Privacy Policy: Capshy

Effective date: 2026-11-13 · Last updated: 2026-11-13

Draft. Prepared with reference to Israeli law (Protection of Privacy Law, 5741-1981, including Amendment No. 13), the EU GDPR and UK GDPR, California law (CalOPPA, CCPA/CPRA), COPPA, and the Google Play User Data policy. Before publication, every statement must be checked against what the app actually does, and the document should be reviewed by a lawyer. Fill in every [TO COMPLETE] before publishing.

1. Summary

Your video itself never leaves your phone. To create captions, the App sends only the audio of the part you chose to our transcription server. The server transcribes it. To improve transcription, we keep this audio and the captions before and after your corrections, and train our transcription models on them (section 3A). This is on by default, and you can turn it off at any time in Settings, and then nothing new is kept. You can ask us to delete what was already kept (section 14). With the setting off, the audio is discarded from memory as soon as transcription ends.
We collect very little: your Google account details (name and email), your purchases and usage quota, and basic technical data.
Only if you choose to use cloud AI features (such as "improve transcript" or translation), the text of the transcript (never the video or audio) is sent through our server to our AI provider, processed and discarded. We do not store it.
We do not sell data or use third-party tracking or analytics tools. Ads are shown only if you choose to watch one in "Free credits", for another minute of transcription or credits. Then, and only then, Google AdMob receives technical data from your phone that it needs to show the ad (section 5). The ads are not personalized, and without that choice no ad is ever shown, on any plan.
You can delete your account and its data at any time, in the app or on a web page.

2. Who we are

Capshy, formerly Yiddish Captions AI (the "App" or "Service") is operated by Shalom Vider, a sole proprietor (osek patur) registered in the State of Israel ("we", "us"). We are the controller of your personal data under the GDPR and the "database controller" under Israeli law.

Postal address: 122 Rabbi Akiva St., Apt. 13, Bnei Brak 5146205, Israel
Privacy contact email: privacy@capshy.com

This policy covers the App, its server and its website. It does not cover third-party services that the App merely opens or links to (section 8).

3. What we collect, why, and on what legal basis

You are under no legal obligation to provide personal data. Providing it is voluntary. Without a Google account, however, features that require an account (purchases and cloud AI features) are unavailable. You can also use the App without an account ("Continue without signing up"); we then keep only a random install ID, for the transcription quota.

