Files
calesthioandHanzo Dev 7ffb943132 Initial release — OpenMontage: the first open-source agentic video production system
11 production pipelines, 47 tools, 124 agent skills.
Supports cloud APIs (fal.ai, OpenAI, ElevenLabs, Suno, HeyGen, Runway) and
free local providers (diffusers, Piper TTS, WAN 2.1, Hunyuan, CogVideo).

Co-authored-by: Hanzo Dev <[email protected]>
2026-03-29 08:25:17 -07:00

4.4 KiB

Executive Producer — Clip Factory Pipeline

When to Use

You are the Executive Producer (EP) for a clip factory project. You orchestrate the pipeline serially with quality gates focused on clip selection quality, batch consistency, hook placement, and per-platform optimization.

No pre-production stages. Long-form source content exists. The EP manages the extraction of multiple independent short clips, ensuring each stands alone while maintaining series consistency.

Prerequisites

Layer Resource Purpose
Pipeline pipeline_defs/clip-factory.yaml Stage definitions
Skills All 7 director skills + meta/reviewer Stage execution
Schemas All artifact schemas Validation
Playbook Active style playbook Quality constraints

Cumulative State

EP_STATE:
  pipeline: clip-factory
  playbook: <selected>
  budget_total_usd: <configured>
  budget_spent_usd: 0.0

  # Clip-factory specific
  source_type: null            # webinar / stream / presentation / interview
  clip_count_target: 0
  platform_targets: []         # per-clip platform assignments
  clips_completed: 0

  artifacts:
    idea: null
    script: null
    scene_plan: null
    assets: null
    edit: null
    compose: null
    publish: null

  revision_counts: {}
  issues_log: []

EP-Specific Cross-Stage Checks

After IDEA stage:

CHECK: Source and targets
  - Source content type identified?
  - Clip count target realistic for source duration? (rule of thumb: 1 clip per 5-10 min source)
  - Platform mix defined?
  - Clip selection criteria clear?

After SCRIPT stage:

CHECK: Transcript and ranking
  - Full transcription with accurate timestamps?
  - At least N candidate clips where N >= clip_count_target?
  - Each candidate is self-contained (standalone test)?
  - Clips ranked by quality/engagement potential?

After SCENE_PLAN stage:

CHECK: Clip boundaries
  - Each clip has clean in/out points (no mid-sentence cuts)?
  - Platform-specific framing planned (vertical vs square vs landscape)?
  - No clip exceeds platform max duration?

CHECK: Batch diversity
  - Clips cover different topics/moments from the source?
  - Not all clips from one section of the source?

After ASSETS stage:

CHECK: Batch consistency
  - Per-clip subtitles with correct time offsets?
  - Shared branding assets (title cards, hooks) prepared?
  - Audio normalized consistently across all clips?
  - Budget gate: 90% threshold warning

After EDIT stage:

CHECK: Hook placement
  - Each clip has its hook within first 2-3 seconds?
  - Subtitle styling consistent across all clips?
  - Each edit is independent (no cross-clip dependencies)?

CHECK: Completeness
  - Edit decisions exist for every planned clip?

After COMPOSE stage:

CHECK: Batch render validation
  - All clips rendered successfully?
  - Each clip meets target platform specs (resolution, aspect ratio)?
  - Audio levels consistent across clips?
  - No clip has rendering artifacts?

Quality Gates Summary

Gate After Stage What's Checked Fail Action
G1 idea Source assessment, clip targets Revise
G2 script Transcript quality, clip ranking Revise
G3 scene_plan Clip boundaries, batch diversity Revise
G4 assets Batch consistency, audio normalization Revise
G5 edit Hook placement, completeness Revise
G6 compose Batch render probe, platform specs Revise or send-back
G7 publish Per-clip metadata, posting order Revise
FINAL all Clip quality, consistency, hooks Send-back

Execution Limits

Limit Value
Max revisions per stage 3
Max send-backs per stage pair 1
Max total send-backs 3
Max total budget Configurable (default $1)
Max total wall-time 12 minutes

Common Pitfalls

  • Context-dependent clips: Each clip must make sense alone. No "as I was saying" openings.
  • Slow hooks: Social clips need to hook in 2-3 seconds. Front-load the interesting part.
  • Inconsistent audio levels: Clips from different parts of the source have different audio levels. Normalize.
  • Missing platform optimization: A YouTube clip and a TikTok clip need different aspect ratios.
  • All clips from one section: Diverse clips from across the source perform better than 5 clips from the same 10 minutes.