Privacy policy

1. The short version

How-To Maker runs on your computer, and your recordings stay there. The video file itself is never sent over a network. Turning a video into a transcript, and writing a tutorial from one, can send content to a service, and which service that is depends on a choice you make. Every action that sends anything says what it will send before you take it.

2. The three ways a model can run

Each of the two jobs, turning your video into a transcript and writing your tutorial, is done by a model. There are three ways to run one, and you choose:

• Our own cloud service. The application sends the work to our server, which passes it to the provider that runs the model. This is the only path we charge for and the only one metered against the credit on your account.
• Your own account with a service you chose. You supply an API key and the application talks to that service directly. The request does not pass through us, we cannot see it, and no credit is spent.
• A model on your own machine. Nothing crosses a network at all.

Some of the services that run a model on your own machine can be pointed at a different machine instead. When you have done that, the work is sent there, and the application says so on that row rather than telling you nothing leaves.

This policy can only describe the first path in full. On the other two the arrangement is between you and whoever you chose.

3. What is sent, and what is not

The whole video file is never sent over a network. No part of the application uploads it today. If a feature ever needs a model to see what was on screen, it may send a short clip for that one purpose, and only where you have agreed to that action at the time it happens. The complete recording is not part of any of it.

Beyond that, what leaves depends on the job:

• Turning your video into a transcript can send the sound from your video to the service you chose. Your video file itself is never sent.
• Writing your tutorial sends your transcript, and the text of the document being written, to the service you chose. Your video and its sound stay on your computer.
• A model running on your own machine sends neither.

Every network operation in the application declares which classes of content it sends, and what you are shown is generated from that declaration rather than written next to it. A change in what an operation sends changes what you are told, without anyone having to remember to edit a sentence.

4. Where your work goes after it leaves us

This policy does not name the services that run the models. That is deliberate rather than evasive. You pick which one does the work, several of the available choices can be pointed at a cloud we have no visibility of, and naming one would be us making a promise about somebody else’s product that we cannot keep true.

What we can tell you, and keep telling you in a year, is what we send and where our part of it stops. How the service you chose handles your work is up to them. If that matters to you, check before you use it.

5. Our own transcription service, and where the text sits

If you use our transcription rather than one of the other two paths:

• The sound from your video streams through our server to the speech to text provider we use. We do not keep a copy of the audio.
• When the transcript is ready, our server fetches the text, holds it in our transcript storage, and deletes both the transcript and the uploaded audio at the provider. The job is not complete until that deletion has succeeded, and an hourly job keeps retrying any deletion that failed until it does.
• The application collects the text and then tells us it has it, which deletes our copy.
• A transcript that is never collected is deleted by a seven day expiry rule on that storage. Nothing waits there indefinitely.

6. What we store about your account

• Your email address, to identify your account.
• Your subscription status, the date your access runs to, and your credit balance and the entries that make it up.
• A record of each metered call: which model, the token counts or the duration of audio, what it cost, and the status of your account at the time. We do not store the text of a request or of a response.
• A record of each transcription job: its state, the duration charged for, and any error. Transcript text is held only as described in section 5.
• A salted hash of a machine identifier for each device you activate, with the label and platform. We never store a raw machine identifier.
• A hashed IP address against each download from this website, which is what makes rate limiting possible. The address itself is not stored.
• Which version of each legal document you accepted and when, with a hash of the address it was accepted from as evidence that it happened. The address itself is not stored.

We do not sell any of it, we run no advertising, and we do not use your work to train models.

7. Who else processes your data

• Supabase, for authentication and database hosting.
• Cloudflare, for website hosting, our API, and the storage described in section 5.
• Stripe, for payments. Card details are handled by Stripe and never reach us.
• The provider that runs the model behind our own cloud service, and the speech to text provider behind our own transcription. We do not name either of them, for the reason in section 4. If you want to know who they are before you use our cloud service, ask through the contact form and we will tell you.
• Where you chose your own service or your own machine instead, nobody, because nothing on that path reaches us.

8. Retention and deletion

You may delete your account through the contact form. Deletion removes your profile, entitlement, devices, wallet and usage history. Transcripts held under section 5 go with it. Records we are required to keep for tax or accounting purposes are retained for as long as the law requires.

9. Changes

Each revision is published as a new numbered version with an effective date, and previous versions remain readable on this site.