Product Demo and Live Sales¶

This case is for e-commerce guides, showroom explainers, course narration, and live interactive demos. Unlike customer support, the content is usually more controlled by scripts, product facts, and operator timing. It benefits from stable speaking style, long-form narration, and manual interruption.

Suitable Scenarios¶

A digital guide on a product page or live commerce room.
Product explainers on showroom screens.
Course, news, or event narration.
Operator-prepared scripts spoken by a digital human.

Recommended Pipeline¶

flowchart LR
    Operator[Operator console]
    Script[Product facts / script]
    API[OpenTalking API]
    LLM[LLM rewriting and Q&A]
    TTS[TTS voice]
    Avatar[QuickTalk / Wav2Lip / FlashTalk]
    Audience[Audience browser]

    Operator --> Script
    Script --> API
    Operator -->|Questions / interruption| API
    API --> LLM
    API --> TTS
    API --> Avatar
    API -->|WebRTC| Audience

Prerequisites¶

Finish AI Customer Support or at least validate mock.
Prepare a display-friendly avatar. See Custom Avatar.
For real lip-sync video, deploy at least one talking-head backend in Model Deployment.

1. Prepare Product Facts¶

Keep product facts short and structured:

product-brief.txt

Product: Smart Conference Camera Pro
Audience: remote meetings, online courses, small live rooms
Key benefits:
- Automatic speaker framing
- Dual-microphone noise reduction
- USB plug-and-play
Limits:
- No waterproof guarantee
- Not a replacement for professional cinema cameras
Promotion:
- Use the live business system as the source of truth

2. Configure the Persona¶

.env

OPENTALKING_LLM_SYSTEM_PROMPT=You are a live product-demo digital human. Speak naturally and rhythmically. Keep each answer under 80 Chinese characters when speaking Chinese. Explain user benefits before features. For price, inventory, and after-sales policies, rely on the business system and do not invent facts.
OPENTALKING_TTS_DEFAULT_PROVIDER=edge
OPENTALKING_TTS_EDGE_VOICE=zh-CN-XiaoxiaoNeural

If you already have a product retrieval service, let the business layer assemble the relevant facts before sending text to OpenTalking.

3. Start and Select a Model¶

Validate script pacing with mock first:

terminal

bash scripts/quickstart/start_mock.sh

For a lightweight local path:

terminal

bash scripts/start_unified.sh --backend local --model quicktalk

For a remote high-quality path:

terminal

bash scripts/start_unified.sh --backend omnirt --model flashtalk --omnirt http://127.0.0.1:9000

4. Design Interaction Modes¶

Input type	Handling
Prepared narration	Send fixed scripts or segmented copy from the operator console.
User questions	Answer through the LLM/Agent, with product retrieval when needed.
Manual interruption	Stop the current narration and switch to the new segment or answer.
Risky questions	Return fixed fallback copy for pricing, contracts, medical, or financial claims.

Validation¶

Long text is split into natural segments.
The TTS voice fits the brand or character.
Avatar and model type match; /models reports the real model as connected.
Old product narration does not continue after interruption.
Dynamic price, inventory, and promotion facts come from the business system.

Troubleshooting¶

Symptom	Action
Speech sounds mechanical	Shorten each text segment and make the prompt more conversational.
The LLM drifts away from the product	Pass structured product facts from the business layer and require grounded answers.
First frame is slow	Warm up the real model or create the session before the live segment starts.
The live room needs concurrency	Use API/Worker split and external Redis; see Deployment.