7.5 KiB
Executable File
Workflow Recipes
Real-world tool chains for common post-production tasks. Each recipe shows what to ask Claude and which tools fire under the hood.
Delivery QC Pipeline
Scenario: Final timeline needs quality sign-off before client delivery.
"Run a full QC check on /path/to/project.fcpxml"
Tool chain: analyze_timeline → detect_flash_frames → detect_gaps → detect_duplicates → validate_timeline
The validate_timeline tool returns a 0–100% health score. Anything below 80% flags specific issues. Follow up with:
"Fix all flash frames by extending previous clips, then fill any gaps"
Tool chain: fix_flash_frames → fill_gaps
Both tools generate _modified output files — your original XML is never touched.
YouTube Chapter Export
Scenario: 45-minute podcast edit with chapter markers needs YouTube-formatted timestamps.
"List all markers in my timeline formatted for YouTube chapters"
Tool chain: list_markers (with format filter)
If chapters don't exist yet but you have a transcript:
"Import these YouTube chapters as markers: 0:00 Intro, 2:15 Topic One, 14:30 Deep Dive..."
Tool chain: import_transcript_markers → list_markers
For SRT/VTT subtitle files from auto-transcription services:
"Import chapters from /path/to/captions.srt as markers"
Tool chain: import_srt_markers
Beat-Synced Music Video Assembly
Scenario: 200 B-roll clips tagged by keyword, one music track with beat analysis.
Step 1 — Import beats:
"Import beat markers from /path/to/beats.json"
Step 2 — Generate assembly:
"Create a rough cut using clips tagged 'performance' and 'broll', target 3:30 duration, accelerating pacing"
Tool chain: import_beat_markers → auto_rough_cut or generate_montage
Step 3 — Snap to beats:
"Snap all cuts to the nearest beat marker"
Tool chain: snap_to_beats
The beat JSON format expects an array of timestamps in seconds:
{ "beats": [0.0, 0.48, 0.96, 1.44, 1.92] }
Cross-NLE Handoff
Scenario: Timeline edited in FCP needs to go to a colorist on DaVinci Resolve and an audio mixer on Pro Tools (via Premiere).
"Export my timeline for DaVinci Resolve and also as FCP7 XML for Premiere"
Tool chain: export_resolve_xml + export_fcp7_xml
What changes in each export:
- Resolve (FCPXML v1.9): Compound clips flattened, unsupported attributes stripped, simpler element tree
- FCP7 XMEML: Spine-based model converted to track-based model — primary storyline becomes Track 0, connected clip lanes map to higher tracks
Documentary A/B Roll
Scenario: Interview footage (A-roll) with cutaway B-roll needs structured assembly.
"Generate an A/B roll edit — 'interview' clips as A-roll, 'broll' clips as B-roll, 8-minute target"
Tool chain: generate_ab_roll
The generator alternates between A-roll and B-roll clips, placing B-roll on connected lanes (above the primary storyline). This matches the standard documentary editing pattern where interview audio runs continuously and visuals cut between talking head and supplementary footage.
Social Media Reformat
Scenario: 16:9 master edit needs vertical versions for Reels/TikTok and square for feed posts.
"Reformat my timeline to 9:16 for Instagram Reels"
Tool chain: reformat_timeline (preset: 9:16)
Available presets: 9:16 (vertical), 1:1 (square), 4:5 (portrait feed), 4:3 (classic), 16:9 (widescreen). Custom resolutions also supported.
Note: This changes the project format metadata — it doesn't re-frame or crop footage. You'll still need to adjust framing in FCP after import.
Timeline Version Comparison
Scenario: Director sent revision notes, you made changes, now need to document what changed.
"Compare /path/to/edit_v2.fcpxml with /path/to/edit_v1.fcpxml"
Tool chain: diff_timelines
Returns structured diff: clips added, removed, moved, or trimmed. Marker changes, transition changes, and format changes are all tracked. Useful for revision logs and client communication.
Silence Cleanup
Scenario: Long-form interview has dead air that needs trimming.
Step 1 — Detect:
"Find silence candidates in my timeline"
Tool chain: detect_silence_candidates
Uses heuristics: gaps, ultra-short clips, naming patterns (clips named "silence", "room tone"), and duration anomalies. Results include confidence scores.
Step 2 — Review and remove:
"Remove all silence candidates with high confidence"
Tool chain: remove_silence_candidates (mode: delete or mark)
Mark mode adds markers instead of deleting — safer for first pass.
Dynamic Subtitles: Generate, Then Always Validate
Scenario: Word-by-word progressive-composition subtitles (the diagrammed look — small supporting words, one key word large in a display italic) need to go on a cut before delivery.
"Generate dynamic subtitles for /path/to/project.fcpxml"
Tool chain: generate_dynamic_subtitles → validate_subtitle_layout
Run generate_dynamic_subtitles on the final cut, after cuts/zooms are already applied — a connected title anchors to its parent clip's source-media coordinates, so re-cutting the timeline afterward can silently detach captions from the words they were built for.
Never treat generation as done without the second call. The layout only guarantees non-overlap by construction for what it itself lays out — it cannot see a hand-edited title, stray content left over in a reused base file, or a word long/uppercase enough to have needed shrinking. validate_subtitle_layout re-measures every <title> independently and reports a severity (none/warning/probable/severe) plus per-issue suggested corrections:
"Validate the subtitle layout in /path/to/project_dynamic_subtitles.fcpxmld"
none/warning(onlyoutside_safe_area, nooutside_frameorcollision) — safe to deliver; the emphasis word sitting close to the 5% margin is expected on the diagrammed look.probable/severe— investigate before touching code. Read the issue's exact FCPXML fraction times (not the rounded float) before deciding whether it's a real overlap; two titles that are only touching at a shared boundary can still round to "equal-looking but not bit-identical" floats and misreport. SeeEngine/docs/03_SERVER_TOOLS.md's "Legendas dinâmicas" section andEngine/docs/05_EXPERIENCIAS.md(2026-08-19 entries) for the concrete bugs already found and fixed this way, and the checklist for the next one.
Composing Tools in AI Agent Workflows
Each tool in this MCP server follows the same pattern: read FCPXML → process → write modified FCPXML. This makes them composable — the output of one tool is valid input for the next.
When used through Claude Desktop or any MCP client, you describe intent in natural language and the agent selects and chains tools automatically. The 5 built-in MCP prompts (qc-check, youtube-chapters, rough-cut, timeline-summary, cleanup) are pre-built chains for the most common workflows.
For custom workflows, describe the full pipeline in one message:
"Analyze my timeline, fix any flash frames, add chapter markers at every 5-minute
interval, then export for DaVinci Resolve"
The agent will chain: analyze_timeline → fix_flash_frames → batch_add_markers → export_resolve_xml, passing the modified file through each step.