- Home
- Skills
- Cdeistopened
- Opened Vault
- Reel Builder
reel-builder_skill
- HTML
4
GitHub Stars
2
Bundled Files
2 months ago
Catalog Refreshed
3 months ago
First Indexed
Readme & install
Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.
Installation
Preview and clipboard use veilstrat where the catalogue uses aiagentskills.
npx veilstrat add skill cdeistopened/opened-vault --skill reel-builder- HANDOFF-ELIJAH.md3.0 KB
- SKILL.md8.6 KB
Overview
This skill builds multi-scene animated reels from AI-generated images using Fal.ai tools and ffmpeg. It produces a style-consistent set of stills, morphs scenes using Kling start/end frame interpolation, and assembles a normalized final video ready for vertical platforms. The pipeline is optimized for Instagram Reels, TikTok, Shorts, or any short-form multi-shot content.
How this skill works
Generate an anchor image that establishes the visual style, then produce matching scene images with Kontext Max (or flux-general as fallback). For each scene transition, submit a Kling start/end interpolation job that morphs Scene N into Scene N+1 with a short motion prompt. Normalize clip dimensions with ffmpeg, concatenate clips, and add text overlays after assembly.
When to use it
- Creating vertical reels (9:16) with consistent character or art style
- Converting a storyboard into smooth animated transitions without frame-by-frame animation
- Producing short multi-shot promotional or narrative videos from AI images
- Rapidly iterating on social video concepts with low cost per reel
Best practices
- Always generate an explicit anchor image in the target aspect ratio (e.g., 768x1344 for 9:16).
- Use Kontext Max with the anchor image for style/character consistency; fall back to flux-general with reference_strength 0.7 if Kontext times out.
- For Kling transitions, prompt the CHANGE between frames (1–2 actions per 5s clip) rather than re-describing both frames.
- Normalize every Kling output to 1080x1920 via ffmpeg before concatenation to avoid mismatched resolutions.
- Do not bake text into images or Kling prompts; add overlays after assembly using ffmpeg or an editor.
Example use cases
- A 5-scene children’s story reel with consistent watercolor characters and smooth scene morphs.
- A product demo reel that shifts from product close-ups to lifestyle scenes with fluid transitions.
- A marketing short that stitches multiple mood shots into a single vertical video for Instagram/TikTok.
- Batch-generate multiple short variations of the same storyboard for A/B testing ad creatives.
FAQ
Use 9:16 with explicit pixel dimensions like 768x1344 for generation and normalize final clips to 1080x1920 before concatenation.
What if Kontext Max times out or fails to match style?
Fall back to flux-general with reference_image_url and reference_strength 0.7 to retain style cues while being more flexible.
Why add text overlays after assembly?
Text baked into images or Kling prompts distorts in interpolation. Apply text post-assembly with ffmpeg, Remotion, or an editor for crisp results.