# Your custom image model, trained on your brand Ideogram publishes a public search API over its indexed pages; the guide is at https://ideogram.ai/agents/agent-access. Source: https://ideogram.ai/features/custom-models/ Source observed: 2026-11-08T15:45:19.060Z Last checked: 2026-11-08T15:45:19.06+00:00 Indexed: 2026-11-08T15:46:42.982+00:00 --- # Your custom image model, trained on your brand We let teams train a custom image model on their own visual identity. Upload **15–100 on-brand images** representing a specific style; new generations from the custom model follow the supplied art direction, typography, and visual identity. [Start training](https://ideogram.ai/models) ## What custom models can preserve We show example models for book illustration, custom lettering, photography style, poster design, character design, and optical illusion. The examples include: - A book-illustration model trained on 27 illustrations from *The Tale of Peter Rabbit* by Beatrix Potter. It retains soft English watercolor washes, delicate pen linework, anthropomorphic character proportions, pastoral garden compositions, muted earth tones, and countryside warmth. - A custom-lettering model that maintains consistent lettering weight, color palette, and decorative character across generations. - A photography-style model trained on antique photographic processes from the 1850s through the 1920s, reproducing period grain, hand-colored tinting, and tonal compression. - A poster-design model trained on vintage poster art from the 1890s through the 1940s, capturing flat lithographic color, period display typography, and bold compositional hierarchy. - A character-design model trained on 15 images of a specific 3D character style, maintaining proportions, matte materials, and clean white backgrounds. - An optical-illusion model trained on 22 hand-crafted optical artworks, rendering prompted subjects as intricate illusions. A comparison uses the same prompt across nine custom models: “Woman seated in bistro chair, coffee cup in hand, Paris street with Eiffel Tower in background. Text ‘Paris is always a good idea’ is displayed across the top.” These are illustrative examples of different custom-model outputs. ## The problem ### A model trained on the internet does not know your brand Off-the-shelf models blend billions of unrelated images. A custom model is fine-tuned on your assets to learn your style, visual identity, and aesthetic choices. You can feed strong outputs back into training and train a different model for each expression of your brand. ## Why Ideogram ### Your LoRA is only as good as its base model Open-source LoRAs inherit the limitations of their open-source base models, including inconsistent text, generic composition, and limited style range. We describe Ideogram as a foundation built for custom-model training, with a focus on legible text rendering and enterprise workflows. ## Enterprise customization [Talk to our team about advanced customization](https://ideogram.ai/features/custom-models/#). Enterprise options include: 1. **White-glove setup and training:** access to research, engineering, and annotation teams from asset sourcing through deployment. 2. **Large datasets, unlimited iteration:** train on thousands of images with custom captioning pipelines and iterate until the model matches your standards. 3. **Custom typography and exact logos:** render brand marks as layerized text and SVG for on-brand identity. 4. **Your data, your model:** Enterprise training data and models are not used to improve our foundation models; the model and data are yours alone. The page also shows these figures and customer references: - **100K+** branded concepts generated by [Skechers](https://ideogram.ai/blog/skechers-case-study/). - **30+** custom models trained for enterprises. - **3B+** designs generated on Ideogram. - **30M+** creators, designers, and marketers. - Customer logos shown: Netflix, Adobe, Skechers, Gamma, Wix, Printify, and Canva. ## How it works: Upload. Train. Generate. No infrastructure or GPU provisioning is required from you. 1. **Upload images:** Select the Models tab, click **+** to create a model, and upload 15–100 images of your brand, subject, or style. 2. **Train your model:** We auto-generate captions. Review and refine them, then start training. Quality matters more than quantity. 3. **Generate at scale:** Select your Custom Model and start prompting. ## API support You can create datasets, upload training images, start training, and generate at scale from your own infrastructure without using the UI. 1. Create a dataset and upload 10–100 images. 2. Start training with a single POST request. 3. Generate images using `custom_model_uri`. [Read the API docs](https://developer.ideogram.ai/tutorials/custom-model-training) · [API reference](https://developer.ideogram.ai/) ```python # Generate with your trained model response = requests.post( "https://api.ideogram.ai/v1/ideogram-v3/generate", headers={"Api-Key": API_KEY}, data={ "prompt": "A photo in my custom style", "custom_model_uri": "model/my-brand/version/1", } ) ``` ## Custom Models for teams On the team plan, train once and share the model across your organization. The page describes: - Any team member can train and share; there is no gatekeeper. - Unlimited models, with no cap, for uses such as product photography, social content, and print campaigns. - Models can be trained from shared collections of approved assets. - Generated images can be shared through shared collections. [See the team plan](https://ideogram.ai/pricing/) ## Frequently asked questions ### How many images do I need? Self-serve supports 15–100 images. Larger datasets require Enterprise training. Complex concepts and styles may need more than 15 images; the page emphasizes that quality matters more than quantity and says 40 exceptional images with descriptive captions can outperform a larger mediocre set. ### What can I train a custom model on? Products, characters, mascots, illustration styles, brand aesthetics, photography styles, custom typography, and graphic-design lettering styles can be trained when the visual identity is consistently represented across images. ### What if the model cannot generate certain scenes? An overly constrained training set can cause overfitting. That may be useful for a specific character or optical illusion, but desired scenes and compositions should be represented in the training data because the model can reproduce only what it has seen. ### Is training data used to improve Ideogram’s base models? On Pro and Team plans, data may be used to evaluate and improve Ideogram’s foundation models. On Enterprise, we say that training data and custom models will not be used for AI training or shared outside the organization. ### Can I share a custom model? Team members can share trained models with the organization on Team plans. Pro and Team users can also share models publicly with the broader Ideogram community. Team-generated images can be shared through shared collections. ### Can I train on a custom font or logo? Enterprise custom models support custom typography and exact logo placement as layerized text and SVG. Enterprise also includes research, engineering, and annotation support, larger datasets with custom captioning pipelines, and unlimited training runs. [Learn about Enterprise features](https://ideogram.ai/features/custom-models/#enterprise) or [talk to our sales team](https://ideogram.ai/features/custom-models/#). ### How does Enterprise differ from self-serve? Enterprise includes on-site visits, embedding with design teams, PLM and DAM integration, vendor material digitization, and ongoing iteration with our team. It is built for brands generating at significant scale. ### How do I access Custom Models? Custom Branded Models are available on [Pro and Team plans](https://ideogram.ai/pricing/) and through the [API](https://developer.ideogram.ai/tutorials/custom-model-training). For advanced customization, contact us about [Enterprise](https://ideogram.ai/features/custom-models/#enterprise). --- [Read this page’s summary](https://ideogram.ai/agents/api/public/v1/page-tldr?url=https%3A%2F%2Fideogram.ai%2Ffeatures%2Fcustom-models%2F&trace=t_380edc68d53d) ## Ask a follow-up about this page Need a more specific answer? Ask the exact question you have. It returns a fresh, cited answer from Ideogram’s knowledge base. No account or API key required. Open this URL with your own question, URL-encoded. Include your question in q. All research inputs are optional; no account or setup. `GET https://ideogram.ai/agents/agent/faq?q={question}&trace=t_380edc68d53d` [Tool guide and examples](https://ideogram.ai/agents/agent-access) · [OpenAPI](https://ideogram.ai/agents/api/public/v1/openapi.json) · [Page TLDR form](https://ideogram.ai/agents/agent-tools/page-tldr?url=https%3A%2F%2Fideogram.ai%2Ffeatures%2Fcustom-models%2F). Opening a form does not execute a tool.