Content AI FAQs
All your questions, answered.
Frequently asked questions
In order to maximize the quality and effectiveness of our Content AI features, you will get full Content AI product for a 2 month period with fair usage caps to help us evaluate and make the appropriate improvements.
Yes, this program will help us identify bugs and issues while we look to improve the quality and user experience. We are actively investing into this area which means we will take your feedback very seriously and will seek to resolve issues fast.
We have no official date as of now, but we're aiming to roll things out throughout Q1 and Q2 2025.
Use of your data is limited to your bespoke Turtl Content AI models, which means that no other model could ever access it. All of the core Content AI infrastructure is deployed on our AWS, which is already listed as a secure vendor in our ToCs. For generative AI media, we use Replicate. This functionality can be avoided if registering Replicate to the list of vendors is problematic for any reason.
We’re working with two distinct features: image generation and document content manipulation. Here’s how each is handled:
For image generation: We use a service called Replicate, which lets us fine-tune a model called Flux using customer images. If customers choose, they can have their own fine-tuned models created. Importantly, these models remain within our system and aren’t shared outside of Replicate.
For content manipulation: We aren’t training any models for this feature. Instead, we use an existing AI model that operates entirely within our secure AWS infrastructure. Your document content is not used to train models—now or in the foreseeable future. If this ever changes, it will be on an opt-in basis, and you will be informed beforehand. You content will never be used to train generative models without your explicit consent.
Yes! You have full ownership of any images generated through the system. This includes images produced using fine-tuned models based on your data. However, the underlying AI model itself (think of it as the “engine”) is licensed to us and cannot be redistributed or sold commercially. You get all the benefits of the output, but we can’t sell or share the tool that generates it.
All AI generated images can be used for commercial purposes, including outside of Turtl.
Any media provided by the customer will not be used by Turtl for any purpose other than generating new media for the customer.
When it comes to media as a part of Content AI, there are two primary options of finding / sourcing the right image for your content:
Find the best match in your Turtl media libraryOur AI finds top 4 best image matches in your media library for your content, or just by manually describing what kind of image you’re looking for. The prerequisite for this is to upload a bank of pre-approved images into the Turtl media library. This feature works entirely within the AWS environment and does not require any additional subproccessors.
Generate new images with AI based on your bespoke AI modelThe first step is to upload a bank of pre-approved photos / icons / illustrations into the Turtl media library. It is important that you own commercial rights to those images. Second step is to select between 20-30 images of a similar style to train your very own bespoke AI model. You can train multiple bespoke models, one for icons, one for photos, one for banners, for various brands and so on. From then on, our AI will generate new images in the style of training images of a particular model.
For example, asking our AI the following “Generate a photo of marketers running a workshop using sticky notes and drinking coffee” will either find the best match in your existing media library, or generate a new image in the exact style of your bespoke AI model, depending on your preference.
#8 What LLM powers Turtl's translation capabilities?
Turtl leverages Anthropic's Claude LLM through their official API for all translation needs.