AI Talking Avatar Twin Creation

I want to build a true “digital twin” that can turn my ideas into polished talking-avatar videos that look and sound like me. The core job is to train an AI model on two data sources I’ll supply:
• a folder of past social-media posts and longer articles so the language stays in my voice
• scripts, recordings, and video transcripts so the avatar’s delivery, tone, and cadence match mine

Using that material, the system should let me type (or paste) a script and receive a rendered video where a photoreal or stylised avatar of me speaks the words naturally. Lip-sync, facial animation, and cloned voice quality must feel authentic enough that viewers recognise it as me, not a generic model.

I’m open to whichever toolchain you prefer—e.g., voice cloning with ElevenLabs or similar, avatar generation with D-ID, HeyGen, or a custom Unreal/MetaHuman rig—as long as the final workflow is streamlined: upload or write text ➜ get downloadable video in HD. A lightweight web interface or a local desktop prototype is fine for the first version.

Deliverables
1. Trained voice model and avatar files (with usage rights transferred).
2. A working MVP that outputs talking-head MP4/Mov videos from user-supplied text.
3. Setup guide so I can reproduce the build or move it to my own cloud account.

Acceptance criteria: the finished clip must keep lip-sync within 1–2 frames, pass a basic TTS detect test, and preserve my signature phrases pulled from the supplied posts. I’ll provide feedback rounds with sample scripts to fine-tune the look and sound before sign-off.