Apple's new SpeechAnalyzer API, benchmarked against Whisper and its predecessor | Hacker News Reader

Apple's new SpeechAnalyzer API, benchmarked against Whisper and its predecessor

get-inscribe.com

570 pointsget-inscribe238 comments

Whisper is the wrong model to benchmark against, or rather, there are better models that are state of the art now like Nemotron and Parakeet both by Nvidia, as well as Mistral's Voxtral and Cohere Transcribe.

However, what's funny is, RIP to a lot of the paid apps that simply wrap Whisper, I'm sure Apple will make a native GUI such as a recorder app for macOS that obviates the need for these wrappers, which everyone seems to be vibe coding these days.

For multilingual and noisy audio the best right now is MOSS-Transcribe-Diarize which was released just a few days ago

Superwhisper does a lot more than just provide a whisper/parakeet UI so I’m not sure Apple will destroy them so easily

Thanks, was looking at a better diarization model.

Even for those sorts of apps, MacParakeet which I've been using is FOSS so no payment needed. In reality these days with AI the ability to spin up a free and/or OSS competitor falls to zero.

I’m not even using it for diarisation just transcription and it’s amazing. It also doesn’t need a VAD

A new VAD I found though is FireRedVAD which has better benchmark results than TEN and Silero by far

Is MOSS a streaming model or only for offline? For that VAD how are you integrating it into a model like Whisper etc?

I'm using it offline. But it's much faster than realtime so it should be usable for streaming. I just asked Codex / Sol to integrate FireRedVAD with Whisper...

Interesting! And what would you say are MTD top competitors?

I'm only dealing with Japanese audio so for me it's Anime-Whisper, a Japanese-specific fine-tuning of Qwen3 ASR, and Apple SpeechTranscriber.

16GB! (edit: this is wrong, I was looking at TTS, the transcribe model is 1.7GB). Compared to Parakeet 2.3GB (but no diarization).

Also doesn't seem to be tailored to Apple hardware (i.e. no MLX or ANE variant/implementation)

I am running the MLX fork https://huggingface.co/vanch007/mlx-MOSS-Transcribe-Diarize

Generally labs don't release MLX or ANE versions and we must rely on finding someone who's converted it

Parakeet is not multilingual so not directly comparable

Where do you see 16GB? MOSS is smaller than Parakeet at 1.82GB

Thanks, I didn't see vanch007 version at first (only ~30 downloads), I usually look at mlx-community. For the size I was looking at the wrong model (TTS not transcribe-diarize), thanks for the corrections.

> there are better models that are state of the art now like Nemotron and Parakeet both by Nvidia

Is parakeet state of the art? It always transcribes speech fragments for me, like if I stutter and say "m-m-m-map" parakeet will dutifully transcribe "m m m map". Which I guess could be a good thing or a bad thing depending on what you want. Whisper does not do that however.

I do like cohere transcribe a lot.

Agree on this point. Recent anecdotal testing I did found Whisper is still better than Parakeet.

Parakeet is certainly faster on my machine (m3 max), but I can't stand using it over Whisper for dictating my prompts. It makes a lot more mistakes, possibly because (like you mentioned) large portions of the speech will pause / stutter as I think about what to include.

With whisper v3 turbo, I can almost always live with the few mistakes because the overall stream-of-thought word-salad I provide is still clear at a high level. The bits and pieces of context seem to help, that I might leave out if typing and focused more on traditional conciseness / clean writing. With parakeet, I needed to do frequent editing even for shorter bits of speech.

I realize some applications prioritize the latency.

It sounds like post processing should be the job of an LLM. I would like the voice model to be faithful to what was said and then that output can be smoothed over or postprocessed as needed for the use case

To be clear, I'm talking about high word error rate with parakeet vs whisper, not post processing and cleaning up my speech. Re: being faithful to what was said, one small example, Whisper will often put ellipses when I pause.

The near-instantaneous nature of Parakeet has led me to keep using it and occasionally simply re-dictating a second time as needed.

For round two after a typo-laden transcript, I’m dictating and annunciating with great passion as I read the first transcript to make sure I don’t miss a beat. It’s kind of fun because it’s a little performance.

I use Parakeet V3 via this tool and it is actually quite reliable for me (in English): https://github.com/cjpais/Handy

If you are using Parakeet for English only then you should be using V2. V3 is for several languages and is worse at English only.

I think that's parakeet doing its job there. That is a closer reflection of what you've actually said. The trick is then throwing that output through some additional deterministic and non-deterministic steps to tidy it up however you prefer. It's exactly what I do with my free and open source dictation app (dictator.robgough.net) for Mac+iOS. And of course, everything stays entirely on-device. Gemma E4B is really wonderful for that second step, it's great at language – but takes up 6-7GB RAM.

Oh yeah, this one’s worth hyperlinking

https://dictator.robgough.net

Great work Rob! Indeed private as promised per App Privacy Report, “Domains contacted directly by app”:

  cas-bridge.xethub.hf.co; huggingface.co; mzstorekit.itunes.apple.com

Site could identify our device and send iOS visitors to the iOS page (or maybe that’s against the vibe and we should tap it ourselves).

App might be able to launch the keyboard settings directly but I suppose Apple doesn’t like devs using those undocumented URIs (uhg but maybe can understand part of it).

Keyboard, given manual app switchbacking + manual pasting, is less convenient for some of my use cases compared to an action button shortcut. (Reference Spokenly w/Local-Only Mode.)

Separating dictation and the history, and having a syncable scratch pad, are some welcome innovations!

>Gemma E4B … takes up 6-7GB RAM.

