Start with a complete moment

A short clip should make sense to someone who never watched the stream. Look for a setup, an action and a payoff: a disagreement, an unexpected mistake, a reaction or a difficult play. A loud sentence alone is not a story.

Upload the recording to the library and open Editing. Confirm the import has finished, its duration matches the source and audio is present. File size alone does not prove a recording is complete.

Turn speech into a useful transcript

Transcription creates text with timecodes. Those timecodes help you find lines and choose boundaries. Processing speed depends on recording length, model and available resources; use the actual job status rather than an estimate.

Keep the original source after transcription. Saved speech can be reused for another cut of the same recording with compatible settings. A changed source or model may require new processing.

Check the setup and the payoff

Watch a few seconds before the proposed start. One question can explain the entire reaction. Check the ending too: avoid cutting a word in half or ending before the reply.

Remove pauses deliberately. A short silence before a joke may be part of its timing. Automatic suggestions still need an editor’s final decision.

Prepare the vertical frame

Choose a layout first. Put the camera above the gameplay when the creator’s reaction matters. Without a camera, give the action more space. Check that cropping does not hide the target, interface or captions.

Watch the export with sound on a phone before posting. Check wording, audio, boundaries and your permission to use the material. A moment score is not a promise of views.