Privacy Policy
This Privacy Policy explains how SORI handles personal information when you use AI-MIDI, including how Web Transcription Inputs and Outputs are processed for model training, evaluation, validation, quality measurement, and improvement under the separate Training Data Use Agreement. This English edition includes notices for users in the United States and European Union. Mandatory local law applies.
- Effective
- 2026-08-06
- Displayed language
- English · original
This English edition includes notices for users in the United States and European Union. Mandatory local law applies.
1. Scope and controller
SORI is the controller of personal information processed through AI-MIDI. This edition includes additional notices for users in the United States and European Union. Mandatory local law applies.
2. Information we process and why
Account and authentication
Email address, username, name, country, password hash, email-verification status, and Google sign-in identifiers are processed to create and secure your account, authenticate you, and provide account features.
Audio-to-MIDI service
Uploaded files, file name, format, size and duration, storage reference, job ID, selected model, processing status, and generated MIDI are processed to perform the conversion you request. A file that contains voice, lyrics, tags, or other identifying details may contain personal information.
Web Transcription model training, evaluation, and improvement
After the separate Training Data Use Agreement is accepted, we process copies of Web Transcription Inputs and generated MIDI Outputs, file size and duration, an opaque training-intake identifier, consent-document version and acceptance time, creation time, and anonymization status to train, evaluate, validate, measure the quality of, and improve AI-MIDI models. After a successful conversion, a training copy may be created in storage separated from operational service data. Account identifiers, original file names, and operational storage references are not included in the training-intake record. Until the file content can no longer reasonably identify a person, including through voice, lyrics, tags, or other available information, the copy is treated as personal information rather than anonymous information.
Legal bases
Processing needed to perform the requested conversion is based on performance of the service contract where that basis is available. Processing Web Transcription Inputs and Outputs for model training, evaluation, validation, quality measurement, and improvement is based on the separate Training Data Use Agreement and applicable consent requirements. Other processing may rely on legal obligations, security and fraud-prevention requirements, or another lawful basis described by applicable law.
Operations and security
IP address, browser, operating system, device information, request and error logs, timestamps, job identifiers, and security events are processed to operate the service, prevent abuse, diagnose failures, and protect users and SORI.
Billing and support
Paddle customer, subscription, product, transaction and payment-status identifiers, currency, credit usage, support messages, and attachments are processed for billing, subscription management, tax, refunds, customer support, and disputes. SORI does not store full payment-card numbers.
3. Service providers, disclosures, and international transfers
We use service providers for cloud infrastructure, storage, queueing, MongoDB Atlas database hosting and backup, authentication, billing, email, support, analytics, and monitoring. For Web Transcription, cloud infrastructure providers may also process training copies and training-intake metadata for separated storage, anonymization, model training, evaluation, validation, quality measurement, and improvement under the Training Data Use Agreement. These providers process information only for the contracted service, subject to applicable safeguards. Information may be processed in the United States, the European Union, and other locations where SORI or a provider operates. We may disclose information when required by law, to protect rights and security, or in connection with a corporate transaction with appropriate notice and safeguards. We do not provide personal information to an unrelated third party for that party's independent purpose without consent or another lawful basis.
4. Retention and deletion
Account and operational data
We retain information only while needed for the purposes described here. Personal information that becomes unnecessary after account deletion or completion of its purpose is deleted without undue delay, subject to legal retention obligations. Operational Web Transcription Inputs and Outputs are deleted when account deletion is completed, and isolated disaster-recovery backups are deleted within no more than 30 days.
Web and API files
Operational Web Transcription files follow the account-deletion and backup periods above. AI-MIDI API upload bodies and generated MIDI files are not used for training and are deleted within 24 hours. Necessary job-status, usage, credit, billing, security, error-response, and legal records may be retained for their applicable purpose.
Training copies and anonymous training data
A training copy awaiting anonymization is retained only until anonymization is completed and never for more than 30 days after creation. If anonymization is not completed within that period, or consent withdrawal or account deletion occurs first, the pending copy is deleted. Data is classified as anonymous and incorporated into a training dataset only after SORI determines that a person can no longer reasonably be identified when available information and the time, cost, and technology required for identification are considered. Anonymous training data may be retained until SORI permanently discontinues all AI-MIDI model training, evaluation, validation, quality-measurement, and improvement purposes. SORI does not reconnect anonymous training data to account information or use it to identify a person. Data that retains a reasonable possibility of identifying a person is not treated as anonymous and remains subject to the applicable personal-information deletion rules.
Legally retained records
Records required for transactions, tax, fraud prevention, security, disputes, or legal claims are segregated where practical, limited to what is necessary, and deleted when the applicable period ends.
5. Cookies, security, and children
We may use essential cookies for authentication, security, and core service functions. Non-essential analytics technologies are used in accordance with applicable notice and choice requirements. We use administrative, technical, and physical safeguards such as access controls, encryption in transit and at rest where supported, logging, separated training storage, backups, vulnerability response, and service-provider oversight. AI-MIDI is not directed to children under 14, and we delete an underage account and related personal information when identified, subject to legal exceptions.
6. Your choices and rights
Depending on your location, you may request access, correction, deletion, restriction, portability, or an objection to processing, and may withdraw consent where processing is based on consent. A request to withdraw the Training Data Use Agreement may be submitted to team@sori-ai.com. We delete pending or otherwise identifiable training copies covered by a valid withdrawal request and stop collecting later Inputs and Outputs for training under that consent. Data already verified as anonymous may continue to be retained and used as described above. If information can still reasonably identify a person or be retrieved from a model, we evaluate the request under applicable law rather than treating the information as anonymous. You may also delete your account through Account settings. We may verify your identity and may retain or refuse a request where permitted or required by law.
7. United States notice
Where an applicable U.S. state privacy law applies, residents may have rights to know or access categories and specific pieces of personal information, correct inaccuracies, delete information, obtain a portable copy, and appeal a denied request. We do not sell personal information or share it for cross-context behavioral advertising, and we do not discriminate against a user for exercising an applicable privacy right. We limit retention of personal information to what is reasonably necessary and proportionate for the disclosed purpose. A verified request may be submitted to team@sori-ai.com. An authorized agent may submit a request where local law permits and appropriate authority can be verified.
8. European Union notice
For users in the European Union, processing may rely on performance of a contract, compliance with legal obligations, consent, or legitimate interests such as service security, abuse prevention, and troubleshooting, after balancing those interests against your rights. Processing Web Transcription Inputs and Outputs for model training, evaluation, validation, quality measurement, and improvement relies on the Training Data Use Agreement and applicable consent requirements rather than the service-improvement legitimate interest described here. You may exercise rights of access, rectification, erasure, restriction, portability, objection, and withdrawal of consent, and may lodge a complaint with your local supervisory authority. Where personal data is transferred outside the European Economic Area, we use an available transfer mechanism and supplementary safeguards where required, such as adequacy decisions or standard contractual clauses.
9. Contact and policy changes
Privacy questions, Training Data Use Agreement withdrawal requests, and other rights requests may be sent to team@sori-ai.com. We will publish material changes on this page with the new effective date and version, and retain prior published versions for audit purposes.
team@sori-ai.com