Video links
Paste a video URL into the chat and Tale ingests its transcript for the agent — supported platforms, the ingestion flow, and what each failure state means.
3 min read
Paste a video link into the chat and Tale fetches the video's transcript so the agent can read, quote, and answer from it — no manual download, no copy-paste of a transcript. It is the fastest way to bring a talk, tutorial, or recorded meeting into a reply.
This page covers the chat's video-link chip. For pasting files rather than links, see Attachments.
A worked example
Paste a YouTube URL into the chat. Tale recognises it as a video link and drops a chip below the message with the video's title and a spinner. Behind the chip, Tale fetches the captions (or, when there are none, the audio, which it transcribes), indexes the transcript, and flips the chip to Ready. Send the message and the agent answers from the transcript, citing the passages it used. A long video keeps indexing in the background; the chip shows its progress and the transcript becomes searchable the moment it finishes.
Supported platforms
Tale ingests links from YouTube (including youtu.be, m.youtube.com, and Music), Vimeo, Dailymotion, Twitch, and Bilibili. A link to any other host is left as ordinary text in your message — no chip appears. Only public videos work; anything behind a login, a paywall, or a region block cannot be fetched.
What Tale extracts
Tale prefers the platform's own captions when they exist — they are exact and cheap to fetch. When a video has no captions, Tale downloads the audio and transcribes it with speech-to-text, so a caption-less upload still becomes a searchable transcript. Either way the result is indexed like any other knowledge: the agent retrieves the relevant passages at reply time and the citations point back to the transcript.
Failure states
The chip turns red when ingestion cannot complete, with a short reason:
Every failed chip carries a Retry, and a retry is safe — Tale never double-indexes a video that already succeeded.
For self-hosted operators
A managed Cloud deployment handles the anti-bot measures for you. If you self-host and video ingestion keeps hitting the bot wall, the deployment ships a proof-of-origin token provider that is wired up by default, and you can add an egress proxy or a pre-warmed session pool as escalation. The configuration lives in Video ingestion.
Where this fits
Video links are a chat-scoped way to ground a reply in a recording, the same way Attachments ground it in a file. Both feed the agent's retrieval; neither persists beyond the chat. To make a video's transcript reusable across chats, copy the ingested text into a Knowledge document.