AI agent skill · by @Pluviobyte
Rn Bw Text Opener
Create reusable black-background white-text opening animations for new videos. Use when the user asks for "黑底白字开场", "纯黑加白字", "白色文字逐渐出来", "一个个字出来", "打字效果", "打字音效", "字体变化开头", "opening title animation", "kinetic text opener", or a short black-and-white intro before a HyperFrames, Remotion, ffmpeg, SaaS/product, tutorial, commentary, or course video. This is for original opener creation, not pixel-level reference matching.
Archived 2026-10-23 · Source updated 2026-09-14. Check the original repository for the latest version.
SKILL.md instructions
# Black White Text Opener ## Role Create a short, reusable opening animation: black background, large white text that appears one character at a time, synchronized typing click SFX, optional word replacement, and a clean transition into the main video. This skill is generative. Do not copy source frames or promise pixel-level alignment with a reference. If the user asks to exactly match an existing video, route to `rn-replica-qc` first. ## Required Reads Before building an opener, read: 1. `references/motion-contract.md` 2. `references/timing-presets.md` Read `references/tool-recipes.md` when choosing HyperFrames, Remotion, or ffmpeg. ## Inputs Collect or infer: - main opening phrase - optional replacement words or contrast words - final hold phrase, logo text, or next-scene lead-in - target ratio and resolution - duration preset: `micro`, `standard`, or `slow` - tone: cinematic, clean tutorial, product launch, or opinion opener - transition into the next scene: fade, white velocity wipe, text smear, or cut - typing cadence and SFX preference, default `14 chars/s` with subtle key clicks If the user gives only a topic, write 2-3 short phrase options and choose the cleanest one. Avoid long sentences. ## Defaults - Duration: `standard` 4.8-5.8s. - Background: pure or near black, no busy decorative background. - Text: large white sans-serif, centered or slightly above center. - Motion: black hold -> white text types in character by character -> phrase holds -> replacement text types in -> final typed phrase holds -> transition out. - Audio: quiet typing clicks aligned to visible character appearances. Do not add clicks during black hold, final hold, or transition-only frames. - Build tool: HyperFrames for polished video projects; Remotion when the project already uses Remotion; ffmpeg only for simple utility openers. ## Workflow ### 1. Create The Timing Plan Use the bundled script to generate a deterministic plan: ```bash python3 scripts/create_opener_plan.py \ --title "AI workflow engine" \ --replace "from idea" "to shipped video" \ --final "Build in minutes" \ --preset standard \ --typing-cps 14 \ --out opener-plan.json ``` Use the plan as the source of truth for scene timing and copy. Do not improvise new copy during rendering unless the user asks. ### 2. Build The Animation Follow the motion contract: - hold black before text appears - reveal each text state one character at a time - align typing click SFX to the same character timestamps - keep the first readable phrase large - use only 1-3 replacement states - hold the final phrase long enough to read - transition out cleanly into the next scene For dark SaaS/product videos, this opener can precede `rn-dark-saas-video`; keep the opener simpler than the main product scene. ### 3. QC Before handoff, verify: - first frames are black or near black - characters appear sequentially rather than all at once - typing clicks line up with the character reveal and stop during holds - white text is readable at mobile size - no text clips outside frame - no accidental colored flash - no persistent neon-line background - transition lands before the next scene starts - media specs match the target project If this opener is inserted into a longer video, export both the standalone opener and the full composition when practical. ## Output Return: - opener MP4 path or composition path - timing plan path - key text states and timestamps - typing cadence and SFX schedule - QC notes, especially first-frame blackness and text readability