YouTube celebrates Deaf Awareness Week by killing crowd-sourced captions

• 

  Here's what the community caption feature looked like. [credit: YouTube ]

Today's the day YouTube is killing its "Community Contributions" feature for videos, which let content creators crowdsource captions and subtitles for their videos. YouTube announced the move back in July, which triggered a community outcry from the deaf, hard of hearing, and fans of foreign media, but it does not sound like the company is relenting. In one of Google's all-time poor-timing decisions, YouTube is killing the feature just two days after the International Week of the Deaf, which is the last full week in September.

Once enabled by a channel owner, the Community Contributions feature would let viewers caption or translate a video and submit it to the channel for approval. YouTube currently offers machine-transcribed subtitles that are often full of errors, and if you also need YouTube to take a second pass at the subtitles for machine translation, they've probably lost all meaning by the time they hit your screen. The Community Caption feature would load up those machine-written subtitles as a starting point and allow the user to make corrections and add text the machine transcription doesn't handle well, like transcribed sound cues for the deaf and hard of hearing.

YouTube says it's killing crowd-source subtitles due to spam and low usage. "While we hoped Community Contributions would be a wide-scale, community-driven source of quality translations for Creators," the company wrote, "it's rarely used and people continue to report spam and abuse." The community does not seem to agree with this assessment, since a petition immediately popped up asking YouTube to reconsider, and so far a half-million people have signed. "Removing community captions locks so many viewers out of the experience," the petition reads. "Community captions ensured that many videos were accessible that otherwise would not be."

Read 1 remaining paragraphs | Comments

from Tech – Ars Technica https://ift.tt/3cDTdH7