Google has a spyware-adjacent dictation app (maybe not really, but they demand to connect to servers after you enable their offline toggle). Just like yours, they thought of a cute name too (Google AI Edge Eloquent). Do you know what language model they install on the iPhone? Not a very good one but I’m sure over the next year or two…

I appreciate the kind words. Thank you.

I really wanted to have background audio and make it so the keyboard would directly record audio etc, but my first pass didn't make it through app review (and that was just keeping background audio listening AFTER you'd already started a recording). I could maybe have fought it, but figured if I was already butting up against app review there was little point as they'd likely reject in a future release anyway.

Re: analytics, It is quite weird having no idea how many people are using the app. It does leave you a little blind, but I figure people will get in touch if they have big enough problems with it.

For the Google app, I believe Gemma E2B and E4B both have audio input, so I suspect they're using one of those.

Analytics: if you generate them on my device, and bury an option to manually send them to you when I have a bug--or...

you instead bury an opt-in to automatic telemetry, and let us CC ourselves each ~month when the analytics get sent in to you (so we can verify it's all boring data)...

It's hard for me not to opt in to stuff like that, at least periodically. If I'm opted-in by default? Ehhhhh.... I totally get it, weird "flying blind" (quoting a different dev w/similar philosophy), but I guess I'm still weird myself and hence looking for that autonomy or something? Oh and if somebody makes something opt in, and then certainly if they furthermore stick the option somewhere slightly off the beaten path, that seems pretty darn trustworthy.

(I wonder if we'd have enough public data for a decent statistician to calculate a likely number range of how many users you have by extrapolating from the 1% to 10% or 20% of users who'd chose to opt in...)

PS: "bury" concept likely not actually important :)

No, there are better open weight models: https://artificialanalysis.ai/speech-to-text/non-streaming

Apparently MOSS-Transcribe-Diarize is quite good too as it released only a few days ago.

Also, this test is English-only, while a strong point of other models is to understand different languages without first having to say which one (so you don't need 3 different keyboard shortcuts if you wanna dictate in 3 languages day-to-day)

As an Australian, Apples voice models have always sucked. I've tried using stt (again) more recently and its improved, but i'm so tired of having to Americanize my voice to get it to figure out what the hell i'm saying.

It also struggles with my NYC-area accent, which is only medium thick.

The Two Yoots problem. Do you use d's in place of t's such as dees/dems/dose/dere? I have a heavy queens accent so you'll hear me say things like "deres tree uh dem ova dere."

Surprised to read this as a Queens thing, this sounds perfectly at home in the Midwest.

Yah thats definitely Minnesota, eh?

Oh jeez

It's spoken fast with a nasal tone.

As a Texan first, American second, I sympathize with this statement. Siri can't understand me probably 25% of the time. I use STT for iMessage while in the car, and half the time it will take 3+ times to either get it right or me give up, and hope to remember to text them by hand when I next stop.

What does this mean? When I pledged my oath to become a citizen, I had to promise to put America above all other allegiances. Is it in relation to allegiance in the general sense or some weird statement that only relates to STT?

I took it to mean “as someone with a Texas accent”

When they say everything's bigger in Texas, maybe that includes the patriotism.

Texans (generally) consider themselves Texans firsts. Similar to born and raised New Yorkers (NYC). That said, I really meant it tongue-in-cheek that my accent was too thick, and doesn't work well for Siri's STT model. ChatGPT's does way better, but I really don't like talking to an LLM.

Yeah, but I had my fingers crossed behind my back. I hail from, and only stand for, the Republic Of Scouse.

(and if you think Texans have it bad regarding being understood, try being a Scouser [1])

1: https://www.youtube.com/watch?v=LJ5Isx6Hi0E

As a Brit, I concur.

That checks out.

Does anyone have any experience with Mandarin STT? What's a good model for this? The use-case I have is subtitling of Mandarin speech.

MOSS-Transcribe-Diarize [0] is by a Chinese team so apparently it's quite good. Try it out here [1].

[0] https://huggingface.co/OpenMOSS-Team/MOSS-Transcribe-Diarize

[1] https://huggingface.co/spaces/OpenMOSS-Team/MOSS-transcribe-...

Thank you very much.

I'd also look at the Qwen ASR models, also Chinese made, by Alibaba

I use Systran/faster-whisper-medium for real-time subtitling, but you need to get used to the context it's used it and the weirdness it translates into. Parakeet has great mandarin>CN text, but running that + a translation model has been tricky and I never got it fast.

Thank you for this too. I am running in an offline scope so I don't need speed just quality. I'm willing to do it overnight as well if required.

I run it offline too, don't want to depend on separate services.

I use TranscriptionSuite, which is focused on offline transcription, and supports Parakeet as well as whisper and others (https://github.com/homelab-00/TranscriptionSuite) The dev was sweet enough to improve the live mode GUI so that I and some friends can use it when playing on Chinese mmo servers.

Reminds me of the time my neighbours must have wondered if I was having some kind of a breakdown when trying out really basic MacOS voice recognition back in the early 2000s. There was a keyboard shortcut and you could say something like "phone number for firstname lastname" and it would theoretically show you that phone number. Thing is it didn't seem to like a British accent, so I spent a good hour trying out different accents, rotating through various US accents, Australian, South African, Canadian and so on. It seemed to respond best to some kind of a melange of Californian / Australian.

Scottish Elevator - Voice Recognition - https://www.youtube.com/watch?v=NMS2VnDveP8

Not too far off what happened, although thankfully I wasn't actually trying to do anything other than test it. Going to take the opportunity afforded by Scottish TV comedy here, and make a very tenuous link to intercultural exchange so I can post my favourite Rab C Nesbitt scene, hands across the sea indeed: https://www.youtube.com/watch?v=uKxPH_QH940

Thanks for this gem, had a good laugh

Interesting - I don’t think I’ve ever seen anyone from the UK refer to talking in a “British accent” before since we are normally aware of the wild regional variations.

