Netflix introduces a new kind of subtitles for the non-hearing impaired | Hacker News Reader

Netflix introduces a new kind of subtitles for the non-hearing impaired

arstechnica.com

31 pointsfurcyd47 comments

Article tagline: “Just the dialogue—no sound effects or music cues.”

Do any HNers find those cues sufficiently distracting that they’re excited about this feature? I always have subtitles on. While I find those cues occasionally amusing, I’ve never found them to be an impediment to my enjoyment. Maybe they are for people who read slowly? Or because they break the fourth wall?

Eh, I don't need help hearing background noise. I need help understanding the words coming out of the actors mouths because the mix is wrong or directed that way.

Exactly the same issue.

Quiet parts loud, loud parts quiet.

To be honest I usually think they're tacky because I imagine they detract from the experience of nearly every single person.

It's the kind of thing where there is a horse-drawn carriage going by on screen and they print the text "[clip clop clip clop]". If you can't hear it, wouldn't you just want to watch it?

Sometimes the sound can be offscreen, e.g.:

[Window breaking]

[Gun cocking]

[Stairs creaking]

Yeah those make sense to me, because they're part of the plot.

The weird ones for me are where some absolutely unintelligible background murmuring is subtitled. I couldn't tell that any actual words had been spoken and suddenly it's all spelt out.

yeah I totally get it for moments like that where something that you can only hear is critical to understanding what's going on in the movie, but they never seemed to care much about making those kinds of artistic judgements

The funny thing is that I've been seeing this for decades. E.g.: media with a separate "subtitle (audio only)" and "subtitles (with sounds)" tracks.

Not sure why this is being presented a something new here.

This is the difference between just "subtitles" and "closed captioning / subtitles for the Deaf or hard of hearing". Usually one will be labeled as "English" and the other as "English (CC)". Other streaming services have had this distinction and you can also see this classification in places such as open subtitles. Netflix has always been the odd one out to only include "English (CC)".

This is Apple introducing the new iPhone, with the world-turning new features "copy", and "paste" all over again.

Pro Max buyers get "cut" too.

CC are not meant for deaf. They are exact transcripts of speech. They are better for language learners and foreigners who occasionally need to glance at subtitles to get missing word. They dont have advantages for dead afaik.

They real advantage is that you don't need to create priper subtitles that are subject to different standards. So, they are cheaper.

Are you nitpicking or what do deaf people normally?

Deaf are better off using normal subtitles, because those are literally designed to be read while watching the movie.

They have different standards for stuff like how many letters to show on screen at the same time and for how long.

What do you mean by normal subtitles? I have a few Deaf friends and they all like the CC version as sometimes they miss out on the audio cues. Some scenes make no sense without it.

yes they are

The ones that describe the music are funniest. Cheery music! Foreboding music!

I've seen this option for a long time, I always assumed they create them as the starting point of subtitling in foreign languages. Presumably nobody is French, hard of hearing and wants to watch an English language program.

As someone that reads faster than people general speak, I find subtitles very annoying in a language I understand. But I live in a multilingual household so I keep them on for nearly all English content because it is genuinely helpful for others. Anything that gets some of the text off screen is a positive for me

Yeah I hate them.

I expect the target audience for this change is people whose reading speed for whatever reason can’t keep up with dialogue when sound effects captions are included. There are a variety of reasons for that and it’s sensible to provide “sfx/bgm on/off”, though it shows how little investment is being made in accessibility by the media industry that Netflix has to do this in their language selector rather than the TV being able to do it for all feeds.

If you are discussing the movie as you watch it subs are nice to not miss dialogue. And disturbance in general.

I don't care about that. But being presented the choice, I would choose the subtitles without (loud bang). And having a choice is nice, isn't it? So yes, I appreciate the effort.

Me for one

Frankly, I'd rather that content producers and their sound mixing and mastering engineers just improve the audio levels of their dialog tracks in relation to the rest of the mix. I already have my center audio channel goosed +3db beyond the professionally calibrated academy reference levels to help with legibility. The problem is that mixes are increasing in dynamic range, so if I elevate the center channel level (where most dialog is) even more to lift the occasionally inaudible utterance, some of the rest of the center channel content is too loud in relation to the other channels.

While my high-end home theater audio system has DSP functions like dynamic compression which I can apply to smooth out the center channel, that's relying on a blanket algorithm in an attempt to fix something that's much better fixed on the mixing stage in the first place. The mixing engineers have much better tools which they can selectively apply when needed and even have the option of problematic dialog being re-looped if necessary. It's their job to get this right and they certainly have the all the tools and training to do it properly and only when, and as much, as actually necessary. Having home viewers slap some auto-mode plug-in over the entire run time of a sound mix that was painstakingly hand-mastered scene-by-scene is objectively the worst way to solve the problem.

This simply makes no sense because the entire modern signal chain is digital. There's no technical reason the gain shouldn't be correct. The fact that dialog audio levels are still a recurring problem in my properly calibrated, 11.2.4 THX-rated dedicated home theater must be that the audio engineers aren't being given the time and resources to do their job properly or are being instructed to do this as some misguided aesthetic choice (looking at you Christopher Nolan).

I'm really glad you posted this, it's good to know that I didn't mess up the calibration of my (much humbler) 3.1 system.

AppleTV 4K box with the new Enhance Dialogue feature has been working great for me.

Raising the center channel never sounded great to me, because there are so many other sounds in the center channel as well.

At least now I can raise only the dialogue levels on demand.

