ASSETS
What content do you have?
Review originals, derivatives, images and audio of people, and terminology.
LETR AI / ENTERPRISE
A place where your content
keeps creating value.
Start with your assets, workflows and quality standards. Validate features with real content, then prepare the review and operations your team needs.
↓A shared space for content and people
AI possibilities. Human standards.
Our clients and partners
Helping content businesses, professional service providers and public institutions get more value from their content.
BUILT AROUND YOUR TEAM
Understand the work before choosing the tools.
YOUR CONTENT. YOUR STANDARDS.We start with how your team works.
WORKFLOW
What results do you need?
Prioritize discoverability, buyer presentations, documentary footage review and edit planning.
STANDARDS
Who reviews and uses it?
Clarify owners, review standards, schedules and access requirements.
TECHNOLOGY WITH A PURPOSE
Technology, explained
through the work it supports.
From understanding content to reviewing output and operating models. Explore the research and intellectual property behind Twigfarm’s technology.
- 01
Understand the content
Turn content into findable information
Transcription · Speakers · OCR
Use transcription, speaker separation and OCR in search and review.
- 02
Review with people
Apply your team’s standards
Context · Terminology · Review
Produce translations using context and terminology, then review with people.
- 03
Keep models running
Connect evaluation and operations
Evaluate · Deploy · Monitor
A technology foundation for model and data versions, evaluation and operations.
LETR AIContent understanding · Hybrid translation · MLOps
Registered patents and applications are identified separately. Confirm applicability for the features you adopt.
Explore patents, applications & researchYOUR DATA, CLEARLY EXPLAINED
Understand how data is used.
Know how it is protected.
LETR AI never uses customer data to train its own models. Only when a customer requests a custom model do we train and supply a dedicated model using the customer’s data under a separate agreement.
Clear scope. Clear rules.
Review separate rules for personal data, external AI processing and account connections.
Encryption & access
Encryption of personal data in transit and at rest, limited handlers, and access records.
External AI services
We use paid APIs from OpenAI, Anthropic (Claude), Google (Gemini), NAVER Cloud (CLOVA Speech) and ElevenLabs. Both direct connections and cloud services are paid.
Google account protection
Data received from Drive and YouTube has separate rules for training use, disconnection and deletion.
Check the rules that apply to your data
Which external AI services are used?
All external AI APIs used by LETR AI are paid services. We use OpenAI, Anthropic’s Claude and Google’s Gemini through direct APIs and cloud services, NAVER Cloud’s CLOVA Speech for speech recognition and transcription, and ElevenLabs for speech synthesis and dubbing. Data processors, transmitted data and processing regions vary by feature and connection route.
External AI service informationDoes LETR AI use customer data to train its own models?
No. LETR AI never uses customer data to train or improve its own models.
Customer data use principleWhat if a customer needs a custom model?
Only when a customer requests a custom model do we enter into a separate agreement and supply a dedicated model trained on the customer data specified in that agreement. The agreement defines the training data, purpose and scope. This data is not used to train LETR AI’s own models.
Customer request · Custom model under a separate agreementDo external AI providers use customer data for training?
LETR AI uses paid APIs only. Our operating policy excludes customer data from external providers’ general-purpose model training and improvement. The providers publish the following API data-use conditions.
-
OpenAI API
API data is not used to train or improve models by default. Explicitly opting in to share data for training is an exception.
OpenAI data-use policy -
Anthropic · Claude API
Commercial API inputs and outputs are not used for model training by default. Submitting feedback or bug reports to the provider, or otherwise allowing data use, creates an exception.
Anthropic training policy -
Google · Paid Gemini API
Paid-service prompts and responses are not used to improve Google products. Prompts include instructions, attached images, videos, documents and cached content.
Gemini paid-service terms -
NAVER Cloud · CLOVA Speech
Used for speech recognition and transcription. Article 4 permits audio and video data to be used to improve the recognition engine only with customer consent. LETR AI’s operating policy prohibits opting in to this data sharing.
CLOVA Speech service terms -
ElevenLabs · Speech synthesis & dubbing
Paid use alone does not automatically exclude data from training. Disabling “Improve the models for everyone” in account settings and saving the choice excludes newly submitted data from model training. LETR AI’s operating policy requires a training opt-out or contractual exclusion from training.
ElevenLabs training policy
LETR AI’s operating policy prohibits opting in to data sharing for external providers’ general-purpose model training or improvement, or submitting customer content as feedback to providers.
Conditions when connecting through cloud services
These policies apply when connecting through the respective service. Direct APIs and cloud services are governed by their respective contracts.
-
Azure OpenAI
Data is not used to train foundation models without customer permission or instruction. Inputs and outputs are not made available to OpenAI.
Microsoft data privacy policy -
Amazon Bedrock
AWS states that customer content is not used to improve base models and is not shared with model providers.
Amazon Bedrock official FAQ -
Google Cloud · Vertex AI
Customer data is not used to train or fine-tune AI/ML models without prior customer permission or instruction.
Google Cloud data-use policy -
Claude in Microsoft Foundry
Anthropic acts as an independent data processor and its commercial terms apply. Its processing arrangements differ from Azure OpenAI.
Foundry Claude-specific policy
Exclusion from training does not mean no data is retained. Provider- and feature-specific rules govern logs and files retained for service delivery, security and legal obligations.
Which rules apply to Google-connected data?
The policy prohibits using data received through Google APIs, including Drive and YouTube, for training, fine-tuning or evaluation. Data is deleted within 30 days of disconnection, account closure or a deletion request, subject to legal retention. These rules are distinct from terms for external AI models such as Gemini.
Terms Article 7 · Privacy Policy Article 5Can you access every file in my Google Drive?
The policy limits processing to files you explicitly select in the file picker, excluding unselected files and the full Drive listing. You can revoke access in your Google account settings.
Privacy Policy Articles 2 & 5How is personal data protected?
Personal data is protected with TLS encryption during transmission and AES-256 encryption during storage. We also manage access permissions and access records.
Privacy Policy Article 9Is every record deleted immediately when I leave?
Personal data is generally destroyed when its purpose is met, the account is closed or consent is withdrawn. Legal exceptions include five years for contracts and payments, and three years for complaints and disputes. Google data has a separate deletion period of up to 30 days.
Privacy Policy Articles 4, 5 & 8How are videos shared with buyers protected?
Media Sales Kit supports sharing with email authentication and DRM. Confirm supported environments and conditions during consultation.
Product feature informationSTART SMALL. BUILD TOGETHER.
Validate with real content. Expand from there.
- 01
Understand the work
Review current challenges, available assets and adoption goals.
- 02
Validate real material
Assess output quality, review workflows and practical value.
- 03
Define the operation
Agree on features, owners, data and access requirements.
Where should your team begin?
Tell us about your content and what you want to achieve.