Creator OS
2026clipperpodcastspeaker-trackingsplit-screen

Podcast Clipper With Speaker Tracking: 2026 Split Screen Guide

A podcast clipper with speaker tracking follows whoever is talking, cuts to split screen when hosts trade lines, and punches in word by word. Here is how Clippy fram

Creator OS · November 18, 2026 · 8 min read

A podcast clipper with speaker tracking follows whoever is talking in the source video, switches to a split screen when two people trade lines, and cuts on exact words so no clip starts or ends mid-sentence. Creator OS builds this into Clippy, its content clipper, which is available on every plan.

source video (podcast, 2 guests)
        |
        v
  mouth movement per frame  ---> who is talking now?
        |
        +-- one speaker  ---> punched-in shot / wide shot alternating
        |
        +-- two trade lines ---> split screen (left | right)
        |
        +-- transcript words ---> cut on exact word boundaries
        |
        v
  captions + virality score (0-100) ---> 9:16 clip
How speaker tracking drives the layout and the cuts in one podcast clip.

What a Podcast Clipper With Speaker Tracking Does

A speaker-aware clipper turns a long podcast into vertical clips where the framing follows the conversation instead of a fixed crop.

Clippy reads the transcript to find candidate moments, then reads the frames to see who is speaking. From there it picks a layout per moment: one person on screen, two people side by side, or a punch-in on the person talking. Every clip is rendered 9:16 by default and captioned word by word.

A few specifics about the input side. You can find a creator on Twitch, Kick or YouTube and pick one of their streams or videos, paste a video link, or upload your own file. You get up to 10 clips per video, and on a stream Clippy scans up to 3 hours, skipping a 4 minute intro on longer streams.

The output is not a rough export. Clippy scouts about three times as many candidate moments as you asked for, an editor pass cuts and scores each one, and the best are kept. Each clip carries a Virality score from 0 to 100.

Get started for $0

How Speaker Tracking Picks Who Is Talking

Clippy decides who is talking from mouth movement in the footage, not from audio levels alone.

That matters for podcasts recorded in one room with one shared microphone track. Loudness tells you someone spoke; it does not tell you which face on screen moved. Mouth movement maps the sound back to a person, so the framing lands on the guest who just made the point rather than the host who laughed at it.

The layout logic that follows from it:

  • One speaker on a single-speaker shot: Clippy alternates a punched-in and a wide shot every few seconds, so the clip has movement without a cut to a second person.
  • Two people trading lines: the clip switches to a split screen, each speaker in their own panel.
  • Seam placement: captions sit on the seam between the two panels in the split screen, so the words do not cover either face.

The same tracking runs in the gaming layout, where Clippy locates the streamer’s facecam once per stream and renders gameplay on top with the facecam below, captions on the seam. Podcast layout is the version built for talking heads.

Split Screen When Two People Trade Lines

The split screen is not on for the whole clip. It appears when the conversation moves between two people and goes away when it settles on one.

Take a 12 minute segment where a host asks a question and the guest answers for 40 seconds, then they go back and forth for 20 seconds. The answer is rendered as alternating punched-in and wide shots of the guest. The back and forth becomes a split screen, because that is the stretch where both faces carry the meaning.

This is the difference between a clipper that crops the center of a two-camera podcast and one that reads the room. A fixed center crop gives you half of each face for the entire clip. Speaker tracking gives you the person who is talking, and both people when the exchange is the point.

For the wider workflow of turning a finished clip into posts on every network, see the guide to clipper captions and scores.

Punched-In, Wide, and Word-Level Cuts in One Clip

Three things happen inside one clip: framing changes, cut points, and captions, and they all come from the same transcript.

Framing changes are the punched-in and wide alternation on single-speaker stretches and the split screen on exchanges. They give the eye something to follow on a vertical screen.

Word-level cuts mean no clip starts mid-word or ends mid-word. The editor pass snaps each cut to exact word boundaries in the transcript. Up to 3 dull stretches can also be cut out of the middle of a candidate, which is how a rambling 90 second answer becomes a tight 35 second clip without a jump cut you can hear.

Captions are transcribed, then proofread against the source to fix misheard names, places and slang, and rendered word by word. You have 8 caption design presets and 12 fonts, with control over colours, position, words per line and a pop or box highlight. A hook title draws as a TikTok-style rounded text box over the first seconds. Optional silence cutting and audio normalised to -14 LUFS round it out.

If a clip scores just under your posting bar, auto polish applies the Virality report’s fixes once, re-renders, re-scores, and keeps the better version. You can also fine-tune any clip by saying what to change in plain words; Clippy plans the edit and re-cuts just that part.

