Gengar narration: what should each sentence include?

Questions

1. Should we always name the hands, or only when it helps explain the action? Do we need to describe the supporting hand too?
2. Should picking up, turning, and replacing an object be three separate labels, or can they stay together in one sentence?
3. Should every repeated movement be narrated, or can we describe the repetition once and speak again when something changes?
4. Are separate recordings per segment okay, or do you also need one continuous audio file? Are pauses okay?
5. Should each segment start and end with the action, even if the sentence takes longer to say? How precise do the timestamps need to be?
6. Should we clean up grammar and use consistent object names? When a hand or object is obscured, should the narration mention the uncertainty, or should we leave a separate note for review?

Reference comparison

The three reference videos all describe what the person does, what they handle, and where they put things. They differ in how much detail they give about hands and individual movements.

We compared all 86 transcript segments and checked nine frames across the videos. The examples below come from that comparison; we haven’t checked every action against the footage.

Hands

The camera-cleaning narration names the left, right, or both hands in 28 of its 33 segments. Shopping and crocheting never explicitly name left/right/both hands, even though both hands are visible doing different things.

For example:

• Shopping, 00:07–00:21: picking up, flipping, and replacing a lemon-patterned notebook. Both hands hold it at 00:10, but the narration doesn’t mention which hands.
• Camera cleaning, 01:51–02:02: picking up the cleaning cloth with the right hand. The frame at 01:55 matches that description.
• Crocheting, 00:14.9–00:29.1: putting yarn on a finger and pulling through a hole. At 00:22 one hand holds the yarn and the other holds the hook, but the narration doesn’t distinguish them.

Our SOP says to name hands when visible and useful. The narrator brief says to name them whenever hands are involved.

Actions and objects

The camera example breaks handling into specific steps: opening the battery lid, pressing the release, removing the battery, and placing it on the desk. Shopping also includes walking, looking at products, turning them over, and putting them back.

That suggests sentences should explain the actual movement and object, rather than just say “I cleaned the camera” or “I went shopping.”

A possible sentence would be: “I held the camera with my left hand and opened the battery compartment with my right.”

Length and repetition

The reference transcript segments vary in length:

Fourteen segments exceed 12 seconds. Some contain several actions; others split a sentence across segments. The narrator brief suggests a sentence every 5–12 seconds, while the SOP says to describe repetition naturally rather than repeat sentences at fixed intervals.

Crocheting describes repeated yarn-over and pull-through movements, sometimes counting them.

Recording and timing

The SOP asks for continuous narration. We currently support recording one segment at a time.

The SOP also says timestamps should match the action in the video, which may happen at a different time from the spoken sentence.

Wording and unclear footage

The examples mix tense and terminology. Crocheting, for example, uses “hole,” “hoop,” and “loop.”