YouTube Studio AI clip suggestions: 10 countries; clips elsewhere

YouTube Studio can turn a segment of your long video into a Short or clip, and its AI suggestions for the best moments are limited to creators in ten countries and select languages. If you are elsewhere, or your podcast is in an unsupported language, you can do the same job with Sume: transcribe with video inspect at $0.01 per audio minute, pick the ranges yourself, cut each with video trim at $0.02, and burn captions at $0.20 per clip up to 60 seconds.

The YouTube side comes from its Video Clips and Shorts in YouTube Studio help page, read on 2026-10-02. The Sume side comes from video inspect, video trim and video captions.

Where are YouTube's AI clip suggestions available?

The help page says AI-powered suggestions for clip extraction are currently available for creators in Brazil, Canada, France, Germany, India, Mexico, South Korea, Spain, the United Kingdom and the United States, and in select languages. It does not publish the language list, and it gives no minimum or maximum clip length.

It also says the creation path is Content, then Videos, pick the source, then Clips and Shorts, then create a clip or Short, preview the auto-generated captions and create a draft. And it says creators will not be penalized for using their own content in a clip or Short.

How do you find the moments yourself?

Run POST /v1/video-inspect with frames: false and transcribe: true, and add segmentation.mode: "sentence" to get caption-shaped sentence segments. Pass language_code if you know the language, since it is an STT hint and the default is auto-detect. The transcription hint is capped at 600 seconds, so cut a long episode into pieces first, which the 1800-second source cap post explains.

Then choose moments with a reader's eye: a question, a surprising claim, an answer that lands in under 45 seconds. If you want a model to propose them, a typed clip plan from an Agent Completion returns start and end times you can feed straight into trim.

Pick the number of clips before you read the transcript. Five 30 to 45 second clips from a one-hour episode is a usual target, and it keeps the reading time small. Write each pick as a start second, a duration and a one-line reason, so a teammate can approve the list without opening the video.

How do you cut and caption one clip?

Trim takes start and exactly one of end or duration. Leave precision at exact for a frame-accurate cut, and pass an output of { "width": 1080, "height": 1920 } if the source is landscape and you want the clip portrait; the output is conformed to those pixels, so check the framing of a landscape source before you publish. Each trim is $0.02.

Then caption the trimmed MP4. Passing style is optional: omit it and Latin wording gets slam, Korean gets black-outline. Pass script_text when you have a corrected transcript so the words match your spelling.

Keep the exact timings in a small table or sheet and give each clip an ID used in its Idempotency-Key. A rerun after a failed job then reuses the same key and does not create a duplicate trim.

curl -X POST https://api.sume.com/v1/video-trim \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: ep42-clip-03-trim" \
  -d '{
    "video_url": "https://media.sume.com/artifacts/artf_demo/episode-42.mp4",
    "start": 1312.4,
    "duration": 38,
    "precision": "exact"
  }'

What do you give up compared with Studio?

Studio's tool sits inside your channel: it picks from the transcript, drafts captions and publishes in one place. Sume does not publish to YouTube, does not rank moments by itself and does not read your analytics. You carry the clip from the result URL to YouTube yourself.

One more point: YouTube's originality rules for reposted clips changed on October 1, 2026, and our post on clipping your own long video covers what counts. A clean cut with a fresh hook helps with that, but it is not a guarantee.