AI Disclosure

RiffAds is an AI tool for making video ads. It is run by [Company legal name] ("we" or "us"). The content RiffAds produces is made or changed by AI models, and we want you, and the people who watch your ads, to know that.

This page explains where we use AI, what is synthetic, how outputs are marked today, and what you must do before you publish. Many laws and platform rules now cover AI generated content. One example is the EU Artificial Intelligence Act, whose transparency rules for AI generated and AI edited content apply from 2 August 2026.

This page is part of our Terms of Service. Please read it together with our Acceptable Use Policy and our Privacy Policy.

When you use RiffAds, AI models create or change the content. This includes:

• Videos: talking actor videos, where AI animates a face so it speaks a script or an audio track, and videos made from text or images.
• Images: images made from a text prompt, and images changed by AI, for example edits, upscaling, background removal, or showing a product held or worn by a person.
• Voices and audio: speech made from a script, designed voices, cloned voices, changed voices and music.
• Scripts and text: scripts, captions and other text that our AI tools write, improve, translate or transcribe.

This is true even when you write the script yourself or upload your own files. The result (for example an AI actor speaking your script, or a photo turned into a video) is AI generated or AI edited content.

Treat every file you get from RiffAds as AI generated or AI edited content.

The actors in the RiffAds library were created with AI. They are not real people. Their faces and bodies are synthetic, and any likeness to a real person is by chance. Their speech is also made by AI, using voices from our voice provider (see "AI voices and voice clones" below). To confirm: that every library actor was made with AI and not from a photo or video of a real person. This includes the first seed actors (Maya and Leo), whose files have no recorded origin in the code, and any actor an admin publishes to the library from a workspace, which the publish path does not check.

When an AI actor appears in your ad, the words, the voice, the facial expressions and the movements are all made by AI. No real person said your script or approved it.

An AI actor is not a real customer, reviewer or expert. It has never used, tested or recommended any product. Do not present it as a real person who has.

Custom actors. You can also create your own actors. You can describe a person in words (and add reference images), and AI makes a portrait. Or you can start from a photo. If you start from a photo of a real person, the actor will look like that person. A reference image of a real person can also make the actor look like them. Every video made with such an actor shows that person saying words they never said. You need that person's written consent first (see "Real people must agree" below).

The speech RiffAds creates is made by AI speech models. There are three kinds of AI voices:

• Library voices. These come from our voice provider, ElevenLabs. Some are ElevenLabs' own default voices, and some were shared by their owners in the ElevenLabs voice library. The speech in your ad is made by AI: the voice owner never said your script. Our process for choosing shared library voices leaves out voices in the ElevenLabs "famous" category. A shared library voice can stop being available if its owner withdraws it.
• Designed voices. You describe a voice in words (when you create an actor, AI can suggest the description), and AI creates a new synthetic voice from that description. No voice sample is uploaded.
• Cloned voices. You upload a sample of a real voice, and AI creates a voice that sounds like it. Before you clone a voice, you must confirm that you have the right to clone it and that you will not use it to impersonate anyone without their permission. We keep a record of that confirmation.

You can also upload or record your own audio instead of using an AI voice. AI then moves the actor's lips to match it, so the video is still AI generated.

To use a real person's face, likeness or voice in RiffAds (other than your own), you must have their written consent first. This applies when you create a custom actor from a photo or reference image, clone a voice, or upload images, video or audio of someone.

Do not make content that shows a real person, including a public figure, saying or doing something they did not say or do, without their written consent. The full rules are in our Acceptable Use Policy.

RiffAds does not verify who is in a photo or voice sample, or whether you really have their consent. That responsibility is yours.

Besides making content, RiffAds uses AI in these ways:

• Assistants. Some tools talk with you, for example the chat that helps you describe and design a new actor. When you use them, you are talking with an AI model, not a person.
• Writing help. AI can draft scripts, improve prompts, suggest how a line should be spoken, and translate a script before it is spoken.
• Imports. When you give us a link to a web page, AI reads the page to help pull out product and brand details, such as claims, tone and audience.
• Workflows. AI can build a workflow from a short brief, or fill in a workflow's scripts, prompts and settings from a short instruction.
• Actor setup. When you create an actor, AI looks at its image (for example a photo you chose) and suggests details for the actor, such as gender, an age range, ethnicity, a short description and a first name. It also suggests a voice for the actor. These are guesses made by a model, not facts about any person. To confirm: this step infers ethnicity and an age range (including under 18) from an image that may show a real person. The EU AI Act has banned, since 2 February 2025, biometric categorisation systems that use biometric data such as face images to infer a person's race, and this step may fall under that ban. Get legal advice before launch, including on privacy law, and decide whether to remove the ethnicity field.
• Safety checks. Automated tools, some of which use AI, check scripts, prompts, uploaded images and voice clone samples before they are used.

