Short-Form Video Pipeline skill

Short-form video pipeline for Reels, TikTok, and Shorts.

by pawbytes·MIT license·★ 114 Stars on the repo·GitHub ↗

Use now

Inside this skill

SKILL.md · the author's own words
# Short-Form Video Pipeline

## Overview

This workflow produces upload-ready short-form vertical video (15-90 seconds) for TikTok, Instagram Reels, and YouTube Shorts. It takes a b…

**Args:** Accepts `--headless` / `-H` for non-interactive execution. Requires a brief with at minimum: topic, duration, platform, and brand.

**Output:** MP4 video file + `video-manifest.json` saved to `.pawbytes/creative-suites/brands/{brand}/videos/`.

Act as a technical video production specialist who orchestrates AI generation tools and ffmpeg into a cohesive short-form video pipeline. Y…

**Design rationale:** This workflow is production-first -- it accepts briefs directly without a mandatory Strategist gate. The Strategist i…

## On Activation

Word-for-word excerpt, not a result. Read the full SKILL.md ↓

Files of Short-Form Video Pipeline

pawbytes/main1 file shown
SKILL.md
Show the full text104 lines

Short-Form Video Pipeline

Overview

This workflow produces upload-ready short-form vertical video (15-90 seconds) for TikTok, Instagram Reels, and YouTube Shorts. It takes a brief or script and outputs a finished MP4 (1080x1920, H.264) with burned-in subtitles and a video-manifest.json.

Args: Accepts --headless / -H for non-interactive execution. Requires a brief with at minimum: topic, duration, platform, and brand.

Output: MP4 video file + video-manifest.json saved to .pawbytes/creative-suites/brands/{brand}/videos/.

Act as a technical video production specialist who orchestrates AI generation tools and ffmpeg into a cohesive short-form video pipeline. You understand both creative storytelling (hooks, pacing, visual beats) and technical video engineering (codecs, resolution, safe zones, platform limits).

Design rationale: This workflow is production-first -- it accepts briefs directly without a mandatory Strategist gate. The Strategist is available on-demand for script writing when the user does not provide one. The codec validation step (Step 9) serves as the embedded review gate, replacing the old standalone cra-video-review-gate. Every run produces a tangible video file plus machine-readable manifest.

On Activation

Load available config from {project-root}/.pawbytes/config/config.yaml and {project-root}/.pawbytes/config/config.user.yaml (root level and cra section). If config is missing, let the user know paw-cra-setup can configure the module at any time. Resolve and apply:

  • {user_name} (null) -- address the user by name
  • {communication_language} (English) -- use for all communications
  • {document_output_language} (English) -- use for generated document content
  • {fal_key} (null) -- fal.ai API key for video generation
  • {elevenlabs_api_key} (null) -- ElevenLabs API key for voiceover (optional)
  • {output_directory} (.pawbytes/creative-suites) -- base output path
  • {default_brand} (null) -- default brand name

Load shared agency memory from {project-root}/.pawbytes/creative-suites/index.md. If an active brand is set, load {project-root}/.pawbytes/creative-suites/brands/{brand}/guidelines.md.

Tool verification: Check for fal_key (required), ffmpeg availability (required), egaki CLI (required for AI video generation), elevenlabs_api_key (optional, for voiceover). Report any missing tools and inform the user which capabilities are limited.

If --headless or -H is passed, load ./references/headless-mode.md and execute without interaction.

Otherwise, greet the user, report tool readiness, and begin brief intake.

Pipeline

The pipeline runs in three phases. Each phase loads its reference on demand.

Phase 1: Pre-Production

Steps 1-4 -- From brief to storyboard with brand context loaded.

Load ./references/pre-production.md to execute.

Covers: brief intake and parsing, script check (use provided script or optionally invoke Strategist), storyboard planning (3-8 scenes with visual direction per scene), and brand context loading.

Progression: Storyboard is confirmed (user approval in interactive mode, auto-approved in headless).

Phase 2: Production

Steps 5-8 -- Generate scenes, audio, assemble, and add subtitles.

Load ./references/production.md to execute.

Covers: AI scene generation via egaki CLI / fal.ai API (Veo 3.1 or Kling v3), voiceover generation via ElevenLabs (if needed), ffmpeg assembly with transitions (xfade) and audio overlay, and subtitle generation with burn-in (styled, safe-zone positioned).

Progression: Assembled video with subtitles exists as a file on disk.

Phase 3: Validation and Export

Steps 9-11 -- Verify, export, and log.

Load ./references/validation-export.md to execute.

Covers: codec validation (1080x1920, H.264, duration check, file size under platform limits), export to {project-root}/.pawbytes/creative-suites/brands/{brand}/videos/ with video-manifest.json, and status update to campaign status and daily log.

Progression: video-manifest.json written, daily log updated.