Fair point! I think it's a tic from being English and having lived in Scotland for quite a while so I autocorrect "English" to "British", but I've over-corrected here. (Also perhaps something to do with "British English").

They were probably mainly wondering why you were doing this at their place instead of your own!

Apple's own Voice Memos app already does automatic transcription since macOS 15 / iOS 18.

Speech-to-text is also already built into the keyboard as well, so it can be used in any app where a user would type.

From my experience, Speech-to-text falls way short of Wispr flow and I would assume the ones that are said to be better than that. It lacks context awareness and formatting

I've never been able to record a voice memo on my watch and have it sync to my Mac. It's never once happened. Not sure if I have it configured wrong or what.

Came here to post this. I use handy on my own machine and it's perfect with parakeet. If I switch to whisper it makes lots of mistakes

> However, what's funny is, RIP to a lot of the paid apps that simply wrap Whisper, I'm sure Apple will make a native GUI such as a recorder app for macOS that obviates the need for these wrappers, which everyone seems to be vibe coding these days.

What's insane to me is that you have all of these low-quality me-too apps, and literally no one could bother to read the damn Human Interface Guidelines or follow iOS design conventions.

Doing so is literally LESS WORK than trying to make your own custom awful iOS UI.

Not if your app is a Web wrapper, which so many of these are.

If you use SwiftUI (the native recommendation by Apple), it severely penalizes you, if you want to paint outside the lines (which is a big reason that I don't use SwiftUI for shipping apps). It's insanely easy to write a native app that is 100% in line with HIG.

Of these only Parakeet is <1B, it looks better than Apple's model, however it's not builtin. I wonder how it compares on latency and efficiency.

Parakeet is incredibly fast and accurate even on CPU, and it supports streaming now also in TDT3.

Apple likely needed a model that ran on their NPU natively.

- parakeet usually runs on Bfloat16. NPU doesn't support that

- CPU is not as fast as the NPU for these ops on A-series, and even on modern CPUs, there's a latency delay

- Parakeet latency is fine but "fine" may not be good enough for Apple's UX team.

- CPU increases power consumption over dedicated float blocks

So I would say that Parakeet was a non-option for Apple to ship, although it should be in the benchmarks anyways!

Fluidaudio implements Parakeet on ANE. I'd like to know how SpeechAnalyzer compares in speed.

https://github.com/FluidInference/FluidAudio

Just tried test using yap on a single ~1hr mp3: yap/Speechanalyzer is about 50% slower than fluidaudio on M1. yap interface is nicer though.

https://github.com/finnvoor/yap

Recently contributed a patch to FluidAudio that sped up Parakeet V2 and V3 to 320x and 282x faster than real time, respectively:

https://github.com/FluidInference/FluidAudio/pull/507

That means one hour of audio transcribed in 11.25 and 12.75 seconds.

The Inscribe post doesn't give a speed factor for SpeechAnalyzer. However, this Argmax blog post reports 70:

https://www.argmaxinc.com/blog/apple-and-argmax

Based on that, FluidAudio is ~4.6x and ~4.0x faster.

The difference using an mp3 seems to be smaller: yap seems to use about the same time but fluidaudio seems to take twice as long. Do you happen to know why?

Investigated this and it turned out to be an amusing bug: audio decoding was happening three times instead of just once lol. I've put up a PR to remove the wasteful redundant decoding:

https://github.com/FluidInference/FluidAudio/pull/799

With the updated PR code, ran a test comparing transcribing (using Parakeet V3) a 1 hr stereo 44.1 kHz mp3 vs the same audio in 16 kHz mono wav format. The result was about 21.3% slower with the mp3 vs the wav, i.e. that's the overhead of decoding + resampling.

Currently the decoding + resampling is done up front. If it was done in a pipelined fashion with the inference, that overhead can be eliminated. This is what I did in a recent app I made:

https://apps.apple.com/us/app/drea-podcast-ad-blocker/id6759...

It uses FluidAudio as well, but I forked it and replaced the audio decoding code to (a) use mpg123 instead of the native Apple API and (b) do audio decoding and inference in a pipelined fashion. These two changes effectively eliminated the overhead. mpg123 is quite a bit faster than the native Apple API at mp3 decoding (has some very optimized arm64 assembly routines), and the pipelining ensures that the inference is never starved by the mp3 decoding.

Contributing this pipelined setup to FluidAudio would be good.

That app was exactly what I was looking for, something like SponsorBlock but for podcasts but I suppose using AI for finding the ads works too. Any chance it'll release on Android?

Yep it's something I wanted for a while too; there were existing apps that did this, but had two issues: they were paid, and the UI was subpar. So for mine, I made sure it's fully free and that the UI is on par with Apple Podcasts, Spotify, etc.

Making the ad-finding cheap enough such that I could make it free turned out to be harder than expected. The main issue you run into is dynamic, location-targeted ads. So I came up with a novel technique that uses Shazam-style audio fingerprints for accurate matching, instead of their normal use case, which is identification. This technique is what allows the ad finding to be very cheap, allowing me to make it free.

The SponsorBlock model would actually not work for podcasts, due to dynamic ads. I.e. the location and content of the ads in episodes these days varies by download location. You need the media to be static, like YouTube, for SponsorBlock model to work. Therefore, using an LLM to find the ads + the fingerprints matching in combination is an efficient technique.

Android has def been the most requested thing thus far haha. It'll be a decent undertaking due to me having written the app fully in Swift, i.e. it'll be a complete rewrite. I'll also need to replace FluidAudio with some good, fast Android equivalent.

