video/reel-from-reference

Reel From Reference — analyze a reference reel and produce an original video in the same format.

Category: video
Source: workflows/video/reel_from_reference.py

Input Schema

Field	Type	Default	Description
`audience`	`string`	`""`	Target audience.
`background`	`string`	`"ai"`	Background video source: ‘ai’, ‘random’, ‘stock’, ‘solid’.
`bgm`	`boolean`	`true`	Add background music.
`bgm_volume`	`number`	`0.12`	Background music volume (0.0-1.0).
`creator_persona`	`string`	`"confident young creator, natural and conversational"`	Creator persona description.
`enable_ocr`	`boolean`	`false`	Run OCR on sampled frames.
`platform`	`string`	`"instagram_reels"`	Target platform.
`quality`	`string`	`"standard"`	Quality preset: free, budget, standard, premium, local, local-power
`reference_video_path`	`string`	`""`	Local path to reference video (MP4/MOV).
`reference_video_url`	`string`	`""`	URL to download reference video (YouTube, TikTok, Instagram, direct link).
`regenerate`	`object`	—	When set, this run is a regeneration. Workflows may read `direction` / `keep` / `extra_instructions` to modulate prompts; the engine persists `parent_run_id` and `parent_variant_index` as run lineage columns.
`subtitles`	`boolean`	`true`	Burn karaoke captions onto video.
`tone`	`string`	`"energetic"`	Tone: energetic, calm, sarcastic, etc.
`topic`	`string`	required	Topic for the new video.
`variants`	`integer`	`1`	Number of independent variant executions (1–10). When > 1, the engine runs the workflow N times with different sampling, producing N outputs.
`whisper_model`	`string`	`"large-v3"`	Whisper model for transcription.

Output Schema

Field	Type	Default	Description
`asset_id`	`object`	—	Fabric asset ID.
`format_archetype`	`string`	`""`	Detected format archetype.
`format_dna_json_path`	`string`	`""`	Path to format_dna.json.
`kind`	`object`	—	Variant card shape: video / carousel / image / text. Surfaced on the per-variant entry of the run-output API and used by gallery UIs to pick the right layout.
`script_text`	`string`	`""`	The narration text used.
`summary_markdown_path`	`string`	`""`	Path to format summary.
`video_path`	`string`	`""`	Path to the final rendered MP4.

Task Pipeline

validate_and_probe → detect_shots_and_visual_analysis → transcribe_and_analyze_audio → infer_format_dna → generate_recreation_script → prepare_input → generate_voiceover → generate_background → merge_branches → compose_video → add_subtitles → add_bgm → collect_output → collect_reel_output

Task	Description
`validate_and_probe`	Download (if URL provided), validate input video, and extract metadata.
`detect_shots_and_visual_analysis`	Detect shots, classify composition, analyze motion and color.
`transcribe_and_analyze_audio`	Transcribe speech, map pauses, detect retention hooks, analyze audio layers.
`infer_format_dna`	Use LLM to infer structural patterns, then assemble FormatDNA.
`generate_recreation_script`	Generate an original script from format DNA, then bridge to quick-shorts.
`prepare_input`	Validate input and generate script if needed.
`generate_voiceover`	Generate TTS voiceover — reuses SDK voiceover stage.
`generate_background`	Generate per-segment AI video clips, or fall back to catalog/stock.
`merge_branches`	Merge parallel voiceover + background branches.
`compose_video`	Compose per-segment video clips with voiceover.
`add_subtitles`	Burn karaoke subtitles if enabled.
`add_bgm`	Generate and mix background music if enabled.
`collect_output`	Persist artifact, build upgrade payload, and return output.
`collect_reel_output`	Collect final output from quick-shorts and add format metadata.

Run-spec example

Save the YAML below as my-run.yaml, edit the values, and run with the CLI or POST it to the API. Required fields are uncommented; optional knobs are documented above the input: block — copy any line under input: and uncomment to set.

workflow: video/reel-from-reference

# Optional fields — copy any line(s) under `input:` and uncomment to set:
# Target audience.
#   audience: ""
#
# Background video source: 'ai', 'random', 'stock', 'solid'.
#   background: ai
#
# Add background music.
#   bgm: true
#
# Background music volume (0.0-1.0).
#   bgm_volume: 0.12
#
# Creator persona description.
#   creator_persona: "confident young creator, natural and conversational"
#
# Run OCR on sampled frames.
#   enable_ocr: false
#
# Target platform.
#   platform: instagram_reels
#
# Quality preset: free, budget, standard, premium, local, local-power
#   quality: standard
#
# Local path to reference video (MP4/MOV).
#   reference_video_path: ""
#
# URL to download reference video (YouTube, TikTok, Instagram, direct link).
#   reference_video_url: ""
#
# Burn karaoke captions onto video.
#   subtitles: true
#
# Tone: energetic, calm, sarcastic, etc.
#   tone: energetic
#
# Whisper model for transcription.
#   whisper_model: large-v3
#

input:
  # Topic for the new video.
  topic: ""

Run it locally:

fab-workflow --from-file my-run.yaml

Or submit over the wire — the same file is the request body:

curl -X POST 'https://gofabric.dev/v1/workflows/runs?name=video/reel-from-reference' \
  -H 'Authorization: Bearer fab_xxx' \
  -H 'content-type: application/yaml' \
  --data-binary @my-run.yaml

Every workflow also accepts the universal WorkflowInput fields — variants (1–10 fan-out) and regenerate (creative-direction hints with run lineage). See Run-specs (YAML / TOML / JSON) for the full top-level shape (metadata, priority, bundle, parent, etc.).

Warnings

Task generate_voiceover has no Pydantic types — contract is opaque to consumers.
Task merge_branches has no Pydantic types — contract is opaque to consumers.