Data	Details	Purpose	GDPR legal basis	Retention
Account data	Google account ID, email address, name and profile picture (if provided by Google), sign-up date, preferred language	Creating your account and signing you in, identifying your plan, support	Performance of a contract (Art. 6(1)(b))	Until you delete your account, plus up to 30 days in backups
Purchases and subscriptions	Google Play order ID, product, dates, status (active, cancelled, refunded), purchase token, country and currency. We never receive your card details, which stay with Google	Verifying purchases, unlocking your plan, handling cancellations and refunds, bookkeeping	Contract. Legal obligation (tax law)	Financial records: 7 years under Israeli tax law. Other data: until account deletion
Usage and quota	How many videos or minutes you processed with quota-limited features, when, which feature, our cost per request, number of trial videos, coupons redeemed	Managing quotas, preventing abuse, cost accounting	Contract. Legitimate interests (fraud prevention)	Until account deletion. Aggregated, non-identifying statistics may be kept longer
Audio for transcription	The audio of the part of the video you chose, without the picture, and the language you chose	Creating the captions you asked for (recognizing the words and their timing). If you asked for "more than one speaker", also telling the voices in the clip apart, to give each speaker a colour. This does not identify who the people are, and no voice characteristics are kept	Contract	With "Help improve transcription" off, not stored: it is held in the server's memory only while it is transcribed and discarded right after. With it on, it is kept to improve transcription (the row "Audio and text for improving transcription"). The transcript sent back to the App stays in server memory for up to 30 minutes, until the App downloads it, and is then discarded. Only the clip's length and the time are kept, for the quota
Install ID (only without an account)	A random number the App creates when installed, unrelated to your identity or device, and how many seconds you transcribed with it each month	Managing the transcription quota for users without an account, preventing abuse	Legitimate interests	Up to 13 months after last use
Text for cloud AI processing (only if you use such a feature)	Transcript text, its language and the target language	Performing the action you requested (improving, correcting or translating)	Contract	Not stored by us. It is passed to the AI provider and discarded from memory when the request completes. For retention by the provider, see section 5
Technical data	IP address, app version, Android version, device model, request times and error codes in server logs	Security, troubleshooting, abuse prevention	Legitimate interests	Server logs: up to 30 days. For a guest who transcribes, a hash of the IP address (not the address itself) is also kept with the transcription record, to limit free transcription from one network
Integrity check (Play Integrity)	With each transcription and AI action: Google's answer on whether the request came from our genuine app, as installed from Google Play, on a genuine Android device. We keep only the result ("genuine", "modified app" and so on), the action, the time and the account or install ID	Preventing fraud and abuse through modified copies of the App	Legitimate interests	90 days
Reports (only if you send one)	The content you choose to attach to a report (for example an offensive caption line), your note and the date	Reviewing the report and improving the Service; also a Google Play requirement for AI apps	Legitimate interests	Up to 12 months after handling, or until account deletion (the report then stays without a link to you)
In-app feedback (only if you send it)	Your message and its kind (idea, problem or general), whether you agreed to be answered by email, app version, device model and language. For a guest: the random install ID instead of an account	Improving the App, fixing problems, and answering you if you asked	Legitimate interests	Up to 24 months. The message is also emailed to the team through our email provider Resend
Automatic error and crash report	When the App crashes or an action fails: the error type, code and technical message, and the code's call chain (stack trace); the screen and step; technical details of the video being worked on (resolution, length, frame rate, HDR, source and export quality); app version, device model, Android version, language, free memory and free storage; and the account or installation ID. A crash is stored on the phone and sent on the next launch. Never the video, the audio or the caption text	Finding and fixing problems	Legitimate interests	Up to 24 months. Also emailed to the team through Resend
Invitations and partner codes (only if you used one)	The invitation or partner code you redeemed, who invited you and when, and whether the benefit was granted	Granting benefits, calculating the partner's share of profit, and preventing fraud	Contract. Legitimate interests	Until the account is deleted. The partner never receives your name or email, only aggregate numbers
Audio and text for improving transcription (while the setting is on)	The audio of the transcribed clip, and the caption text as transcribed and after your corrections, with timing	Training and improving our transcription models and automatic correction (section 3A)	Legitimate interests, with the right to object by turning the setting off. In Israel: consent when accepting the Terms	Until the setting is turned off or the account is deleted, for as long as it is used to improve transcription. A clip no video was created from: 14 days
Why you deleted your account (only if you chose or wrote one)	The answer you picked and what you wrote, before the deletion	Improving the App	Legitimate interests	Up to 24 months, not linked to the deleted account
Correspondence	What you send us by email or form, and your email address	Answering you and providing support	Legitimate interests. Contract	Up to 24 months after the matter is closed

Our legitimate interests are keeping the Service secure, preventing fraud and quota abuse, and fixing faults. We have weighed these against your rights: the data is minimal and not sensitive. You may object at any time (section 9).

Marketing email: we do not currently send marketing email. If we start, we will send it only to people who ticked a separate consent, and every message will be marked as advertising and include an unsubscribe link. Service messages, for example about your account or a purchase, go to everyone.

Sources: account data comes from Google when you choose to sign in. Purchase data comes from Google Play. Everything else comes from you or from the App.

3A. Improving transcription: audio and text

The "Help improve transcription" setting is on by default, and it is stated on the terms screen you accept before first use. You can turn it off at any time in Settings. Turning it off stops keeping anything new. What was already kept, including the audio, is deleted when you ask us to (section 14) or when you delete your account. Without an account, the setting is stored on your phone, and for a deletion request we find what was kept by your installation ID.

What is kept while the setting is on:

The audio of the transcribed clip (not the video or its picture).
When you create the video: the caption text as our model transcribed it, and the text after your corrections, with each line's timing.
The language, the clip's length, whether the AI tools corrected the lines, and the App version.

All of it is linked to your account (or, without an account, your installation ID), only so that we can delete everything when you ask us to or delete your account. Your name and email are not stored with the audio and are never used in training.

Never kept: the video and its picture. Text you pasted yourself ("I already have the text"), and the audio of such a clip. Clips transcribed with "Exact transcription" in the closed trial. Captions translated into another language.

Why: to train our transcription models on real speech, in every language in the App, and to learn from corrections where transcription goes wrong: for example Hebrew-origin words, names, and speech with background noise or music. The trained models serve our Service. We do not use the audio to identify who is speaking, and do not create a "voiceprint" from it. Only we use this material. We do not sell it or give it to others to train their models. Our hosting and computing providers (section 5) process it only on our behalf. Authorized staff may listen to and read samples to check training quality, and every such view is logged.

Legal basis: in Israel, the consent you give when you accept the Terms, which you can withdraw at any time by turning the setting off. Under the GDPR, our legitimate interest in improving the Service. Your right to object (section 9) is exercised by turning the setting off, and by asking us to delete what was already kept, with no reason needed.