The goal of making this app was to create something impressive so that I could get a job. Haven't gotten a job yet, but if and when I do, then I'll have time & resources to think about doing an Android version. Currently a bit stressed and occupied from the job search lol.

Could you speak more on the Shazam style part? Don't you need a database of all ads to figure out if some snippet is an ad or not? That's how Shazam works for songs at least.

A native implementation I had thought of was running the audio through a STT LLM that then detects the ad timestamps then returns that to the UI to block.

Yeah that's what I do (STT -> LLM -> timestamps to block), but if you did that for every episode download, due to location-targeted dynamic ads, your LLM API bill would run up very fast. So that's where the fingerprints matching comes in, for further versions of an episode (version = same episode but with different dynamic ads inside).

I explained how the full system works here (someone emailed me and asked): https://pastebin.com/raw/r2YUEkK5

Thanks!

I use fluidaudiocli and it's unfortunate that it doesn't support streaming (e.g. from a named pipe); that would have been an easy workaround to both the pipelining problem and the faster-decoder problem.

Yep, agreed, streaming would be great to have. Although via pipes I think is better suited for a macOS. Since FluidAudio has to support both macOS and iOS, I think using Swift concurrency primitives would be the better fit. That's what I did for my app (TaskGroup, actor, AsyncSequence).

I hope they replace their awful voice to text on their keyboard. I can't stand that terrible bit of software.

I’m not sure I agree. There may be better models, but the comparison is still useful so long as whisper is so widely used.

> RIP to a lot of the paid apps that simply wrap Whisper

I started using a few open source apps for transcription and eventually subscribed to a paid one...

On paper, it's not hard to compete, but for this use case, a few rough edges make it really frustrating to use. Like a keyboard that sometimes doubles the letter "e"

Automatic dictionary, seamless language switch, no issues with accents, etc... Putting the effort in the last mile makes a world of difference.

If anyone has better options, I'm willing to have a look. The best open source solution I found was Handy, and I currently use Wispr Flow

I built my own because I was frustrated with a lot of the free options. Largely because a lot of them had an upsell to be able to do the secondary post-processing step with an LLM. And it wouldn't pick up things like emojis properly or say numbers. Because of that, I left quite a lot of options in there for customising and adding additional steps, etc. Feel free to take a look: dictator.robgough.net

My initial Mac version actually had three additional steps that you could toggle, obviously at the cost of some speed. That is what the website talks about, although nowadays for my own use I've reduced that to just one step and found that it's pretty great. I've got a new version in test to tidy that up, but still lets you define as many steps as you want.

App works great. Parkeeter TDT V3 not so much. I speak Polish and it detects Russian. Also there is no possibility to force language:

> unfortuntately parakeet-v3 model doesn;t recieve or output language id https://github.com/NVIDIA-NeMo/Speech/issues/14799#issuecomm...

Also this:

> If you are using Parakeet for English only then you should be using V2. V3 is for several languages and is worse at English only. https://news.ycombinator.com/item?id=48897908

Hey. Yes. I did vive code one as an exercise yo learn how to publish to apple store.

Listen and transcribe felt like the easiest thing to do.

Distavo.com

The source is open for anyone to use, and the builds are in github.

I found quite interesting that claude didn't help too much on how to publish to SetApp until Fable.

How many of the Whisper competitors will work at a reasonable speed using only CPU (on Linux, not Apple)?

(Genuine question - I'm a happy Whisper user but am always looking for improvements).

Yeah, apple will be optimizing a model to work on ANE and then turn it into a native app. My only hope is that it has a reasonable api so that I can use that as a generic input source across iOS / macOS that’s equivalent to the ubiquity of the keyboard.

Presumably the existing transcribe button on the keyboard will route through this on iOS 26?

One would hope, though I suspect they may want to make things a bit flashier than “we made the audio transcription on the keyboard not terrible” in the changelog given the amount of work that’s gone in.

I use handy.computer and it is pretty much everything I want from a transcribing app.

I don’t know how Apple divides computation between the GPU and the Neural Engine, but one major benefit, especially for real-time transcription on laptops, is the improved power and thermal efficiency. I noticed better accuracy after switching my app to SpeechAnalyzer, and I suspect part of that improvement for me came from the microphone no longer having to compete with jet-engine fan noise.

What Apple laptop has “jet-engine fan noise”?

I’ve regularly used an M1 Max, M4 Pro, and M5 Max. They all get pretty loud when driving local LLMs. “Jet-engine” would be hyperbolic, but it’s definitely noisy.

This particular product used Whisper, so that was obviously the right model to compare it against. Further this is explicitly on device, and Nemotron 3.5, as one example, is 2.5GB for the model.

And if someone were broadly comparing all on-device models (instead of just looking at how this new on-device ones compares to what a specific product uses), Nemotron 3.5's WER are actually a bit higher than what they report for SpeechAnalyzer, for both tests.

I am curious, what are the use cases people are using voice transcription for?

I find voice memos really annoying, so I’ve been dropping them into a Whisper frontend for a while. Something built into the OS would be nicer though.

Pretty sure iOS is providing this? At least for messages and voicemail.

Yup but on third party apps, where you can download an mp3, there is not always a built-in transcription option

Turning dialogue audio into text for searchability and accessibility (particularly for deaf users)

The canadian government will provide lots of historical data for curious citizens, many of which are recordings of interviews from decades and decades ago. For a book project this allows me to make a hours of audio searchable through a GUI application I have developed that has a voxtral backend.

Whisper v3 is still the best (by far) when it comes to poor quality input (say background audio from a security camera), though remains more susceptible to hallucination so it's a bit of a tradeoff.

Any models that can understand thick accents better than I can?

Anytime I’m talking to an Indian on the other end, I have to have them repeat everything 2 or 3 times.

