The things we talk about

Producing images from text descriptions or reference images using diffusion or transformer-based models. Euryka's Imaginations brings 10+ image models — Flux, Stable Diffusion, Midjourney via Sparks — into one interface with shared brand context, image-to-image remixing, and inpaint editing.

Coordinating multiple AI models — text, image, video, voice — from a single workspace so creative teams can move between modalities without breaking context. Includes routing the right prompt to the right model, persisting brand state across calls, and aggregating outputs into one project.

Generating short video clips from text prompts, images, or other videos using models like Kling, Veo, Sora, and Runway. Euryka exposes seven video models in one place so creative teams can match the right model to the right shot without juggling separate accounts.

Optimising websites to be cited by AI answer engines — Google AI Overviews, Perplexity, ChatGPT search — rather than only ranked in classical search results. AEO requires structured data, answer-shaped content, and entity binding so AI engines can reliably identify the page as a citation source.

The consistent personality, tone, and language a brand uses across every channel and asset. In AI workflows, brand voice is encoded as guardrails the model checks against — preventing off-brand outputs at generation time rather than catching them in review.

Claude is a family of large language models developed by Anthropic, first released in March 2023. Trained using Constitutional AI for improved safety and alignment, Claude models are available in three sizes—Haiku, Sonnet, and Opus—and support text, coding, reasoning, and image-input tasks.

The discipline of feeding AI models persistent, structured context — brand guidelines, approved assets, audience knowledge, prior outputs — so every generation stays on-brand without manual re-briefing. Successor to prompt engineering: less about crafting one perfect prompt, more about maintaining the state the model reasons against.

Gemini is a family of multimodal large language models developed by Google DeepMind, succeeding LaMDA and PaLM 2. Announced December 6, 2023, it processes text, images, audio, video, and code natively, powering Google's Gemini chatbot across sizes like Ultra, Pro, Flash, and Nano.

AI that produces new content — text, images, video, audio, code — rather than just classifying or analysing. Generative AI is the umbrella category Euryka operates inside; the platform makes it usable for marketing teams by adding brand context, governance, and multi-model orchestration on top.

Image-to-video is a generative AI technique that animates a static image into a short video clip, using diffusion models to predict plausible motion, camera movement, and temporal progression from that image, often guided by an accompanying text prompt describing the desired action or style.

A type of AI trained on vast amounts of text that can read, write, and reason in natural language. LLMs power Euryka's Threads, Docs, and Sparks — GPT, Claude, Gemini, and others all sit behind the workspace, routed transparently based on the task.

Luma AI is a generative artificial intelligence company that builds multimodal models for creating photorealistic images, videos, and 3D content from text, image, or video prompts. Founded in 2021 and based in Palo Alto, its flagship platform, Dream Machine, uses models like Ray and Uni for creative production.

Model Context Protocol (MCP) is an open standard introduced by Anthropic in November 2024 that provides a universal interface for connecting AI systems, such as large language models, to external data sources, tools, and services using JSON-RPC 2.0 messaging, replacing fragmented custom integrations with a single interoperable protocol.

AI models that work across multiple input and output types — text, image, audio, video — within a single pipeline. Multimodal AI is what lets Euryka turn a written brief into an image, then a voiceover, then a video, all carrying the same brand context end to end.

Node-based programming is a visual programming paradigm in which a program is constructed by connecting discrete functional units called nodes within a graph. Each node has inputs and outputs; edges between nodes represent data or control flow. It is widely used in 3D graphics, audio, machine learning, and IoT tooling.

The practice of crafting input prompts that get reliable, high-quality outputs from AI models. Foundational discipline that Euryka content positions as the predecessor to context engineering — useful for one-off tasks, insufficient for sustained brand work where the same instructions get repeated to every model.

Retrieval-augmented generation combines a retriever — vector search over an indexed corpus — with a large language model so answers are grounded in cited source documents instead of the model's training data. RAG cuts hallucinations and lets answers stay current as your corpus updates.

A text-to-video model is a form of generative artificial intelligence that uses a natural language description as input to produce a video relevant to the input text, with recent advances driven largely by video diffusion models and diffusion transformer architectures.

Synthesising speech in a specific person's voice using a short audio sample as reference. Powers Euryka's Voiceovers product — record a voice once, generate unlimited voiceovers in that voice across languages, all carrying the original speaker's tone and timbre.

Workflow automation is the use of technology to execute a sequence of business tasks or processes automatically, following predefined rules, with minimal human intervention. It connects applications, routes work between steps, and reduces manual effort in repetitive, rule-based processes across departments or systems.