Intel / article

Processing Images And Documents In Elevenagents

Source: elevenlabs.io
Author: ElevenLabs editorial sitemap
Date: Published Sep 18, 2026

Why it matters

Clarifies which channels accept image and PDF input (web, mobile, WhatsApp) and how files enter model . Useful when designing agents that must handle photos and documents inside a voice or chat thread.

Key takeaways · AI-distilled

ElevenAgents passes images and PDFs to the model as native file references bound to the user's turn rather than text summaries, while voice notes, location pins and contact cards are normalized to plain text in the transcript.
On web and mobile, file input needs file_input enabled plus a vision- and document-capable model, and the upload must happen before the message: a message sent without the returned file_id gives the model no reference to the file.
On WhatsApp, ElevenLabs downloads customer files from Meta server-to-server, so the integrating app never holds the file and can retrieve it only through the file_url in the post-call webhook.
Nothing carries across conversations automatically. The post recommends storing post-call webhook data against a customer identifier and injecting it as dynamic variables next session; signed file URLs expire after 15 minutes, with the GET conversation API as a fallback.

Terms in this piece · Glossary

multimodal — A model that works with more than text — reading images, audio, or video, and sometimes generating them too.
context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.

Read the source elevenlabs.io