Does a voiceover make a reused YouTube Short original?

Not by itself. On October 1, 2026, YouTube's Creator Liaison Rene Ritchie said, as reported by Relevant Audience, that remixers should not rely on voiceover descriptions of what is happening on screen, minor technical edits, or template-based bulk changes. Laying a voice track over someone else's footage is the first thing on that list, so it is not a safe way to make a reused Short count as yours.

What YouTube encourages instead is your own commentary, analysis and storytelling. Sume cannot decide for you whether a Short is original, and nothing in its API produces an originality score. What it can do is make the footage and the voice yours from the start, so there is nothing borrowed to narrate over.

What exactly did YouTube say about voiceovers?

The facts below come from news coverage of a Creator Insider video, not from a YouTube help page. As reported by Relevant Audience, 9to5Google and Music Ally (all read 2026-10-03), the change updates the recommendation system to reduce the reach of channels that mainly aggregate or re-upload clips without significant changes.

The same coverage says the update concerns recommendations only. Monetization stays under YouTube's existing reused-content policy. No start date, timeline or dashboard showing whether a channel was affected has been reported.

What does YouTube's own policy page say about narration?

The monetization side is a published help page. YouTube's reused content policy (read 2026-10-03) allows reaction videos with commentary, critical reviews that use clips, and edited footage with an added storyline and commentary. It lists non-verbal reactions without voice commentary, content copied without substantive modifications, and readings of unoriginal material such as websites or news feeds as not allowed.

Read together, the pattern is consistent: the page rewards a contribution that is visible and substantive. A narration that only restates what the viewer already sees adds none, which matches what was reported on the recommendation side.

How do you make the footage yours instead?

If the clip itself is the problem, replace the clip. Sume's Video generation API makes a new video from a prompt, with aspect_ratio and duration fields. The docs say Auto create controls default to 720p and 8 seconds, with 3 to 10 second clips at 16:9 or 9:16, and that limits vary by catalog model (Video generation, Video Router). This script submits one vertical clip and prints the download URL:

import os
import time
import requests

key = os.environ["SUME_API_KEY"]
h = {"Authorization": f"Bearer {key}"}
job = requests.post(
    "https://api.sume.com/v1/videos",
    headers=h,
    json={
        "model": "sume/auto",
        "prompt": "Handheld vertical shot of a ceramic mug on a workbench, steam rising, morning light",
        "aspect_ratio": "9:16",
        "duration": 8,
    },
).json()
while True:
    time.sleep(30)
    s = requests.get(job["polling_url"], headers=h).json()
    if s["status"] == "completed":
        print(s["unsigned_urls"][0])
        break
    if s["status"] == "failed":
        raise SystemExit(s.get("error", "failed"))

Check GET /v1/videos/models before pinning a model, because durations are not uniform. YouTube's Shorts upload page (read 2026-10-03) lists Shorts of up to 3 minutes, vertical, with a maximum resolution of 1080p, so a few stitched 8-second clips on Sume's Timeline fit well inside it.

If your source is your own longer recording rather than a generated clip, Video trim cuts a range into a new MP4 and leaves the source untouched. Whether YouTube treats Shorts cut from your own long video as original was not addressed in the announcement; see Clip your own long video into Shorts.

What should you do with the voice?

Keep the voice, change what it talks about. A voice that explains an opinion, a result or a method is commentary; a voice that reads out what a stranger's clip already shows is not. Write the script from your own point of view first, then choose footage that supports it, rather than starting from a clip you found.

One disclosure point applies. YouTube's disclosure page (read 2026-10-03) asks you to disclose realistic content that makes a real person appear to say or do something they did not, or a realistic scene that did not occur. It lists cloning your own voice and caption creation among edits that need no disclosure, and says disclosing does not restrict audience reach or monetization eligibility.

• Do: shoot or generate footage you can show the origin of.
• Do: write a script that argues or explains, not one that captions the picture.
• Do: disclose realistic AI scenes when the page says to.
• Don't: expect a voice track, a crop or a speed change to turn a stranger's clip into yours.
• Don't: assume the reported update is limited to monetization; it is about recommendations.