Yes, poor comparison to what's now a relatively low end model.

I'm curious how well it'll work in the real world, but I would be ecstatic if I could ditch all my Whisper apps.

- transcribed using MacWhisper.

> RIP to a lot of the paid apps that simply wrap Whisper, I'm sure Apple will make a native GUI such as a recorder app for macOS that obviates the need for these wrappers

I'd love this, but updated spotlight did not obviate my need for Raycast. I question Apple's ability to make good software at this point.

Being Apple’s model, it will support like 8 languages and leave the rest hanging for 10 years, just like Apple impotently ignored 10 or 20 million countries even with basic “just download open dictionary and run a deploy script” keyboard autocomplete.

Just ran it against Whisper-Large-V2 on a math lecture (my primary use case for ASR is subtitling math lectures), and it was substantially faster and only slightly worse. Very usable for live transcription though I'll probably stick with whisper for the time being since I don't really need the subtitles to be generated in real time.

If it was faster but worse, maybe compare it to a smaller whisper model?

I imagine because quality of transcription is what matters.

Been using it for a podcast app I have been developing for half a year lol (I hope I publish it by version 27) and I can confirm it’s real fast.

Splitting the audio in multiple segments and firing it up without hitting the maximum limit of concurrent decoding streams makes it blazing fast. Fair enough you loose the cut, but it’s good enough for just podcast. In one minute it chews through one hour of audio. This on an iPhone 17 Pro.

What's different about your podcast app?

Nothing really, except that I get to play with SpeechAnalyzer APIs, foundation models, translations. It’s basically my playground where to try all things. Been listening a lot of Chinese podcasts lately, transcribed and translated by local models.

Edit: all that said, the app is irrelevant. What I want to say is that live transcripts on iOS using Apples frameworks works very well. Only thing I miss is diarization support.

You could perhaps run over the segment splitting points (plus a few seconds back and forward) in a second batch then merge the results in the end so you don't miss anything.

You probably know that already, but there's a great ATP episode about using Apple's models for podcast transcribing at scale for Overcast.app (https://appleinsider.com/articles/26/04/07/giant-mac-mini-cl...)

I will plug Willow for mac recording. IMO it's basically to me a "better than perfect transcription" as it cleans things up and is almost instant. I liked Superwhisper but switched to Willow as it was a big difference.

Its so good that I'm not sure that it's possible to get any better. Speech to text seems like basically a solved problem, if not now then definitely in 5 years. I don't know if any of these speech to text businesses will work in the long run, but for consumers they are great. My guess is the 2030 version of Apple's SpeechAnalyzer will be so good that nobody will need to use 3rd party software.

What isn't solved is domain-specific jargon with these tools. When i talk to my coding agent, i want to be able to speak the names of symbols and files and have it be aware of that stuff, like having LSP integration.

If I say 'useSuspenseQuery' I want it to come out as useSuspenseQuery not 'use suspense query'. Even if I had to say 'symbol useSuspenseQuery' to give a hint that i'm referencing a symbol, that would be fine.

Tried running it through something that knows what you're working on?

It doesn't stream text does it? If it could it would be perfect in my book.

Vs Voxtral would be a better comparison. No other model, open or closed, has been able to hit such a low AER (Acronym Error Rate ;)) for my meeting transcripts. Seems to understand/infer all the technobabble I use at work. Never have to edit anything. Whisper was catastrophically bad.

I typically disable autocorrect on Apple products because of this, cautiously optimistic about their improved speech models, but definitely worried that it's going to 'correct' technical jargon to more common words.

One tangible thing this doesn't touch on: SpeechAnalyzer supports streaming, so you can see what it's hearing from you as you talk. A massive UX improvement. Many of the other models force you to record, then it transcribes the audio to text as a single job, then returns the entire blob of text. It's slow, and frustrating if you're talking, only to realize it stopped listening after 2 minutes.

Whisper small/tiny/base are almost four years old (they were not updated for Whisper v2 or v3). Is there really nothing better to benchmark against by now?

There's tons, Parakeet was the last I remember seeing which seemed to gain traction (independent lightweight implementations etc).

Parakeet isn't as good as whisper large.

Parakeet is way faster (on Nvidia hardware) but not quite as accurate in my experience.

It's also super fast on CPU.

Nvidia's Nemotron subsumes their older Parakeet model now even for real time streaming.

I have tried everything (that will run on a 12GB RTX 4070) and I have yet to find anything with better accuracy than Whisper V2 Large for my dataset (discord audio from TTRPG sessions, isolated per-speaker, mostly non-American accents)

Same, for my English-only podcast

Not v3?

Voxtral to me what better

There are many [0], you can search and filter by streaming and open weight only as well.

Looks like Voxtral and Nvidia's Nemotron are best.

[0] https://artificialanalysis.ai/speech-to-text/non-streaming

Just this week I built a live subtitles app for my mother in law who is hard of hearing (she has a hearing aid but still has trouble decoding, but can still read faster than me)

So I ended up organically testing and ending up with SpeechAnalyzer because it was not only fast and accurate enough, but you also see live results as you talk. It also has speaker identification and people can register their voices. And it does all processing on device.

It also had the best model for Indian accented English, given she lives in India.

So I was quite impressed, but the holy grail to me is transcription that does speaker identification but also works in a standard family conversation, where multiple people interrupt each other all the time.

I will say though, I'm really curious as to what Claude Code Desktop uses for their voice mode, because it seems even better than Apple's, and it provides realtime feedback. Maybe they're using apple's model?

Impressive. Apple said they improved the models in 27 didn’t they? It would be interesting to see the numbers the beta turns in.

