<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: WaveXML</title>
    <description>The latest articles on DEV Community by WaveXML (@wavexml).</description>
    <link>https://dev.to/wavexml</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4247651%2F427512c0-4d33-4efd-91bc-e994adebf91f.png</url>
      <title>DEV Community: WaveXML</title>
      <link>https://dev.to/wavexml</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/wavexml"/>
    <language>en</language>
    <item>
      <title>Syncing dual-system audio across multiple cameras without timecode, then letting an AI rough-cut it via MCP</title>
      <dc:creator>WaveXML</dc:creator>
      <pubDate>Mon, 16 Nov 2026 11:43:29 +0000</pubDate>
      <link>https://dev.to/wavexml/syncing-dual-system-audio-across-multiple-cameras-without-timecode-then-letting-an-ai-rough-cut-it-2p6n</link>
      <guid>https://dev.to/wavexml/syncing-dual-system-audio-across-multiple-cameras-without-timecode-then-letting-an-ai-rough-cut-it-2p6n</guid>
      <description>&lt;p&gt;You shot a wedding, a panel or an interview on three cameras plus a separate audio recorder. Nobody jam-synced timecode, and now you have a pile of clips that need to line up before real editing starts. This workflow doesn't rely on timecode, and the limits are stated plainly.&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Disclosure: this is written by the WaveXML team, and the tool in it is ours.&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  1. Shoot so the sync is possible
&lt;/h2&gt;

&lt;p&gt;Waveform sync only works if the software can hear the same sound on every device.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Keep the camera mics on, on every camera.&lt;/strong&gt; Their audio is "scratch": thin and noisy is fine, but it is what gets matched against the recorder. Even quiet room tone can be enough. A clip with no audio can't be matched.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Give yourself a sync point.&lt;/strong&gt; A clap or slate at the start of each take, and again at the end of long ones, makes checking easy later.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Keep the recorder running&lt;/strong&gt; through the event, so every camera overlaps with it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Set camera clocks as well as you can.&lt;/strong&gt; Sync by sound doesn't need them exact, but they help when you sort a day of footage.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  2. Lay out the timeline
&lt;/h2&gt;

&lt;p&gt;In your editor (Resolve, Premiere or Final Cut), build one timeline:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;One camera per video track.&lt;/li&gt;
&lt;li&gt;The recorder on its own audio track, not mixed under the cameras.&lt;/li&gt;
&lt;li&gt;Camera mics left on and linked.&lt;/li&gt;
&lt;li&gt;Don't nest the multicam yet.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Export it. In Resolve that's File → Export → Timeline → OTIO. WaveXML's tutorial recommends OTIO because stereo stays stereo; FCP7 XML works too but returns stereo as mono. Final Cut exports FCPXML.&lt;/p&gt;

&lt;h2&gt;
  
  
  3. Sync, then spot-check
&lt;/h2&gt;

&lt;p&gt;&lt;a href="https://wavexml.com/" rel="noopener noreferrer"&gt;WaveXML&lt;/a&gt; is a free Mac and Windows beta. Drop the exported timeline on its start screen and press Synchronize. It matches the recorder to each camera by sound and writes a synced file you import back into your editor. No timecode needed.&lt;/p&gt;

&lt;p&gt;Then check it yourself. Find a loud transient, like the clap, and look at it on every angle. If a clip stays unsynced, check that it has audio and that two cameras aren't sharing a track.&lt;/p&gt;

&lt;h2&gt;
  
  
  4. The honest limits of waveform sync
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;It needs usable scratch audio on each camera.&lt;/strong&gt; It needs overlapping sound; a muted camera can't be matched.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Long recordings can drift.&lt;/strong&gt; Camera and recorder clocks run independently. For recordings of an hour or more, the site recommends splitting the timeline into segments of about 30 minutes and syncing each.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;It's a beta.&lt;/strong&gt; Full app, no clip limits, but still a beta. The Windows installer is unsigned, so SmartScreen may warn; Mac needs macOS 14 or later.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;No AAF.&lt;/strong&gt; It reads FCP7 XML, OTIO, FCPXML and XMEML, not AAF.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  5. Optional: an AI first draft through MCP
&lt;/h2&gt;

&lt;p&gt;MCP (Model Context Protocol) is the standard way AI apps call outside tools. WaveXML 2.0 has no AI model of its own: it syncs, measures shots and writes the edit file, and an AI you already use makes the editing choices. That can be your own Claude, ChatGPT (Codex) or Cursor, or any MCP-capable app.&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Click &lt;strong&gt;Talk to your AI&lt;/strong&gt; in WaveXML, or connect an AI app separately.&lt;/li&gt;
&lt;li&gt;Drop the whole footage folder straight off the cards. You get a row per camera and recorder, with a warning if a camera's clock looks wrong. Use &lt;strong&gt;Sync everything&lt;/strong&gt;, or drag across part of the day and use &lt;strong&gt;Sync this part&lt;/strong&gt;.&lt;/li&gt;
&lt;li&gt;Ask in plain words: "Cut a multicam edit of the first speech, switching to whoever is talking."&lt;/li&gt;
&lt;li&gt;Your AI uses WaveXML's tools: frame sheets (still images), shot measurements for shake, soft focus and darkness, and a check of its own cut list before you see it.&lt;/li&gt;
&lt;li&gt;The edit opens in WaveXML's editor. Play it, switch angles, trim, then export to Resolve, Final Cut or Premiere.&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;&lt;strong&gt;Privacy:&lt;/strong&gt; the site says your video files are never uploaded and the AI only sees the still frames it asks for. Those frames and the facts it reads do go to your AI app's maker, under their privacy policy.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Limits:&lt;/strong&gt;&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;The edit is only as good as the AI you connect. Expect a first draft to fix, not a finished film.&lt;/li&gt;
&lt;li&gt;Switchable multicam is tested in DaVinci Resolve; Final Cut and Premiere get the same kind of XML but have had less testing.&lt;/li&gt;
&lt;li&gt;WaveXML doesn't render video or recognise faces. Your AI looks at the frames.&lt;/li&gt;
&lt;li&gt;You need your own AI account.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  A practical routine
&lt;/h2&gt;

&lt;p&gt;Shoot with scratch audio on every camera, one device per track, sync, check a clap on every angle, and only then decide whether you want an AI draft. If you want to try the sync tool, it's at &lt;a href="https://wavexml.com/" rel="noopener noreferrer"&gt;wavexml.com&lt;/a&gt;.&lt;/p&gt;

</description>
      <category>filmmakingvideoeditingaimcp</category>
    </item>
  </channel>
</rss>
