d-id-avatar-video

What it does

Instructs the agent to use D-ID for planning, generating, streaming, localizing, and QA of avatar/talking-head videos, covering V2 Photo Avatar Talks, V3 Pro/Instant Avatar Clips, V4 Expressive Avatar Scenes, D-ID Agents, voices, consent, lifecycle, safety, and artifact custody. It emphasizes selecting D-ID routes based on deliverable needs (talks, clips, scenes, agents), and to treat D-ID as presenter-focused rather than a general cinematic generator.

How it works

• Advises choosing the D-ID route before scripting by mapping needs to specific endpoints: V2 Talks (POST /talks), V3 Clips (POST /clips), V3 Instant Avatars, V4 Expressive Avatar creation (POST /scenes/avatars and POST /scenes), or live/live streaming options (Agents/SDK).
• Describes production preflight steps: verify API key and credits, confirm rights and plan constraints, obtain explicit likeness permissions, assess personal data handling, decide between text-to-speech vs external audio, determine custody and storage, and build a QA checklist.
• Details input preparation for source image, presenter selection, and custom avatar/consent footage, including consent verification for V3 Instant Avatars and consent flow for Express Avatars.
• Covers script, voice, and language direction, including available voices, localization, lip-sync guidance, and prioritization of pre-rendered audio when appropriate.
• Outlines API lifecycle: create/select avatar/presenter, provide script, add name/user_data, use webhook, handle result_url, and store payload minus secrets.
• Lists common endpoints and their required fields and potential failures (auth, credits, permissions, moderation).
• Describes real-time streaming with Agents for interactive faces via WebRTC/LiveKit and notes SDK capabilities.
• Enumerates safety and privacy gates and QA pass criteria for finished videos, including consent, non-misleading identity, lip-sync quality, and audio clarity.

When to use it

Use when the selected provider is D-ID or when the user requests a D-ID talking avatar, photo-to-video, presenter clip, expressive avatar, live avatar stream, or localization/training workflow.

What it can touch

Endpoints and fields include: POST /talks, POST /clips, POST /scenes/avatars, POST /scenes, GET /tts/voices, GET /credits, webhook integrations, result_url, source_url, consent_id, presenter_id, driver_id, name, user_data, config, and various script payloads. These are referenced as part of the production and lifecycle steps.

Caveats

• Production preflight requires validation of API keys and credits, rights, consent, and data handling per D-ID terms.
• Trial outputs may be watermarked or restricted to non-commercial use unless plan-enabled; do not promise commercial use without plan authorization.
• Real person likeness requires explicit permission; consent processes are mandatory for custom avatars.
• Lip-sync, pronunciation, and safety checks are to be validated in QA; disclosure and custody considerations must be maintained throughout.