Finally. I‘d be delighted though if they actually implemented language autodetection (like everywhere else) though. There’s little more frustrating in my day to day than having dictated half a page to find that it‘s complete gibberish because Apple forces you to select the right language first…

Same with the keyboard. Apple is completely incapable of taking context into account for the input mechanisms of the operating system.

If I start typing and the existing text is in Spanish, then a sensible default is to select the Spanish keyboard I have installed and let me adjust otherwise.

App developers should also be allowed to supply mini-dictionaries within a context to allow autocorrect to work correctly in that context, so for example in this thread [SpeechAnalyzer, API, Whisper, Parakeet, Nemotron] should be supplied so that these terms are autocorrected.

I took a swing at bringing this into Handy.computer if anybody's interested: https://github.com/cjpais/Handy/discussions/1031 . Looks like there has been past demand for someone to implement it, but no proposed PRs. This article was inspiring.

Is this the new dictation engine that I'm not allowed to run on my 1-YEAR-OLD IPHONE 17 because it's not Pro?

Why did you buy the cheap one, that's your own fault

At the time, there was very little advantage to buying the Pro model. Ironically, it's the first time I've ever bought the base model.

Serves you right

I hope this is sarcasm.

I make an iOS app that uses this API heavily for transcribing diverse audio of varying bitrate and recording quality. The audio often contains music, multiple speakers, sound effects. SpeechAnalyzer almost always gets it.

It can struggle with proper nouns but will return something phonetically similar.

My main gripe is that it requires a separate model download per language. I understand the why they did this (to save disk space). But it makes multi-lingual audio hard to transcribe unless you know ahead of time the languages in the audio.

As an app developer the biggest win from using Apple's model is I don't have to bundle it in my app so my app looks much smaller. If a user has many transcription apps each one could have their own model. If Apple's model is used only one copy is needed.

I use Spokenly, offline-only mode with the Nvidia model. All local, totally free. Highly recommend

For just desktop, I use Handy (open source): https://handy.computer/

And yea, Nvidia's Parakeet v3 is good enough for my own just local transcription most of the time.

When I need local transcription to be more reliable and I don't have the energy to proof read a long ramble, I still often just pop open chatGPT, dictate, cut, paste.

But we're pretty much already to the point where local transcription models can replace cloud ones for personal use. They're still a bit rough around the edges in terms of polish and latency, but plenty of people are fine with that to avoid yet another app subscription and not having to worry about wondering what's potentially happening with their data.

Thanks will check it out, hadn't heard of it

Spokenly is not 100% free.

If you use the offline-only models (I use the Nvidia one and it's great) and put it in offline mode it is. I've used it for 6 months and haven't paid anything

Any chance you can benchmark against whisper large and large v3 turbo? These run comfortably on older Macbooks and are still far more accurate in real life dictation compared to even the parakeet models( despite ASR leaderboards) with an RTF < 1.

Try MOSS-Transcribe-Diarize from a few days ago. I’m getting better results than those whisper models. And it’s very fast and small. Better suited to noisy audio too.

>The result, up front

It's very welcome, to start with this.

However, just from this heading, I knew with 90% certainty that the post was AI generated.

I closed the website after seeing that heading.

I wonder if I would have less of an issue with this if such blog posts would just start with: “I asked $model $model-version the following: $prompt. Here is what I got.”

I have implemented this into Home Assustant some months (https://github.com/FI-153/wyoming-apple-stt). I find the comparison with Whisper particularly helpful since it is the base model used in Home Assistant to transcribe voice in the Assist pipeline. I find Whisper too heavy to run locally with good quality, SpeechAnalyzer instead comes bundled in macOS and is always available.

I'm always confused by these phrases:

> The new API cuts word error rate by 3.5 to 4x on the same audio: from 9.02% to 2.12% on clean speech

Shouldn't they have said "cuts error rate by 78%" or something?

I don’t like it written that way either, and it always seems like the type of number you put on a slide for a head of sales or something. It rankles because:

- it implies that error could be increased n-times, but a 15x _increase_ in 9% error would be an error rate of 135%, which is nonsensical.

- a reduction from 90% error to 20% error is clearly a bigger improvement in rightness to a reduction from 9% to 2%. One is “almost all wrong to almost all right”, the other is “more right”, but they are both a 4.5x reduction in error which means that the 4.5 quantity doesn’t have a constant meaning.

The answer is something like log odds ratios, but that introduces the additional need for a reader to know what that is, and that would be unusual.

I second that! Can you run your benchmarks against the iOS 27 beta?

OK, but how does it compare against Parakeet TDT2 (english), TDT3 (many languages), and Parakeet TDT3 Streaming? And what about whisper large?

Yeah i do find apple's speech to text very good lately and no need to use openai or anything that seem to market their services better

I didn't expect Apple's SpeechAnalyzer to see such an impressive improvement. However, Whisper's biggest moat remains its cross-platform availability and support for 100+ languages. It would be awesome if SpeechAnalyzer could match that level of platform and language coverage.

Im hoping Apple gets the new Siri working better on older phones. I was excited to use it but the latest beta / Siri runs too slow on my iPhone Pro Max 15.

Im looking for the same experience I have when talking to chatGPT. As for past two years or more talking to GPT within it's app and on my iPhone Pro Max 15 it runs smooth as butter :-). This is the experience I was and still am hoping with Apple, but Im thinking all the extra layers of privacy and security might be slowing them down?

Overall, Apple who is suing Open AI should just buy them and let me have the best conversational AI out there baked into my old ass iPhone. Because as so far the new Siri on my old phone (tho again GPT works great talking to it and for years) doesnt come close. It's the same old "Could you try that again," Siri. BOO!!!