Retention: until you ask us to delete it, or until you delete your account. If you never create a video from a clip, the audio and text kept for it are deleted after 14 days. What is no longer used to improve transcription is deleted. Once a year we review whether what is stored is still needed. After deletion, what a model has already learned cannot be "taken out" of it, but the model does not keep the audio or text themselves.

Where: the text in our database at Railway, and the audio in Railway's private file storage, both in Europe. Training itself runs at Modal (section 5), on a temporary copy deleted when it ends.

Please note: captions may include things said in the video, such as names. If you do not want the text kept, turn the setting off before creating the video, or delete what was kept afterwards.

4. What never leaves your phone

The following data is stored and processed only on your phone, and we never receive it:

The videos you select and the captioned videos the App creates.
The picture of your video. Only the audio of the part you chose is sent to the server, to create captions (section 3).
The captions you edited and any text you type or paste, except when you use a cloud AI feature (section 3), and while transcription improvement is on: then, when you create the video, the caption text before and after your corrections is sent to us (section 3A). Text you pasted is never sent for transcription improvement.
Your design settings, templates and saved project.
Clipboard content: the App reads it only when you tap "Paste", and never sends it anywhere.
Teleprompter recordings: the video you record is saved to your gallery. If you choose to caption it, only its audio is sent, as with any video (section 3).

This data is deleted when you uninstall the App or clear its data. Temporary files are deleted when each task finishes. We do not back it up, and Android's automatic backup is disabled for this App.

5. Who we share data with

We do not sell personal information, we do not "share" it for cross-context behavioral advertising (as defined in the CCPA), and we do not provide it to data brokers. We disclose data only to the following, and only as needed:

Recipient	Role	Data	Location
Google (Google LLC, Google Ireland Ltd.)	Sign-in with Google, Google Play store and payments, and checking the App is genuine (Play Integrity)	Account and purchase data, and the integrity token. Google also processes this data under its own privacy policy	USA and EU
Railway (Railway Corporation)	Server and database hosting, and a backup transcription server	All server-side data listed in section 3, the audio to transcribe when the backup is used, and the audio and text for improving transcription (section 3A)	European Union
Modal (Modal Labs, Inc.)	Running the transcription model on a graphics processor (GPU), and training our transcription models	Only the audio of the clip to transcribe, without your name, email or account ID. Not stored. During training: a temporary copy of the audio and text for improving transcription (section 3A), on the same terms, deleted when it ends	USA
Yiddish Labs	"Exact transcription", only in a closed trial for invited users, and only when you turn it on	The clip's audio, without your name, email or account ID. The text they return is never used to train or evaluate models	United States
Anthropic (Anthropic PBC)	AI provider (Claude) for cloud features	Only the transcript text you submit for processing, without your name or email	USA
Google AdMob (Google LLC)	A video ad, only when you choose to watch one for another minute of transcription or credits	The device's advertising ID, IP address, device and OS type, and a random id of ours (an install or account id, no name or email) so we can give you what you chose. We ask for non-personalized ads, rated for all ages. Google also processes this data under its own privacy policy	USA and EU
Resend (Resend, Inc.)	Emailing the team about new feedback	The feedback, and your account email if you agreed to be answered	USA
Our email provider (Google, Gmail service)	Corresponding with you	Your messages and our replies	USA and EU
Our accountant and tax authorities	Bookkeeping and reporting	Financial records	Israel

About Anthropic: under Anthropic's commercial API terms (as of the date of this policy), Anthropic does not train its models on data submitted through the API, and generally deletes it within 30 days. There are exceptions: longer retention where needed to enforce its usage policy or where required by law. [Before publishing: verify against Anthropic's current terms and execute its DPA.]

Other disclosures:

Legal requirements: where the law requires it, under a court order, or on a lawful request by an authority.
Protecting rights: to protect the rights, safety or property of you, us or others, for example in a case of fraud.
Business transfers: if the business is transferred to a company (for example one we establish) or to another party, your data will move with it, subject to this policy, and we will notify you in advance.

Our business partner has access to the Service's admin area as an authorized person, bound by confidentiality and by this policy.

6. International transfers

We are based in Israel. The European Commission has found that Israel provides an adequate level of protection for personal data (2011 decision, reconfirmed in January 2024). Data is also transferred to providers in the USA (section 5). These transfers rely on one of the following:

The EU-U.S. Data Privacy Framework, where the provider is certified.
The Standard Contractual Clauses approved by the European Commission, incorporated in the providers' data processing agreements.

