vocabulary: - term: talk definition: > An asynchronous job that generates an AI talking-head video by animating a source image with a text script (TTS) or pre-recorded audio file. Talks use V2 photo or V3 Pro avatar technology. tags: - Videos - Core Concept - term: agent definition: > A configurable interactive AI entity that combines a digital avatar presenter with an LLM backend, optional RAG knowledge base, and event triggers. Agents support real-time streaming conversations via WebRTC or LiveKit. tags: - Agents - Core Concept - term: session definition: > A real-time streaming connection established between a client and a D-ID agent. Sessions carry WebRTC or LiveKit credentials and may optionally persist chat history. tags: - Agents - Streaming - term: presenter definition: > The avatar configuration attached to a talk or agent, specifying the visual representation (photo, personal clip, or V4 expressive avatar) and the TTS voice settings. tags: - Avatars - Configuration - term: clip definition: > A short training video (V3 Instant) used to create a personal avatar. The clip is trained to produce a V3 Instant presenter that can generate talking-head videos with the subject's likeness and voice. tags: - Avatars - V3 - term: expressive avatar definition: > A V4 full-HD avatar capable of dynamic facial expressions and lifelike motion, distinct from static photo avatars. Used in both standalone video generation and real-time agent streaming. tags: - Avatars - V4 - term: knowledge definition: > A RAG knowledge base attached to an agent. Can store up to 5 documents and is searched at inference time to provide relevant context to the LLM. tags: - Agents - RAG - term: document definition: > A single piece of content (web page, PDF, text) ingested into a knowledge base. Documents are indexed for semantic retrieval during agent chats. tags: - Agents - RAG - term: translation definition: > A video localization job that uses AI speech translation, voice cloning, and lip-sync to re-dub an existing video into one or more target languages. Supports 100+ languages. tags: - Translations - Core Concept - term: transcription definition: > A versioned text representation of a translated video's speech, linked to a translation job. Transcriptions may be selected or swapped to refine the output. tags: - Translations - term: voice definition: > A TTS voice available on the D-ID platform, sourced from providers such as Amazon Polly, Microsoft Azure, ElevenLabs, or Google TTS. Voices have gender, language, and access-level attributes. tags: - TTS - Configuration - term: client key definition: > A scoped API key automatically generated for an agent that permits read-only / session-level operations from untrusted clients (e.g., a browser SDK), without exposing the full account API key. tags: - Authentication - Agents - term: webhook definition: > An HTTPS callback URL provided by the caller. D-ID posts a notification to this URL when an asynchronous job (talk, translation, etc.) completes or encounters an error. tags: - Integration - Asynchronous - term: credits definition: > The unit of account consumption on the D-ID platform. Credits are deducted per video generation, translation, or streaming minute and are tracked via the GET /credits endpoint. tags: - Billing - Platform - term: result_url definition: > The URL of the completed output video or asset. Set once a job reaches done status. May point to a D-ID-managed S3 bucket or a custom destination if a result_url was specified in the request. tags: - Output - Storage - term: stream definition: > A real-time WebRTC or LiveKit media stream used to deliver live avatar video during an agent session. Streams are created per-session and torn down on disconnect. tags: - Streaming - Agents