# Train a Custom AI Model on Your Brand | Samsa > Train a custom AI model on your brand’s style, products, people or scenes. One image is enough to start, and training takes about 15 seconds. Human version: https://samsa.ai/train/ Language: English Machine index: https://samsa.ai/llms.txt --- **Train AI on Your Brand** ## Turn your brand guidelines into a living AI model. With Samsa you train a custom AI model on your brand from your own reference images: your style, your products, your people or your scenes. One image is enough to start, and 2–5 give the best results. Training takes about 15 seconds, is free and unlimited on every plan, and the model stays private to your organization. [Get started](https://app.samsa.ai/signup) [Get a demo](https://samsa.ai/get-demo/) *(image: Train Generative Ai Branding)* --- Trusted by 250+ brand and marketing teams in Switzerland SRK · SPS · Michel · Switzerland Tourism · Hertz · Klett · Vontobel · PostFinance · Focus Water · Sanitas · Trisa · Rivella · Capri-Sun · Stadler Rail · Marché · Darvida · HUG · Wernli · Transa · Bank WIR --- ## One training. On-brand content whenever you need it. Upload a few reference images and Samsa learns your visual identity – ready for anyone on your team to generate perfectly consistent content. *(embedded: vimeo Video Player)* https://player.vimeo.com/video/1151236742?color&autopause=0&dnt=true&loop=0&muted=0&title=0&portrait=0&byline=0#t= - No design skills needed - New team members productive in minutes - Every output follows your brand guidelines --- ## Your brand, your AI models Styles ### Your visual language, ready to scale Train Samsa on your illustration style, photography aesthetic, or design system – then let anyone create assets that look like they came from your best designer. - Consistent creation with guaranteed brand fidelity - From illustrations to photography – any style works - No more briefing designers for routine visuals [Get started](https://app.samsa.ai/signup) *(image: Styles Generative Ai)* *(image: Products Ai Models)* Products ### Product shots without the photoshoot Generate campaign-ready product imagery in any context – no studio, no photographer, no weeks of waiting. - Photorealistic product renders in any setting - Perfect for ads, social media, and presentations - Update campaigns in minutes, not months [Get started](https://app.samsa.ai/signup) People ### Consistent faces across every touchpoint Whether it’s your CEO, brand ambassador, or virtual spokesperson – maintain recognition across all your communications. - Key executives available for any content need - Virtual models for ads and presentations - Real people or brand characters – both work [Get started](https://app.samsa.ai/signup) *(image: People Ai Models)* *(image: Scenes Generative Ai)* Scenes ### Your signature spaces, anywhere you need them Generate your branded office, store, or any signature environment for consistent visual storytelling – without booking locations or travel. - Branded environments on demand - Signature locations that reinforce identity - Perfect for corporate communications and campaigns [Get started](https://app.samsa.ai/signup) --- ## Frequently asked questions: train a custom AI model on your brand ### What can I train a custom AI model on? You can train a custom Samsa model on four things: your style (an illustration, photography or 3D look), your products, your people and your scenes, such as your office or your store. Each becomes its own model that your team selects when it generates, so new images and videos keep your look, your products and your faces consistent. The help center has a guide per model type: [train AI models](https://help.samsa.ai/en/collections/3-train-ai-models). ### How many images do I need to train a model? One reference image is enough to train a Samsa model, and 2–5 give the best results, for every model type. Choose images that share the characteristics the model should learn, and leave out any that break the pattern: in a style set, one image in a different look pulls the model off course. What to prepare, step by step: [How to train AI on your brand guidelines](https://samsa.ai/guides/train-ai-on-brand-guidelines/) ### How long does training take? Training a custom model in Samsa takes about 15 seconds. You upload your reference images, start the training and can generate with the new model right after. Because training is that fast and free, you can train a new model with a better set of images whenever your brand, product or campaign changes. ### What does training a model cost? Nothing. Model training is free and unlimited on every plan, from Starter to Enterprise. You only use credits when you generate with a model, for example 5 credits for an image at 1K resolution. All plans and credit rates are on the [pricing page](https://samsa.ai/pricing/). ### Can Samsa learn any style? Yes, as long as the style is consistent. Samsa learns the characteristics of the style in your reference images, for example: - Illustrations - 3D visualizations - Abstract styles - Photorealistic styles with a specific color grading - Cinematic styles - Paintings and watercolors - Geometric structures What matters is that every image in the training set shares the same stylistic characteristics. If one image breaks ranks, remove it. ### Can I use my trained models for videos too? Yes. You can use your trained models for video generation too, so your clips match your brand as your images do. Video uses credits per second, depending on the model you choose, from 8 credits for a 5-second clip; a clip runs from 3 to 15 seconds. ### Do my models and images stay private? Yes. Your uploaded images train only your own models, and your models are locked to your organization: Samsa never uses your data to train or optimize other customers’ models. Your data is stored in the EU, with our primary data center in Frankfurt; some AI processing is done by providers outside the EU.