PeterH0323/Streamer-Sales
3,710 estrellas·567 forks·Python·AGPL-3.0·3 vistasopenxlab.org.cn/apps/detail/HinGwenWong/Streamer-Sales↗
Streamer Sales
Streamer-Sales is a digital avatar video generator and large language model sales agent framework. It functions as a multimodal AI pipeline that synthesizes talking-head videos using text-to-speech and face animation to create virtual spokespeople.
The system differentiates itself by combining retrieval-augmented generation with agentic web searching to ground product descriptions and sales scripts in factual data and real-time information. It further integrates a multimodal interaction workflow that coordinates automatic speech recognition and text-to-speech for real-time voice conversations.
The project covers a broad capability surface including optical character recognition for data extraction, synthetic dataset generation for model fine-tuning, and model inference acceleration through precision quantization.
The application is delivered as a dockerized AI suite featuring a FastAPI backend and a Vue frontend.
Features
- Multimodal Response Pipelines - Coordinates ASR, LLM, TTS, and avatar rendering to convert spoken input into visual and auditory responses.
- Interactive Video Avatar Generators - Synthesizes talking-head videos using text-to-speech and face animation to create interactive virtual spokespeople.
- Sales Agent Frameworks - Uses large language models to generate persuasive product pitches and sales scripts.
- Retrieval-Augmented Generation - Employs retrieval-augmented generation with vector databases to ground product descriptions in factual data.
- Multimodal AI Pipeline Orchestration - Coordinates automatic speech recognition, LLMs, and text-to-speech into a unified pipeline for real-time voice interaction.
- RAG Context Retrieval - Retrieves relevant document segments from a vector database to ground AI responses in factual data.
- RAG Grounding Verifiers - Implements grounding mechanisms to ensure sales scripts are factually accurate based on product manuals and web data.
- Speech-to-Text and Text-to-Speech Integrations - Combines speech-to-text and text-to-speech capabilities to enable real-time voice conversations.
- Speech to Text Transcription - Converts spoken audio into text using automatic speech recognition to enable real-time interaction.
- Text-to-Speech - Synthesizes natural human speech from written text using high-fidelity generative models.
- AI Sales Agents - Generates persuasive product pitches and scripts using LLMs to increase customer purchase intent.
- Sales Pitch Generation - Generates persuasive product descriptions and sales scripts to increase customer purchase intent.
- Virtual Spokesperson Orchestrations - Orchestrates animated digital avatars with synchronized speech and video to act as automated brand spokespeople.
- Conversational Avatar Animators - Synchronizes digital character movements and facial expressions with conversational AI input.
- Containerized Deployments - Deploying complex machine learning pipelines using Docker and FastAPI for scalable and consistent environment management.
- Agentic Search Tools - Employs autonomous agents for iterative, multi-step information retrieval and real-time web searching.
- Contextual Information Retrieval - Uses embedding and reranking models to fetch relevant document chunks and ground AI responses.
- Synthetic Media Generators - Produces synthetic video presentations using AI-powered tools to convey information visually.
- Quantized Inference Accelerators - Accelerates LLM response times using precision quantization and caching via LMDeploy.
- Real-Time Web Search Integrations - Allows AI agents to perform live internet searches to ground responses with current data.
- AI Application Deployments - Provides a containerized AI suite with a FastAPI backend and Vue frontend packaged via Docker Compose.
