charactr Download free
charactr › Trust & safety › ai-policy

AI Providers & Disclosure

The third-party AI providers charactr uses, what data is sent to each, and the controls you have.

Last updated: April 29, 2026

Effective: April 29, 2026 · Last updated: April 29, 2026

charactr is voice-enabled software for conversational interactions with virtual persons — built as an AI film studio for talking AI characters. We do not train foundation models ourselves. We deliver the experience by routing each generation step to the most suitable third-party AI provider for that step.

This page exists for transparency and to satisfy our compliance obligations under Apple App Review Guideline 5.1.2(i) (effective November 13, 2025) and EU AI Act Article 50 (effective August 2, 2026), as well as analogous state laws (including California SB 243, New York S.8420-A) and our voluntary commitments under the C2PA Content Credentials standard.

1. What "AI" means in charactr

The Service uses general-purpose AI ("GPAI") models supplied by third parties, including text-to-video generation, image-to-video generation, voice synthesis, automatic speech recognition, language understanding, and image generation. charactr is a "deployer" of these systems within the meaning of the EU AI Act. Some user-facing features qualify as "AI systems intended to interact directly with natural persons" under EU AI Act Article 50(1) and as "synthetic content" generators under Article 50(2).

2. Provider categories and named examples

The provider mix evolves as foundation models improve. The current named providers in each category are surfaced in the App at the point of generation; the categories are:

  • Video generation — for cinematic AI video and visual scene rendering. Example providers (live list maintained in-app): video tier — Seedance / ByteDance and other video-foundation-model providers.
  • Voice synthesis — for voice-acted dialogue and character voices. Example providers: voice tier — ElevenLabs, OpenAI.
  • Speech-to-text (automatic speech recognition) — for voice-driven direction, where you speak to direct or perform. Example providers: STT tier — OpenAI Whisper.
  • Language understanding — for dialogue generation, scene structure, character behavior, and conversational character interactions. Example providers: language tier — OpenAI, Anthropic, xAI.
  • Image generation — for character references, storyboards, and visual design. Example providers: image tier — Black Forest Labs, OpenAI.

The specific provider used for any individual generation depends on the request, availability, and quality routing. The live list of active providers is surfaced in the iOS App at the point of generation. When you tap to render a scene, voice, or image, charactr displays which provider is handling that step. This satisfies Apple's 5.1.2(i) point-of-generation disclosure requirement.

3. What gets sent to providers

  • The prompt or direction you have given (text or transcribed voice).
  • Character references (descriptions, voice parameters, visual references) relevant to the request.
  • Scene context (the immediate scene state — preceding lines, mood, setting).
  • Voice parameters (timbre, pacing, performance notes) for voice synthesis.

4. What is never sent to providers

  • Your account credentials, password, or authentication tokens.
  • Your payment information.
  • Your full conversation or scene history beyond the immediate context window required for the request.
  • Other users' content.
  • Personal data unrelated to the generation.

5. Provider data handling and training opt-outs

Each provider has its own data-handling terms. charactr's posture:

  • We use enterprise / zero-retention provider modes where available.
  • Where providers offer an opt-out from training on submitted data, charactr opts out by default for all user content.
  • We periodically review provider policies and remove providers that do not meet our data-handling bar.
  • If a provider materially changes its data-handling terms, we may switch routing to a different provider — and we will update the in-app disclosure accordingly.

6. Your controls

  • Provider visibility — see the active provider for each generation in-app at render time.
  • Model improvement opt-in — charactr does not use your content to improve its own routing or features unless you opt in. You can change this in Settings.
  • Region-aware routing — where required, generations from EU and UK users are routed through providers and regions that meet applicable data-protection rules.

7. EU AI Act Article 50 compliance

Under Article 50 of the EU AI Act:

  • Article 50(1) — Where users interact with an AI system, charactr ensures the user is informed they are interacting with AI, except where this is obvious from the context. The Service makes its AI nature unmistakable: app naming, onboarding, in-app labels, and output watermarks all communicate that outputs are AI-generated.
  • Article 50(2) — Where charactr generates synthetic audio, image, video, or text content (which is the entirety of what charactr is), outputs are marked in a machine-readable format and are detectable as artificially generated. We embed C2PA-compatible Content Credentials and watermark provenance signatures on shareable exports.
  • Article 50(3) — Where outputs constitute a deepfake within the meaning of Article 3(60), charactr discloses the AI-generated nature; our Content Policy further restricts the categories of deepfake outputs that may be created.
  • Article 50(4) — Disclosure obligations are met for users in the EU at the latest at the time of first interaction or exposure. The disclosures on this page, in-app, and on shared outputs satisfy this requirement.

8. Apple App Review Guideline 5.1.2(i) compliance

Apple's App Review Guideline 5.1.2(i) (effective November 13, 2025) requires apps that share personal data with third-party AI to disclose those third parties to users by name and obtain explicit user permission. charactr discloses:

  • The categories of providers (this page).
  • The active provider for any given generation (in-app, at the point of generation).
  • What data is shared and what is not (this page).
  • The user's controls over data sharing and model training (this page and in-app Settings).

The first time you use a feature that shares data with a third-party AI provider, charactr requests your explicit permission and identifies the provider by name.

9. C2PA Content Credentials

Where supported, charactr embeds C2PA Content Credentials in exported media to communicate (a) that the asset was generated with AI, (b) which charactr generation step produced which part of the asset, and (c) any subsequent edits applied within charactr. C2PA is the industry-standard provenance manifest specification.

10. Adjacent state laws

  • California SB 243 — companion-chatbot disclosure obligations. charactr is not designed as a companion chatbot, but we comply with applicable disclosure obligations where any feature is in scope.
  • New York S.8420-A (effective June 9, 2026) — AI-generated influencer content disclosure. charactr's provenance markers and AI-generated labeling support compliance.
  • California AB 2655 and AB 2839 — election-deepfake disclosure and large-platform removal. The Service prohibits political deepfakes of identifiable real people; see our Content Policy.
  • Tennessee ELVIS Act — voice and likeness rights. The Service prohibits unauthorized voice clones of real people; see our Content Policy.

11. Updates

The provider mix evolves as foundation models improve. Material changes to provider categories, data sharing, or compliance posture are announced in-app and reflected here.

12. Contact

privacy@charactr.com for questions about provider data handling.