Yeah, ChatGPT voice is great experience vs. Siri on that phone. In case you haven't done something like this already:

  1. In Shortcuts app, make shortcut named "AI Voice Mode" (or whatever you want, YMMV)
  2. Set it to run the ChatGPT action "Voice Mode" (requires at least the minimum paid tier, I think)
  3. To trigger, say "Hey Siri, AI Voice Mode" (or whatever you called the shortcut)

This is a pretty slick integration, but yeah, if it were baked in that would be all the better.

You can also map the Action Button to GPT voice mode.

Ridiculous that for past many years we can talk to GPT on our iPhones without any hiccups and this new Siri is still the same old horse crap (at least for me and this latest beta). Buy them already Apple or possibly be replaced by them as their path & trajectory (working on Ai devices now and they are stealing like Jobs did with Xerox) mirrors yours in the late 1970s.

Thanks for the tip and if Im not mistaken it's similar to asking Siri to ask chatGPT to ask XYZ?

> similar to asking Siri to ask chatGPT to ask XYZ

Effectively, it sort of does that, but really it just listens to the wakeword and opens/switches to the requested app & modality.

FWIW, I get a very different functional result using the Shortcut method vs. asking Siri to delegate natively. To compare, I asked Siri (non-beta here) now to "ask ChatGPT <x>" and I got a top-card with some fairly low quality SEO-ranked weblinks.

I’m on iPhone 17 Pro Max, 27 beta 3.

New Siri is impressive in that it answers satisfactorily now 80% of the time vs 10% with old Siri.

But it’s slow as shit. GPT, Claude, and Gemini can answer me in 5-10 seconds. Google AI Mode can answer in 2 seconds.

New Siri usually takes 25 seconds to respond to me. This morning it timed out (with strong network connection) when asked a simple multiplication question.

Damn slow on your newer phone too.

That's crappy, but it's also a beta, so it's not yet time to render a verdict.

Last time I asked my Google speaker for a simple multiplication (#00 x #, embarrassingly), it started with "on the website facebook.com". I'm not sure if this is a good comparison.

> Im hoping Apple gets the new Siri working better on older phones.

Apple would never do that, if anything they did not offer their Siri with the most advanced AI on iPhone 16 Pro Max, which is one year-old only.

For my current purposes, I need a speech-to-text model/API to also emit word-level timestamps - for now, that makes ElevenLabs's Scribe v2 the best multiplatform, multi-language choice though it does look like this SpeechAnalyzer API provides them (although only for English).

If you want to use this on the CLI: https://github.com/finnvoor/yap.

Supports SRT/TXT/VTT or JSON-with-optional-word-level-timestamps output and progress meter.

Also it can transcribe live system audio.

> Apple's new SpeechAnalyzer is the most accurate on-device speech engine we tested. It beat every Whisper model we ship, including Whisper Small, on both the clean and the noisy half of LibriSpeech, while running roughly three times faster than Small.

The article just says that SpeechAnalyzer size is "system." It actually treats each locale like an on-demand resource and Apple doesn't document how much space each one takes up or under what circumstances they get cleared.

In my own tests a few months ago, it was faster than both small/large Whisper models I compared it to with accuracy competitive with both of them (each model had different quirks).

Why is my Siri still so horrible at dictating text then?

Because the new Siri is coming with iOS 27 and according to reports works pretty well now.

I'm running the developer beta and it's still horrible.

Did you sign up for the waitlist for the new one or are you using the stock one?

Yes, I signed up for the waitlist for the new one and got in.

Still sucks

I stopped reading after seeing they compared only with Whisper Small, Base, Tiny

This is useless test and benchmark when you have these day Whisper-V3-Large and Whisper V3-Turbo that you can faster than realtime on 5 years old macbook on apple sillicon (ANE). They didn't even compared to parakeet v2 or parakeet v3. And only english language...

Was intrigued by the Inscribe product so I downloaded it to test it out. I uploaded 1 voice file about 4 minutes long and was promptly given a message that I've reached the free plan limits.

Kind of a bait and switch. How can we test the product with such short time limits and what, exactly are you offering if all the processing is done on device by Apple?

Lots of comments are "you should compare against X and Y" - even better, just get the results on a standard benchmark, so you can compare against all,

https://huggingface.co/spaces/hf-audio/open_asr_leaderboard

well, it's nice, but the multi-lingual diff is rather limited (only European languages).

How would it compare to Wispr Flow? I recently started using it and it feels so much more robust than anything out there (and in fact the one in iOS and MacOS 26 seemed pretty poor in comparison)... and do we know if Wispr Flow uses the Whisper tech behind the scenes or if they have their own model etc?

Whisper large v3 turbo works fine locally on recent iphones, weird to keep it out of the comparison

Does Apple use this model with the default keyboard? Or if the keyboard still uses the old API?

i wish they had benchmarked it against parakeet-unified-en-0.6b and cohere-transcribe-03-2026. i am using parakeet with https://handy.computer daily and it's amazing.

Every single asr model I tested so far did not support timestamps properly though. Some use external aligner to create timestamp, but the accuracy is still much inferior than whipser in case the audio is noisy.

Anyone know the best choice these days specifically for speaker diarization?

I'd be interested too. Last time I looked the state of the art was pretty bad.

I run SuperWhisper on both my Mac (where is uses Whisper) and my iPhone (where it uses SpeechAnalyzer and have found that SA does indeed run faster and anecdotally more accurately. Super exciting!

I've been using Parakeet Multilanguage locally, and it's super snappy

Lifetime subscriber to a Superwhisper. How do you have SpeechAnalyzer to run as a model on iOS??

Cloud is so cheap and quick. I use local too but my api bill is like 3 quid a month. You would have to be very cheap or have compliance needs to tolerate the error gap

Can anyone tell me if this model is used by Mac OS when I press the dictate key on my keyboard? Or is that an older model/setup?