I have no doubt that Dialog Enhancer features, which are available on many streaming devices and AVRs, can work well. They are basically variants of multi-band dynamic compressors tuned to target vocal frequencies on the center channel.

Both my streaming device and AVR each have their own flavor of dialog enhancer and they do help improve the problem substantially. But I don't feel 'good' about just leaving this kind of post-processing feature on long-term and certainly not as a widely recommended best practice (except for hearing impaired users). I guess the reason for my reluctance is more in principle than practical. It's just not the right technical approach to solving the root problem, and any automatic algorithm, no matter how adaptive, dynamic or clever, will sometimes fail to do the right thing. It also bothers me that every implementation of dialog enhancer I've seen is opaque and undocumented.

I've been in and around video and audio engineering for most of my career, especially on the broadcast tooling side creating new gear. I go way back to the analog days when we could only dream of having such a pristine digital signal throughout the entire signal chain. We were always battling generation loss and noise on the production side and on the home theater side, we spent time and money slathering various noise reduction and other 'enhancement' processing to fix the worst aspects of the degraded analog signals we received.

At long last, today we live in that sci-fi future nirvana where the same levels the director approved on the mixing stage can be exactly what you hear at home. So, yeah, I have some PTSD about still needing to slap some post-process on at home to fix issues that should not be there in the first place. We waited decades for this tech to arrive, so the engineer in me wants it to work - without fixes or user patches. And that means addressing the issue at the source.

People very much need this. Making them visually more fitting, less distracting, would be a nice next step.

This won't replace the mixer I run my audio through, though. A bit of a pain, and you need a mixer, but try it if you can.

Wait, is this really a new thing...? I could've sworn I've seen it in streaming services before, where you could choose between dialog only and closed captioning. One is usually called "English" and the other is "English SDH" or similar, where SDH is "subtitles for the deaf and hard of hearing". It's the exact difference this article points out (sound effects and music etc., as opposed to just dialog). https://www.3playmedia.com/blog/whats-the-difference-subtitl...

Video games have also had this as an explicit option for many years, maybe decades (Half Life 2 comes to mind).

Not only did Netflix not invent this, I could swear some of their shows have had it for years already...

The linked Netflix support article introduces a new distinction:

English (CC) is described as separate from English; and then, a sentence or two later, they use the phrase “CC/SDH”.

So, Netflix is changing their definition of the word “subtitles” to represent dialogue-only captions, as a distinct thing from CC/SDH “captions” — even though the S in CC/SDH stands for subtitles.

It might be interesting to look to the .SRT file communities and see whether user-produced contributions contain sfx/bgm subtitles or merely dialogue subtitles, as that could reveal a similar shift in definition/usage prior to Netflix’s press release about their own.

I don't think I've ever seen user supplies SRTs that contains the CC m-style music and sound cues (though they make up for that by including some NordVPN ad at the beginning or end...)

Yes... but that isn't new. It's been around for years.

Correct. (?)

>The performance style of actors in current TV shows and movies is more naturalistic and less elocutive than it once was, so characters are more likely to speak softly.

Yeah, I've given up on new movies. If I want naturalistic without a storyline, I can just go outside. I'm not going to turn the volume up to the level required to hear the mumbled whispers, just to have my hearing damaged by the next gunshot or blast of music.

Subtitles aren't a fix -- they're distracting and therefore annoying. A selectable "classic" audio track would be the fix.

It is bad audio processing; "too good" rather, as if the silver screen was the only place anyone watched movies.

It's not even that, I couldn't hear large parts of Oppenheimer, having paid a hefty premium for a cinema with a good sound system.

I think they mix movies differently than they used to, maybe because people all seem to have soundbars today?

It is absurd to call it naturalistic. If people can understand others in real world situations, but struggle to follow them in a movie, it was not naturalistic.

but that's the unrealistic part. in the real world people don't understand each other all the time, but all those "sorry, what did you say?"s would be annoying on screen (like so many other things that are skipped in movies). actors don't talk naturally at all, it just seems to us that way because we're used to them talking like that on screen.

Oh my goodness it only took like ten years but they finally did it!

one would think this was an explicit usecase/requirement 10 years back, nor was it obvious that this was needed.

Progress is always slow but inevitable.

The problem I frequently run into on various platforms is shows with some occasional foreign language that is meant to be subtitled. I end up getting either no subtitles (and it takes me a while to realise that I'm meant to understand) or I turn subtitles on and suddenly all the English is subtitled too, which I don't want.

I don't get why this is complicated.

All these streaming services need better down mixing from surround sound to stereo, so often people can't hear the dialogue because it's been mixed for a centre speaker which becomes too low in the downmix to 2.0/2.1.

While YouTube still displays it word by word, forcing you to not look at the video.

Cheers to Netflix in the age of user hostile and anti ergonomic patterns (while they still have their fair share of sins in this area, at least one good news)

Source: https://news.ycombinator.com/item?id=43788473

Boost spoken audio + adjust playback speed may get me to watch more movies and tv shows.

A bit off-topic, but one gripe I had with some implementations of subtitles is when they’re delivering on the identity of a person before this was exposed. I don’t know why this happens that it would reveal <Name: Dialog> where it should instead be <Generic trait: Dialog>.

So instead of <Phil: Blah blah> it would be <Raspy voice: Blah blah>.

I don’t have examples for this, but it happens regularly that I noticed it as a trend. Might be because subtitles are outsourced and maybe the importance of exposition is not clear to the people creating the subtitles?

A very few shows are attentive enough to this that the character name will change as plot is revealed within a scene. Most, sadly, are not.

Reply on news.ycombinator.com