaintext media slices after that execution. Checkpoints and finished outputs are stored only where required to resume the job or deliver the service, under the configured retention window; deletion and lifecycle policies remove them when that purpose ends, subject to legal or security holds.

Encryption & access controls

Data is encrypted in transit using TLS 1.3 and at rest using AES-256-GCM where supported by the service storage layer. Access to production systems is role-restricted, logged, and granted only to personnel who need it for operations, support, or incident response. Enterprise private checkpoint and object-storage access is scoped to authorized jobs and configured credentials.

GDPR and CCPA / CPRA rights

Depending on your location, you may request access, correction, deletion, portability, restriction, or objection to processing, and may appeal a decision where law provides that right. California residents may exercise rights to know, correct, delete, and opt out of sale or sharing; VidTensor does not sell or share personal information for cross-context behavioral advertising. We do not discriminate for exercising privacy rights. Contact [email protected] to submit a request; we may verify identity and respond within legally required periods.

Processors, international transfers & retention

We use vetted infrastructure and service providers under contractual safeguards and process data in regions needed to operate the service. Where required, international transfers rely on an applicable transfer mechanism and supplementary safeguards. We retain account and transaction records for service delivery and legal compliance; media retention is limited to the product cache, archive period, or enterprise configuration presented for the relevant plan.

Data Processing Agreements

Business customers acting as controllers may request a Data Processing Agreement (DPA), subprocessors information, and security documentation through the enterprise security contact. A DPA governs processing of personal data on the customer's behalf and does not expand our rights to use customer media.

04 / Trust & safety

Acceptable Use Policy

This policy applies to every account, API integration, job, and output handled through VidTensor AI Studio.

Users may not submit, generate, distribute, or facilitate content or activity that:

Creates non-consensual deepfakes of real people or clones a person's voice without explicit, documented consent and any legally required authorization.
Creates or distributes deceptive political misinformation presented as authentic media, or manipulates civic processes through impersonation or fabricated evidence.
Contains, depicts, solicits, or facilitates child sexual abuse material (CSAM) or sexual exploitation of minors. Suspected material may be reported to competent authorities as required by law.
Infringes copyright, trademark, publicity, privacy, or other rights, or includes maliciously deceptive, harassing, or unlawful content.
Attempts to compromise systems, evade safety controls, distribute malware, or access data without authorization.
Uses VidTensor containers or clusters for any unapproved network scanning or stress-testing workloads, password cracking, or non-video tensor-rendering workload. The fleet is reserved for authorized video synthesis and related video tensor-rendering operations.

Automated safety review and provenance

VidTensor uses automated safety auditing and abuse signals to detect policy violations, protect service integrity, and support incident response. For eligible generated media and supported export workflows, C2PA digital content provenance metadata may be injected or attached to help identify origin and editing history. Provenance metadata is not a guarantee that downstream copies preserve attribution. We may rate-limit, block, or suspend jobs that violate this policy and may preserve relevant records where legally required.

To report abuse or raise a safety concern, contact [email protected].

05 / Contact & infrastructure gateways

Reach the right team.

For business, customer support, enterprise security, compliance, and compute-fleet planning, use the direct channels below. The interactive form creates a local inquiry reference and confirmation; it does not transmit data to a remote ticketing service.

Commercial & Customer Support

[email protected]

Enterprise Security & Compliance