All transfers also comply with Israel's Protection of Privacy (Transfer of Data to Databases Abroad) Regulations, 5761-2001. To obtain a copy of the safeguards, contact us (section 2).

7. Security

Encryption: all traffic between the App and the server is encrypted (HTTPS/TLS).
Access control: database access is limited to two authorized persons, with strong passwords and two-factor authentication.
Keys: API keys (Anthropic and Google) are kept only on the server, never inside the App.
AI and transcription requests: every request first checks your account (or install ID), plan and quota. The transcription server accepts requests only from our server, with a secret key, and the GPU server also with an access key from the cloud provider. Google checks that each request came from the genuine App (section 3).
Audio: audio is never written to logs. With "Help improve transcription" off, it is never written to disk and is discarded from memory when transcription ends. With it on, it is kept in private file storage that only our server can reach, without your name or email.
Legal framework: we follow Israel's Protection of Privacy (Data Security) Regulations, 5777-2017.

No system is perfectly secure. If a breach occurs that is likely to put you at risk, we will notify you and the competent authorities (the Israeli Privacy Protection Authority and, where applicable, a European supervisory authority) as the law requires.

8. Third-party services

At your choice, the App can open external websites in your browser, such as a transcription service like Yiddish Labs, and can share a video to other apps such as WhatsApp. Anything you upload to or share with those services is governed by their terms and privacy policies, not ours. We receive no data from them.

9. Your rights

Everyone, everywhere, may:

Ask what data we hold about you and receive a copy.
Correct inaccurate data.
Delete your account and its data (section 10).
Withdraw any consent you gave us.
Turn transcription improvement off at any time, and ask us to delete what was already kept (section 3A).

Israel: under the Protection of Privacy Law you may inspect the data held about you (section 13) and request that data that is not correct, complete, clear or up to date be corrected or deleted (section 14). If you believe your rights were infringed, you may complain to the Privacy Protection Authority at the Ministry of Justice.

EU, EEA and UK: you have the rights of access, rectification, erasure, restriction, objection to processing based on legitimate interests, and data portability. You may also lodge a complaint with your local data protection authority (in the UK, the ICO). We do not make automated decisions that produce legal or similarly significant effects concerning you.

California and other U.S. states: you have the right to know what we collect, and to request deletion and correction. We do not sell or share personal information, so there is nothing to opt out of. We will not discriminate against you for exercising your rights. The App does not track you across third-party websites or services, so we do not respond differently to "Do Not Track" signals.

How to exercise your rights: email us at the address in section 2. We will verify your identity, usually through your account email, and respond within 30 days. If a complex request needs more time, we will tell you. Exercising your rights is free, unless requests are manifestly excessive.

10. Account deletion

In the App: "My account" → "Delete account".
On the web, without installing the App: https://capshy.com/delete-account.

Audio and text kept for improving transcription are deleted immediately when you delete your account. After deletion, account and usage data are erased within 30 days, including from backups. Only financial records that the law requires us to keep (7 years) are retained, limited to what is required. Deleting your account does not cancel a Google Play subscription. Cancel it in Google Play (see the Terms of Use). On-device data is removed when you uninstall the App.

11. Children

The Service is not directed to children under 13, and we do not knowingly collect data from them. Where local law sets a higher minimum age, that age applies. Users under 18 need the consent of a parent or guardian, in particular for purchases. If we learn that we collected data from a child below the permitted age, we will delete it. A parent who believes this happened may contact us.

12. App permissions

Permission or access	Why
Video selection (Android Photo Picker)	Access only to the video you choose, not to your whole gallery
Saving to Gallery and Downloads (MediaStore)	Saving the finished video and the subtitle file
Notifications	Showing progress and a cancel button during transcription or video creation
Foreground service	Continuing work while the screen is off
Internet	Sending the audio for transcription, sign-in, purchase verification and cloud AI features
Advertising ID (AD_ID)	Showing an ad you chose to watch, and preventing ad fraud. You can reset or delete it in Android settings
Camera and microphone	Only in the teleprompter, when you film yourself. The recording stays on your phone and is not sent to us. "Scroll by voice" measures only the volume, on the phone. Permission is requested only when you open the teleprompter

13. Changes to this policy

When we change this policy, we will update the date at the top. For material changes, such as a new type of data or a new use, we will notify you in advance in the App or by email, and ask for your consent where the law requires it.

14. Contact

Questions, requests and complaints: privacy@capshy.com · 122 Rabbi Akiva St., Apt. 13, Bnei Brak 5146205, Israel.