We do not build or train our own AI models. RiffAds uses AI models from other companies, which we reach through these providers:

• fal.ai, a platform that runs media models from many AI companies. We use it for tasks such as talking actor video, text or image to video, text to image, image editing, upscaling, background removal, virtual try-on, music, voice changing and speech to text for captions. We also use it for some basic editing steps, such as trimming and joining clips.
• ElevenLabs, for text to speech, voice design and voice cloning.
• Deepgram, to turn voice clone samples into text so we can check them.
• OpenRouter, a service that connects to large language models from several AI companies. We use these models for writing help, imports, workflows, assistants, actor setup and safety checks.

The models behind each tool can change at any time, for example when a better model comes out or an old one is retired. Our Privacy Policy explains what data we send to these providers.

When a generation finishes, we record its output files as AI generated in our own database and link them to that generation. This record stays in our systems. It is not added to the files.

Today, RiffAds does not add a visible "AI generated" label, a watermark, C2PA content credentials or any other provenance data to the files you download. Some of the models we use may add their own hidden watermark or metadata. We do not add, check or promise these marks, and editing steps (such as adding captions, resizing or joining clips) may remove them.

A file you download from RiffAds may carry no sign that it was made with AI. If a label is needed, you must add it.

Do not remove or hide any AI marking or provenance data that a file carries in order to hide that it was made with AI. If we start marking outputs, we will update this page. To confirm: whether and when RiffAds will add machine-readable marking to outputs (for example C2PA content credentials or a watermark). The EU AI Act asks providers of AI systems that generate synthetic audio, images, video or text to mark outputs in a machine-readable way from 2 August 2026. Update this section when marking ships.

You decide where your content goes, so you are responsible for labeling it. Where a law, a regulator or a platform's rules require it, you must clearly tell viewers that your content is AI generated or AI edited. These rules include, for example:

• the EU Artificial Intelligence Act, which requires anyone who uses AI in their work to make deepfakes (images, audio or video that look like real people, objects, places or events and could falsely seem real) to disclose that the content was made or changed by AI;
• other laws on deepfakes, synthetic media, political ads, and the use of a person's name, image or voice;
• consumer protection and advertising laws, including rules on fake reviews, testimonials and endorsements;
• the advertising and content policies of each platform where you run ads or post content, some of which require a label on realistic AI content.

Take extra care with talking actor videos. A realistic person speaking to the camera is exactly the kind of content these rules are about. When in doubt, label it.

A label does not make misleading content lawful. The rules on testimonials and endorsements differ between countries and platforms, and in some places a label is not enough.

Use a label viewers will notice, for example a clear "AI generated" or "Made with AI" note on the ad or in its caption, or the platform's own AI label. These rules change often and differ by country and platform. This page is not legal advice, so check the rules that apply to you.

AI makes mistakes. Content from RiffAds can contain errors, for example:

• words that are wrong, missing, mispronounced or badly translated;
• lips that do not match the audio, or strange faces, hands, text or objects;
• product details, prices or claims that are not true, including in scripts and import details written by AI;
• content that looks like a real person or a real brand by accident.

We do not review your content by hand before you receive it. Our automated safety checks can miss things, and they do not check the finished output. Watch, listen to and read every output before you use it, and make sure every claim in your ad is true and allowed. You are responsible for the content you choose to publish.

RiffAds does not post, publish or send your content to any social media or advertising platform. You get files and download links, and you decide what to do with them.

This means the choice to publish, and the duty to label, stay with you. It also means we cannot remove content that you or anyone else has published somewhere else.

You can connect AI agents to RiffAds through our API, our MCP server, our command line tool (CLI) and Agent Skills. An agent acts for you. Everything on this page applies to content an agent makes: it is AI generated, and the same consent and labeling rules apply.

• You are responsible for every action and every charge made with your account, API keys, MCP connections and agents, including actions an agent takes without asking you first.
• Spend limits can help control what agents spend, but you are responsible for setting them. Charges from agent calls are final, like any other charge. See our Refund Policy.
• An AI assistant you connect is run by its own company, under its own terms. That company may see your prompts, your scripts and the download links we return. This is one reason download links given to agents expire after a short time.
• If your agent publishes content somewhere else, you must make sure it is labeled where required. RiffAds itself never publishes.

For questions about this page or how RiffAds uses AI, email hello@riffads.com. For questions about personal data, email hello@riffads.com.

To report misuse, email hello@riffads.com. For example, tell us about content you believe was made with RiffAds that uses your face or voice without consent, shows a real person saying something they never said, or is not labeled as AI generated where it should be. Please include a link to or a copy of the content, where you found it, and why you think it breaks the rules.

We look into reports and may take action where we find a breach of our Terms of Service or Acceptable Use Policy. Because we do not publish content, we cannot remove it from other websites or platforms. To get it taken down there, please also report it to that platform.

AI tools, AI models and the rules around them change quickly. We will update this page when the way RiffAds uses AI changes. The date on this page shows when it last changed.