How blablabla handles your scripts

Actors work with confidential material. Some scenes are under NDA, some are pre-release, some you just don't want floating around. We get it - one of us used to be one of you.

This page tells you exactly what happens to your script when you use blablabla. No marketing language, no vague reassurances. Just the facts, kept current. If something here changes, the date at the top changes with it.

What stays on your phone

The following data is stored locally on your device, in private app storage protected by the security controls built into iOS or Android:

• Your saved scene files (JSON)
• All generated audio (cached for offline rehearsal)
• Self-tape recordings, until you save them to Photos or delete them
• Rehearsal history and accuracy scores
• App preferences and settings

We can't inspect this local storage. Other apps can't read it either. Delete the app and its local data is removed from the device.

What never leaves your phone

Fountain (.fountain) and Final Draft (.fdx) files are parsed on the device by our own code. Signed in or not, that text is not sent to our server or to any AI service. If you have a choice of format, this is the most private way to bring a scene in.

The text layer of a PDF is also extracted on the device. Only the extracted text, not the file, goes on to parsing when the scene needs it.

What passes through the cloud

Two things need cloud processing when you're signed in. The app asks first: the first time you import a PDF, Word file or photo, or generate a premium voice, it explains what will be sent and waits for a yes. Decline and you stay on the offline path described below.

Scene parsing

When you import a scene that isn't Fountain or Final Draft, the text goes to our server, which forwards it to an AI model that turns it into structured dialogue. As of this update the order is: OpenAI (GPT-5.6 Luna) first; Anthropic (Claude Sonnet) when the first result is low-confidence; a smaller OpenAI model only if both error out. Photographed or scanned pages go to Anthropic (Claude) for reading. PDF and Word files that need layout understanding are sent to OpenAI as a file with storage disabled.

Our server keeps the text only for the seconds it takes to return the parsed result to your device, and it does not log it. The parsed scene is then saved on your device and, if you're signed in, backed up to your account (see below).

OpenAI and Anthropic receive the text to process it. Neither trains on it. OpenAI keeps API inputs for up to 30 days for abuse monitoring unless the customer has been approved for zero data retention. Anthropic states that API inputs and outputs for the models we use are not retained by default.

Voice generation

Dialogue lines are sent to ElevenLabs through our server for text-to-speech. The audio comes back and is cached on your device for offline use. We don't store the text or the audio on our server.

ElevenLabs receives the dialogue text to generate speech and does not train on it.

What's stored on our servers, and where

Our database and file storage run on Supabase in the AWS eu-west-1 region (Ireland). Our API runs on Vercel. Everything is encrypted in transit (TLS) and at rest.

• Account info (name, email) if you sign in with Google or Apple
• Scene backups: the structured scene, synced so you can open it on another device. Deletable one scene at a time, and removed entirely when you delete your account.
• Share links: when you share a scene with a scene partner, the script and its audio are stored for that link for up to 90 days, then deleted. Search engines are told not to index share pages.
• Ensemble productions: a theatre's scripts are visible only to members the theatre has admitted, and access ends when the theatre removes them.
• Usage logs: character counts for billing. No script content.

Delete your account in Settings and all of this is removed right away. You don't need to write to us or wait for a review.

Crash reports and analytics

Crash reports go to Sentry, hosted in Germany. The app sends no personally identifying information by default. Server-side reports are cut down to a short list of technical fields before they leave our code, so request bodies and script text cannot end up in them. Product analytics stay in our own database and never include scene text. You can switch both off in Settings.

Our subprocessors

These are the companies that process data on our behalf, what each receives, and where it runs. We have a data processing agreement with each of them.

• Supabase - database, authentication, file storage. Account data, scene backups, share files. Ireland (AWS eu-west-1).
• Vercel - hosting and API. Script text in transit only. USA.
• OpenAI - scene parsing. Script text. USA.
• Anthropic - scene parsing and page reading. Script text and page images. USA.
• ElevenLabs - voice generation. Dialogue text. USA.
• Sentry - crash reporting. Technical metadata only. Germany.
• Resend - transactional email (welcome and account-deletion emails). Email address. USA.
• Apple and Google - sign-in and payments under their own terms. We never see card details.

Transfers outside the EU rely on the EU Standard Contractual Clauses included in each provider's agreement. If we add a provider, this list changes before the provider goes live.

The fully offline path

If you use blablabla without signing in, or decline the AI consent, scenes are parsed on-device (Fountain, Final Draft, PDF text extraction) and voices use the speech synthesis built into iOS or Android. Nothing is sent to our servers or to the services above.

You lose premium voices and AI parsing for unstructured scripts, but your script stays entirely on your device.

No training

None of the services we use (OpenAI, Anthropic, ElevenLabs) train on data sent through their API. Each provider's API terms say so in writing. Your scenes are not used to improve anyone's AI, including ours.

Working under NDA

If a production requires it, we will sign the production's NDA or provide our own data processing agreement. Email support@sayblablabla.com and we'll turn it around quickly.

Some practical advice for NDA material: import the Fountain or Final Draft file when you have one, since nothing leaves the phone. Rehearse with the built-in voices if the production forbids third-party processing. Delete the scene from the app when the job is done; deleting a scene removes its backup from our servers too.

Questions

If you work under NDA and need specifics about how blablabla handles your material, email support@sayblablabla.com. Happy to walk through it. For the full legal version, see our privacy policy.