HeyGen

Build a closest honest personal substitute for HeyGen in an empty repository. Use Python 3.12, FastAPI, SQLite, ffmpeg, and a user-owned local TTS model; do not offer alternative stacks. The core loop is: turn user-authored text into a clearly labeled avatar-style video assembled from user-owned media with a local model, never cloning real people, and retain provenance for every output. Make the first run work locally with one documented command. Store all user data locally by default and make export straightforward. Put secrets in .env, ship .env.example, and never commit credentials. Require a project-level rights and consent acknowledgement before generating audio. Ship with no celebrity, public-figure, or scraped voice assets and accept only explicitly licensed models. Generate speech from text with voice, speed, pause, pronunciation, and segment controls. Create a timeline for audio, captions, uploaded visuals, and simple transitions. Embed project metadata and a visible synthetic-media disclosure in exported assets. Store prompts, model identifiers, consent notes, and output hashes in a local provenance log. Include clear empty, loading, success, and recoverable error states. Add input validation, safe filenames, and graceful handling of unavailable APIs. Write focused tests for the core transformation and one end-to-end happy path. Create a README with setup, architecture, permissions, data location, and backup steps. Do not add accounts, billing, telemetry, analytics, or a hosted control plane. Do not claim to reproduce proprietary data, network liquidity, regulated access, or frontier infrastructure. Deliberately leave out cloning a voice without clear consent. Deliberately leave out impersonation or deceptive unlabeled media. Deliberately leave out a frontier avatar model, public hosting, or enterprise rights clearance. Finish by running the tests and listing the exact commands used.