Intel / article

Talk To A Statue Building A Multi Modal Elevenagents Powered App

Source: elevenlabs.io
Author: ElevenLabs editorial sitemap
Date: Published Oct 9, 2026

Context

A photographed monument becomes a conversational character through several distinct AI steps in ElevenLabs' statue-app tutorial. An OpenAI vision model identifies the artwork and depicted figures, while web search can supply historical . Voice Design generates voices from written descriptions, and a voice uses them during conversation. Because those voices come from descriptions rather than recordings, they are creative interpretations of historical figures. The example shows how image recognition and generated speech can support an interactive visitor experience.

Key takeaways · AI-distilled

The vision step asks the OpenAI model to return, alongside the statue's name, location and artist, a ready-made Voice Design description for each character (age, gender, accent, pacing, personality), so identification feeds voice generation directly.
Voice description tips from the post: open with an audio quality marker, say "thick" rather than "strong" for accent strength, avoid vague words like "foreign", and for historical figures borrow a real-world accent as inspiration.
Voice Design is a two-step flow: generate previews from a description plus sample text, then save one as a permanent voice. The post says sample text of 50+ words matched to the character gives more stable results than a generic greeting.
Multi-voice agents work by listing extra voices in a supportedVoices array; per the post, the Agents platform routes each response segment to the right voice when the indicates a different character is speaking.
To avoid a stiff interview feel, the tells the model it plays all characters at once and should have them interrupt, react to each other and hold side conversations, so the user feels like they are joining a real exchange.

Terms in this piece · Glossary

context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
LLM — A large language model — the neural network behind tools like Claude and ChatGPT, trained on huge amounts of text to predict what comes next.
system prompt — The standing instructions a model receives before any user input — defining its role, rules, tools, and tone for the whole conversation.

Read the source elevenlabs.io