How to Set Up Speaker-Aware Clips in Clippy

Setup is a source, a request, and a review, and you can drive all of it from the API or the MCP server if you work with an agent.

  1. Pick the source. Find the creator and the stream or video, paste a link, or upload the file.
  2. Ask for clips. Request up to 10. Clippy scouts more candidates than you asked for and keeps the best scored ones.
  3. Set the bar. Rate clips Good clip or Not for my page. Once 3 or more clips have either, Clippy rewrites what it learned about hooks and picks, and that steers the next runs.
  4. Post or schedule. Send the clips out with a caption and cover, or let an autopilot do it.

An autopilot watches creators you choose, clips each new stream, and posts clips scoring at or above your bar (default 67) to the accounts you assign. Each account gets its own schedule: posts per day (default 6) inside a posting window (default 8am to midnight in your timezone), slots spread evenly and never closer than 2 hours apart. The same video is not clipped twice for the same pages.

An autopilot can also post as Instagram Trial Reels and attach a comment-to-DM funnel with a keyword and link on every clip it posts. That funnel is covered in the podcast clip workflow post and the comments-to-DM docs.

Podcast Clipper Speaker Tracking vs Manual Editing

The trade is control for throughput: a manual edit gives you every frame, and speaker tracking gives you 10 scored clips from a 3 hour stream without touching a timeline.

Job Manual edit in a timeline Clippy podcast layout
Find the moment Scrub the episode, or read the transcript Claude reads the transcript and scouts roughly 3x the candidates
Pick the framing You choose per cut Mouth movement picks the speaker; punched-in and wide alternate
Two guests talking You build the split screen Split screen appears when two people trade lines
Cut points Your ear and the waveform Snapped to exact word boundaries, up to 3 dull stretches removed
Captions and hook Separate transcription and design pass Word-by-word captions proofread against the source, 8 presets
Volume One clip at a time Up to 10 clips per video, plus autopilot on new streams

Manual editing still wins when a clip needs a judgment call no rubric captures. Clippy is built for the volume case: a podcast that ships weekly and needs five clips every week, on schedule, in the same look.

For the code path, the API group is /v1/clipper and the MCP server exposes clipper_ tools on every plan. The clipper API walkthrough shows the request shape.

How Creator Link Tracking Fits Your Podcast Clips

Speaker tracking gets the clip watched. Link tracking tells you which clip earned the click.

Creator OS creates trackable go.creatoros.ca short links with click stats, and you can give each clip, each guest, or each episode its own link. Put the link in the first comment, or in a comment-to-DM keyword funnel, and read clicks per clip next to its views.

The loop closes in the analytics tab, which shows views by posting day, per network, and every posted clip with its stats and links to the live posts. Pair that with per-post analytics and Reel watch stats and you can see whether the split screen clip held attention better than the punched-in one. The MCP tools links_create and links_stats let an agent create and read those links, and get_post_analytics returns the watch numbers.

One rule to keep straight: you need the right to post clips of someone else’s content. Follow each creator’s clipping rules and each platform’s policies.

Frequently Asked Questions

What is podcast speaker tracking in a clipper?

It is the step where the clipper works out who is talking in each moment of the footage, usually from mouth movement, and uses that to choose the framing. In Clippy it decides between a punched-in shot, a wide shot and a split screen for every stretch of a clip.

Can a podcast clipper automatically make split screen clips?

Yes, in Clippy the split screen is automatic and conditional. It appears when two people trade lines, with the captions placed on the seam between the two panels, and it goes away when the conversation settles on one speaker.

How does word-level cutting change podcast clips?

Cuts snap to exact word boundaries in the transcript, so a clip never starts or ends mid-word. Up to 3 dull stretches can also be removed from the middle of a candidate, which tightens a long answer without an audible jump cut.

Is Clippy’s podcast clipper available to everyone in 2026?

No. Clippy is available on every plan, and the Virality agent is available with it. Check the Creator OS docs for details.

How do I track which podcast clips drive signups?

Create a trackable go.creatoros.ca short link per clip or per episode, then read click stats next to the clip’s views. Creator OS provides the posting and the numbers; your funnel and landing page decide what converts.

Get started

Creator OS plans start at $14.99/month for 11 platforms plus Skool, with the API, the MCP server and the CLI included. Start at https://www.creatoros.ca/sign-up.

To see the clipper and the agent workflow running end to end, watch the walkthroughs on the KevBuildsApps YouTube channel and the Creator OS launch video. New subscribers start for $0 with a free trial and can cancel anytime.

Get started for $0
Free trial for new subscribers. Cancel anytime.

Keep reading