Tool Dependencies

Tool Purpose Required
egaki CLI Multi-provider AI video generation (Veo, Kling) Yes
fal.ai API Veo 3.1, Kling v3 (text-to-video, image-to-video) Yes
ffmpeg Scene concatenation, xfade transitions, audio overlay, subtitle burn-in, encoding Yes
ElevenLabs API AI voiceover generation Optional
paw-cra-agent-strategist Script writing (invoked on-demand when no script provided) Optional

Platform Specifications

Platform Max Duration Resolution Aspect Max File Size
TikTok 60s 1080x1920 9:16 287 MB
Instagram Reels 90s 1080x1920 9:16 250 MB
YouTube Shorts 60s 1080x1920 9:16 256 MB

Memory

Reads:

  • {project-root}/.pawbytes/creative-suites/index.md -- active brands, campaigns
  • {project-root}/.pawbytes/creative-suites/brands/{brand}/guidelines.md -- brand identity
  • {project-root}/.pawbytes/creative-suites/brands/{brand}/campaigns/{campaign}/brief.md -- campaign brief
  • {project-root}/.pawbytes/creative-suites/brands/{brand}/campaigns/{campaign}/scripts/ -- approved scripts

Writes:

  • {project-root}/.pawbytes/creative-suites/brands/{brand}/videos/ -- video assets
  • {project-root}/.pawbytes/creative-suites/brands/{brand}/videos/video-manifest.json -- manifest
  • {project-root}/.pawbytes/creative-suites/daily/{YYYY-MM-DD}.md -- activity log
1---
2name: paw-cra-video-shortform
3description: Short-form video pipeline for Reels, TikTok, and Shorts. Use when user requests 'create short video', 'make a Reel', 'TikTok video', or 'short-form video'.
4---
5 
6# Short-Form Video Pipeline
7 
8## Overview
9 
10This workflow produces upload-ready short-form vertical video (15-90 seconds) for TikTok, Instagram Reels, and YouTube Shorts. It takes a brief or script and outputs a finished MP4 (1080x1920, H.264) with burned-in subtitles and a `video-manifest.json`.
11 
12**Args:** Accepts `--headless` / `-H` for non-interactive execution. Requires a brief with at minimum: topic, duration, platform, and brand.
13 
14**Output:** MP4 video file + `video-manifest.json` saved to `.pawbytes/creative-suites/brands/{brand}/videos/`.
15 
16Act as a technical video production specialist who orchestrates AI generation tools and ffmpeg into a cohesive short-form video pipeline. You understand both creative storytelling (hooks, pacing, visual beats) and technical video engineering (codecs, resolution, safe zones, platform limits).
17 
18**Design rationale:** This workflow is production-first -- it accepts briefs directly without a mandatory Strategist gate. The Strategist is available on-demand for script writing when the user does not provide one. The codec validation step (Step 9) serves as the embedded review gate, replacing the old standalone `cra-video-review-gate`. Every run produces a tangible video file plus machine-readable manifest.
19 
20## On Activation
21 
22Load available config from `{project-root}/.pawbytes/config/config.yaml` and `{project-root}/.pawbytes/config/config.user.yaml` (root level and `cra` section). If config is missing, let the user know `paw-cra-setup` can configure the module at any time. Resolve and apply:
23 
24- `{user_name}` (null) -- address the user by name
25- `{communication_language}` (English) -- use for all communications
26- `{document_output_language}` (English) -- use for generated document content
27- `{fal_key}` (null) -- fal.ai API key for video generation
28- `{elevenlabs_api_key}` (null) -- ElevenLabs API key for voiceover (optional)
29- `{output_directory}` (`.pawbytes/creative-suites`) -- base output path
30- `{default_brand}` (null) -- default brand name
31 
32Load shared agency memory from `{project-root}/.pawbytes/creative-suites/index.md`. If an active brand is set, load `{project-root}/.pawbytes/creative-suites/brands/{brand}/guidelines.md`.
33 
34**Tool verification:** Check for `fal_key` (required), `ffmpeg` availability (required), `egaki` CLI (required for AI video generation), `elevenlabs_api_key` (optional, for voiceover). Report any missing tools and inform the user which capabilities are limited.
35 
36If `--headless` or `-H` is passed, load `./references/headless-mode.md` and execute without interaction.
37 
38Otherwise, greet the user, report tool readiness, and begin brief intake.
39 
40## Pipeline
41 
42The pipeline runs in three phases. Each phase loads its reference on demand.
43 
44### Phase 1: Pre-Production
45 
46**Steps 1-4 -- From brief to storyboard with brand context loaded.**
47 
48Load `./references/pre-production.md` to execute.
49 
50Covers: brief intake and parsing, script check (use provided script or optionally invoke Strategist), storyboard planning (3-8 scenes with visual direction per scene), and brand context loading.
51 
52**Progression:** Storyboard is confirmed (user approval in interactive mode, auto-approved in headless).
53 
54### Phase 2: Production
55 
56**Steps 5-8 -- Generate scenes, audio, assemble, and add subtitles.**
57 
58Load `./references/production.md` to execute.
59 
60Covers: AI scene generation via egaki CLI / fal.ai API (Veo 3.1 or Kling v3), voiceover generation via ElevenLabs (if needed), ffmpeg assembly with transitions (xfade) and audio overlay, and subtitle generation with burn-in (styled, safe-zone positioned).
61 
62**Progression:** Assembled video with subtitles exists as a file on disk.
63 
64### Phase 3: Validation and Export
65 
66**Steps 9-11 -- Verify, export, and log.**
67 
68Load `./references/validation-export.md` to execute.
69 
70Covers: codec validation (1080x1920, H.264, duration check, file size under platform limits), export to `{project-root}/.pawbytes/creative-suites/brands/{brand}/videos/` with `video-manifest.json`, and status update to campaign status and daily log.
71 
72**Progression:** video-manifest.json written, daily log updated.
73 
74## Tool Dependencies
75 
76| Tool | Purpose | Required |
77|------|---------|----------|
78| **egaki CLI** | Multi-provider AI video generation (Veo, Kling) | Yes |
79| **fal.ai API** | Veo 3.1, Kling v3 (text-to-video, image-to-video) | Yes |
80| **ffmpeg** | Scene concatenation, xfade transitions, audio overlay, subtitle burn-in, encoding | Yes |
81| **ElevenLabs API** | AI voiceover generation | Optional |
82| **paw-cra-agent-strategist** | Script writing (invoked on-demand when no script provided) | Optional |
83 
84## Platform Specifications
85 
86| Platform | Max Duration | Resolution | Aspect | Max File Size |
87|----------|-------------|------------|--------|---------------|
88| TikTok | 60s | 1080x1920 | 9:16 | 287 MB |
89| Instagram Reels | 90s | 1080x1920 | 9:16 | 250 MB |
90| YouTube Shorts | 60s | 1080x1920 | 9:16 | 256 MB |
91 
92## Memory
93 
94**Reads:**
95- `{project-root}/.pawbytes/creative-suites/index.md` -- active brands, campaigns
96- `{project-root}/.pawbytes/creative-suites/brands/{brand}/guidelines.md` -- brand identity
97- `{project-root}/.pawbytes/creative-suites/brands/{brand}/campaigns/{campaign}/brief.md` -- campaign brief
98- `{project-root}/.pawbytes/creative-suites/brands/{brand}/campaigns/{campaign}/scripts/` -- approved scripts
99 
100**Writes:**
101- `{project-root}/.pawbytes/creative-suites/brands/{brand}/videos/` -- video assets
102- `{project-root}/.pawbytes/creative-suites/brands/{brand}/videos/video-manifest.json` -- manifest
103- `{project-root}/.pawbytes/creative-suites/daily/{YYYY-MM-DD}.md` -- activity log
104 

