13 articles in this topic.
Step-by-step: install the tools, set up a Remotion project, and build a complete Claude Code skill that turns a one-line idea into a rendered vertical MP4 — the foundation for a full AI video studio.
Build the tiktok-script skill: a JSON contract, a craft reference that teaches Claude how short-form scripts work, a validator that enforces pacing and structure, and a generated shooting guide with cue cards.
Build the tiktok-video skill: six stages from preflight to render, Whisper word timing with script-correct spelling, silence-based jump cuts, a typed props contract and a Remotion composition with animated subtitles.
Build the ai-story skill: a fable becomes a narrated, illustrated video generated entirely on a Mac — per-line TTS for exact subtitles, consistent characters from a local image model, and an approval gate before the slow part.
Build the votri-script skill: teach Claude deadpan comedy, revise the script with you in chat, print it as a performance-ready PDF with headless Chrome, and plug it into the editing pipeline with an adapter.
A practical guide: get a script with /tiktok-script, film one clip per scene following the shooting guide, and let /tiktok-video edit it into a finished vertical video — plus fine-tuning cuts and fixing common problems.
A practical guide to narrated, illustrated story videos made on your Mac: one-time setup, reviewing the script before the long drawing step, redrawing single pictures, choosing voices and styles, and reading the warnings.
A practical guide: turn a saying or pickup line into a deadpan comedy script, revise it in chat, print the PDF, perform with a straight face, and let the editor add the music cut, sound effects and snap zoom automatically.
Turn a one-line idea into a finished vertical video: how Claude Code skills, a single JSON contract, validators and a staged Node pipeline drive Remotion to script, edit, subtitle and score short videos automatically.
Speech recognition gets Vietnamese accents wrong, and a misspelled subtitle is the most visible mistake a video can have. Use Whisper only for timestamps, align them to the script with Needleman–Wunsch, and render TikTok-style word captions in Remotion.
Detect silence with FFmpeg, turn it into an editable cut list, and map every word timestamp through the cuts so subtitles and sound effects stay in sync — plus punch-in zooms and a deadpan mode where the pauses are the joke.
Four looping background tracks and fourteen sound effects, generated from scratch with FM synthesis, Karplus–Strong strings, biquad filters and seeded noise — no samples, no licenses, no content-ID claims.
Narrated, illustrated story videos generated entirely on a Mac: per-line Vietnamese TTS for exact subtitle timing, consistent characters from a local image model, and a Remotion edit measured from a reference video.
