<aside> 🛠

Add an AI-Voiced Cold-Open Narration to Every Clip — Powered by Qwen3

Give every clip a spoken cold-open intro. Video Narrate reads each clip's Hook (or Title) text using Qwen3 text-to-speech and prepends the resulting voiceover to the front of the clip. Perfect for "here's what you're about to see…" intros, trailer-style openings, accessible audio-first framing, and multi-language content (Korean, English, or auto-detected). Applied to every clip in a batch in one node.

</aside>

What it does

Video Narrate (Qwen3) reads the LLM-generated Hook or Title text carried alongside each clip (from Video Analysis) and synthesizes a spoken voiceover using Qwen3 — the audio-capable model family from Alibaba (Qwen). The synthesized narration is prepended to the front of each clip, giving it a cold-open intro that tells viewers what's coming out loud rather than just visually. Multi-language support (Auto / Korean / English) and a narrator voice preset let you match the audio style to the content. Batched: applied to every clip in the input list.

Problem it solves

• Spoken cold-open intros – Give each clip a "here's what you're about to see" audio intro without recording voiceover yourself
• Trailer-style narration – Trailer-editor pattern of "In a world…" voiceover openings, automated from LLM-generated hook text
• Complementary to visual hooks – Where Hook Teaser builds a visual cold-open from the clip's own footage, Video Narrate builds a spoken cold-open from the hook/title text — the two can be layered
• Accessible audio-first framing – Not every viewer looks at the screen in the first second; a spoken intro reaches viewers who are listening
• Multi-language content – Auto language detection or explicit Korean / English selection means the same node handles content in either language cleanly
• Fast batch narration – Voice every clip in a batch (highlight reel, episode's worth of shorts) in one node, no per-clip recording required
• Consistent narrator voice – Same voice preset across the whole batch = uniform brand voice
• Fallback text handling – If the primary text field (Hook or Title) is empty, the tool falls back to the other, so no clip ends up with a silent intro

Input/Output

<aside>

• Input: List of video clips + carried metadata
  • Video clip list (top input): From Video Trim, Remove Dead Zone, Video Reframe (9:16), or any upstream clip-producing node
  • Carried metadata (second input): The per-clip metadata originally produced by Video Analysis (title, hook, etc.) that travels alongside the clips through the pipeline </aside>

<aside>

• Output: List of clips with a narrated cold-open prepended
  • Structure: [voiceover intro speaking Hook or Title] + [original clip]
  • Format: Same list shape as input — one output clip per input clip
  • Audio: Synthesized narration mixed at the front; original clip audio follows unchanged
  • Language: Matches the language of the source text (or the explicit hint from Advanced Options) </aside>

Configuration Options

• Which carried text each clip's voiceover speaks (falls back to the other when the chosen one is empty): Which text field to narrate. Fallback: If the chosen field is empty for a given clip, the tool falls back to the other field — so a clip missing a Hook still gets narrated (using its Title, and vice versa).

  Option	What it speaks	Best For
  Hook (default)	The clip's hook line from Video Analysis (e.g., "This is Chunland!")	Retention-focused, attention-grabbing openings
  Title	The clip's descriptive title from Video Analysis (e.g., "We Have Discovered New Land!")	Editorial framing; "this clip is about…" style intros

• Narrator voice preset: Which synthesized voice to use.
  • Default: default — the studio's default Qwen3 voice
  • Purpose: Choose a voice style that matches your content's tone (professional, warm, energetic, etc.)
  • Tip: Keep the same preset across a batch so every clip in the same series sounds like it comes from the same narrator