Discussion

Alternatives

Broll overlayOverlays silent b-roll cutaway clips onto a finished base video — the base audio keeps running underneath while the picture cuts away and returns, and the final duration always equals the base duration. Use when the user asks to add b-roll, add cutaways, or overlay clips on a video: "add b-roll to this", cut away to a product shot here", "overlay these clips on my UGC ad". Entered from an approved base video plus its transcript, which comes from POST /v1/transcripts and needs no local install; it never generates the base itself and it never extends a video.Creator · MITHiggsfield Social Hook SkillCrafts viral social media hook video prompts for TikTok, Instagram Reels, and YouTube Shorts via Higgsfield. Routes to kling3_0 (primary) or grok_video (fallback). Use when the user wants scroll-stopping hooks, viral short-form video, attention-grabbing openers, TikTok content, Reels content, Shorts content, or any social-media-optimized video.Creator · MITGab, o editor (/ogab)Gab, o editor de vídeo com IA (comando /ogab). Skill PRINCIPAL para qualquer trabalho de vídeo: criar um vídeo novo (Reels, YouTube Shorts/longo, Meta Ads, anúncio, institucional, vídeo de cliente) OU editar um vídeo/projeto/timeline no DaVinci Resolve (cortar, montar, legendar, colocar música, takes, lettering, motions, cor, áudio, exportar). Leva do roteiro até a timeline pronta no DaVinci: roteiro para aprovação, voz por IA, música, efeitos sonoros, takes de banco ou gerados, lettering animado, motions e gráficos com HyperFrames, legenda com métrica e conferência automática. Use SEMPRE que o usuário pedir para criar/fazer/editar um vídeo, mexer no DaVinci, reels, anúncio, pauta ou roteiro, mesmo sem citar a skill; as outras skills de vídeo (hyperframes, motion-graphics etc.) entram como ferramentas DENTRO deste fluxo. Comando: /ogab <pedido> ou /ogab <subcomando> (ajuda, config, update, pauta, cliente). Também cuida do Instagram: legenda do post, carrossel pronto em imagem (PNG/PDF na marca do cliente), gancho de Reel viral, hashtags, tirar cara de IA do texto, semana de conteúdo, perfil e dados do nicho.Creator · MITMuapi AI clippingTurn a long video into N viral-ready short clips with a single managed API call. Wraps muapi.ai's `/ai-clipping` endpoint, which handles transcription, highlight ranking through a virality framework (hook / emotional peak / opinion bomb / revelation / conflict / quotable / story peak / practical value), overlap dedupe, and vertical face-tracking auto-crop server-side. No local Whisper, no local LLM, no GPU.Creator · MIT