<aside> 🛠

Auto-Mix Emotion-Matched Sound Effects — Timed to Your Highlights

Add the "creator SFX polish" — boings, dings, laughs, whooshes, record scratches — to every clip in a batch, timed to hit at each clip's emotional peak. When Video SFX is paired with Video Analysis (EDL mode), it reads each clip's emotion tag directly from the analysis metadata and places the SFX at the highlight moment. Without EDL metadata, it falls back to scanning the transcript for emotional / comedic keywords and syncing SFX to matching words. Either way, effects are mixed at a low, non-dominant level so voices stay clear. Default settings are tuned to feel professional (not gimmicky); tune volume, minimum gap, and pop-sync delay in Advanced Options.

</aside>

What it does

Video SFX takes a list of video clips plus their transcript and per-clip metadata (from Video Analysis), and mixes emotion-matched sound effects into each clip's audio. It runs in one of two modes depending on what's wired in:

• Primary — EDL mode (Video Analysis metadata present): Reads each clip's emotion tag from the Video Analysis EDL and places one SFX per clip at the highlight moment (aligned with the clip's hook_sec). This is more accurate than keyword scanning because it uses the LLM's semantic understanding of the whole clip.
• Fallback — Keyword mode (no emotion metadata): Scans the transcript for emotional / comedic keywords (ㅋㅋㅋ, haha, "aha", "wow", etc.) and places matching SFX at each detected word.

Either way, anti-spam gapping and a small SFX-to-caption delay make the result feel like a professional sound designer edit rather than an automated mash-up. Effects like a laugh SFX under a comedy beat, a chime on a revelation, or a whoosh on a scene-change cue — all timed by machine but shaped by the same rhythms a human editor would use

Problem it solves

• Creator-level polish, automated – The SFX layer (laughs, dings, boings, whooshes) is a signature part of high-retention short-form content but painful to add manually; this tool does the whole batch in one pass
• Emotion-aware placement – SFX are chosen from each clip's emotion tag (EDL mode) or the actual meaning of transcript words (keyword mode) — not placed randomly
• Highlight-anchored timing – In EDL mode, one SFX per clip lands at the clip's hook moment, giving each clip a clear sonic beat. In keyword mode, effects land in sync with the matching caption pop.
• Semantic over lexical – EDL mode captures moments that keyword scanning would miss (e.g., a laugh SFX on a wordless comedic pause the LLM tagged as "funny")
• Voice-first mix – The default 0.35 mix volume keeps narration dominant; SFX enhance rather than compete with what's being said
• Anti-spam protection – Rapid runs of laughing text (ㅋㅋㅋ, hahaha) don't stack up into a wall of overlapping SFX
• Batch across a highlight reel – Apply the same SFX treatment consistently across every clip in the input list
• Graceful fallback – Clips missing an emotion tag automatically fall back to keyword scanning, so mixed / partial-metadata batches still get complete SFX coverage
• No sound design tools needed – Skips the entire "browse SFX library / drag onto timeline / align to keyframe" workflow

Input/Output

<aside>

• Input: Video clip list + transcript (+ optional analysis metadata)
  • Video clip list (top input, blue port): A list of video clips (typically from Video Trim, Remove Dead Zone, Video Reframe (9:16), or any upstream clip-producing node)
  • Transcript (bottom input, pink/magenta port): Word-timed transcript for each clip (typically from an upstream ASR / Speech-to-Text node). Provides both keyword-mode matching and the timing basis for SFX placement.
  • Analysis metadata (EDL from Video Analysis): Per-clip metadata that travels alongside the clips through the pipeline. When each clip's metadata includes an emotion tag, Video SFX runs in EDL mode and reads the tag directly. When a clip has no emotion tag, that clip falls back to keyword scanning of its transcript. </aside>

<aside>

• Output: List of clips with SFX mixed into the audio
  • Format: Same list shape as input — one output clip per input clip
  • Audio: Original audio preserved with SFX mixed in on top at the configured volume
  • Video: Passes through unchanged
  • Sync: SFX timed to caption pop-in based on word timestamps (with the configured pop-sync delay applied) </aside>