Automated faceless short-form video factory rendered as a 3D system

GOAL: Build me a faceless channel factory on this machine. MOTION: One topic in, one finished long-form video and 6 vertical shorts out, scheduled to YouTube, TikTok and Reels. Use only free open-source repos. Write every step as a script I can rerun with one command. NICHE: [your niche, e.g. 1930s noir cartoon animals explaining money] STYLE: one locked visual style for the channel (pixel, paper, clay, toon, voxel, neon or noir). Same character every video. THE FACTORY (9 rooms, one conveyor, each room is one folder + one script) 01 SCOUT · yt-dlp: pull the 10 top videos in the niche from the last 30 days 02 TRANSCRIBE · whisperX: word-timed transcripts, then write an original script with a hook in the first 2 seconds 03 VOICE · chatterbox: narrator voice with emotion tags, one wav per scene 04 ART · ComfyUI: character sheet first, then every scene in the locked style, same character on-model 05 MOTION · Wan2.2: each still becomes a 3-5 s moving shot 06 LIP-SYNC · LivePortrait: the character's mouth follows the voice 07 EDIT · remotion: the long-form cut in code, captions from step 02, music under the voice 08 CLIPS · AI-Youtube-Shorts-Generator: 6 vertical 9:16 shorts from the long cut 09 POST · postiz: schedule all 7 videos with titles, descriptions and tags QUALITY BAR - Every shot on-model: same face, colours and proportions as the character sheet - Captions word-synced, big, readable on a phone - Audio: voice clear, music ducked under it, no clipping - Shorts start on motion, never on a black frame or a logo MUST NOT - No stock footage, no slideshow of still images, no robotic TTS voice - No step that needs me to click in a GUI. Everything runs from the terminal - No paid APIs Before you finish: render one full run, pull 10 random frames from every video and look at them. Fix any off-model, blurry or caption-broken frame and rerun that room only. Do not stop until all 7 videos would pass as an upload from a real channel.