Excited to get next years diarization models - ArgMax is pretty incredible but cost prohibitive (when compared to 0 price from iOS)

I'm looking for good and cheap transcription + speaker diarization on my Mac for a small personal project. Recs?

Obligatory mention of Handy which is open-source and cross-platform:

https://github.com/cjpais/Handy

It's great for me when configured to use Parakeet v3.

If you are running ios 27 beta is this the model when you hit voice to text on the native keyboard?

Would this end up replacing the default iOS keyboard dictation functionality in iOS 27?

No, this is the current dictation functionality on iOS 26.

Do you happen to have any links that document this? I haven't been able to confirm.

It’s a great API. I use it in my app and it’s feels so much more accurate.

"The results, up front" AI post. Yes, I can tell.

Anybody compared this against Whisper Turbo?

This is great marketing, I had no idea what inscribe was, but a blog like this going viral did something no ad could do for me.

this is amazing. if i had a mac i would try to reverse engineer the code, extract the weights and port it to something that works on linux/windows like torch or burn. then put the code on github and weights on a torrent site. lifes too short to let apple keep their models exclusive.

Is that copyright infringement?

This hasn't been tested in court. But there's a high chance that model weights are not copyrightable, only the code to generate them is.

Cloud models are usually protected by trade secret laws, leaking them would get you in trouble. However if the model is made available publicly, as long as you don't break the law to get them, anything after that would be fair game unless Apple can prove that humans have significant authorship over the weights, which hasn't been tested and is a significant burden to prove/disprove.

So the models will be protected by contact law instead. Can you get an apple update without agreeing to the iOS terms?

Copyright protects original forms of expression, not arbitrary data. It is very arguable whether it applies to model weights. However, it would likely constitute a license violation.

Aside from the legality of it, I think you are underestimating how complex it can be to do that. It is possible in theory but not something that will be a fun side quest like you are making it seem.

With a IDA Pro decompiler license & MCP server, paired with Codex/Claude Code... it would be a fun side quest.

You likely don't need to disassemble the inference code, the weights are "just an array of numbers" in MLX format.

I believe they are are protected on disk

This.

The Jedi Hand Wave-y nature of the way people talk about AI these days is going to make reigning in the AI superpowers nearly impossible. Because there are people out here who believe models of this quality are easily replicated or reverse engineered. Neither is really doable on any reasonable timeline by people who are not AI experts. Real AI experts. Not TF/PyTorch monkeys or Agent Slop Slingers.

And those people are already highly incentivized to not make anything performing better than SOTA models open source.

Probably not worth the effort (or legal trouble), unless you can show it's better than other recent open models like Cohere Transcribe.

I wish Apple would update there built in TTS model (on Mac). In this day and age, it's quality is terrible.

Still nothing beats OpenAI's VTT. Anthropic's sucks and Apple's isn't even usable.

Edit: Getting downvoted by Apple fanboys for telling the truth is a badge of honor.

Which OpenAI model/API do you mean?

Whisper and GPT-4o (for diarization).

In which world 98% accuracy is not usable?

In a world where you say "tmux" and Apple's VTT writes "T Max".

You can't reasonably expect generic ASR to infer tmux from "tee-mucks". "tee-em-you-ex" works reliably if you're ok with capitalisation for your use case.

agreed. plus all the languages supported!

Author here. I ship both Apple speech engines plus WhisperKit side by side in a transcription app, which made it possible to run all five through identical production code on the same audio: LibriSpeech test-clean and test-other, 5,559 utterances, fully on-device on an M2 Pro.

Apple published no accuracy numbers for SpeechAnalyzer (or for SFSpeechRecognizer, ever, as far as I can tell), so the migration question has been guesswork. Short version: the new API cuts WER 3.5-4x vs the old one (2.12% vs 9.02% on test-clean), and it also beat Whisper Small on both splits at about 3x the speed. The old API came in last on clean speech, behind even Whisper Tiny.

On "why should I trust a vendor benchmark": the Whisper column reproduces OpenAI's published LibriSpeech WERs within +0.11 to +0.42 on all six measurements (same corpus, same normalizer, same scorer for every engine), and the raw per-utterance transcripts are downloadable from the article if anyone wants to rescore with their own normalizer.

Limitations worth stating up front: English only, read speech rather than meeting audio, one machine. Precise per-engine timing isn't in the article yet because the accuracy runs shared the machine with a dev workload; WER is load-independent, timing isn't.

Two things that might interest people migrating: SFSpeechRecognizer sends audio to Apple's servers unless you set requiresOnDeviceRecognition, and with SpeechAnalyzer, finishing your input stream is not enough to end a session. If you never call finalizeAndFinishThroughEndOfInput(), the results sequence never terminates and your await hangs forever. I found that one because it was shipping in my own app.

Happy to answer questions about the harness or the normalizer.

At this point, I would not recommend ignoring Parakeet TDT 0.6b v2/v3 (english-only versus multilingual). Those models have been out for a year, give or take, and they are both accurate and fast. I would choose Parakeet over Whisper in almost all situations these days. Parakeet works great even on my several year old iPhone 15 Pro Max, so if an app is going to ship a dedicated model, I strongly recommend investigating Parakeet.

On the more cutting edge front, Granite Speech 4.1 has proven to be a reliable workhorse for me, but it is larger than Parakeet. Cohere Transcribe is interesting, but how strong it is seems to vary more from task to task.

Parakeet Unified 0.6B came out a few months ago, combining both online streaming and offline transcription into one model, and that is one that I need to test more, but it seems promising.

As others have mentioned macOS 27/iOS 27 is supposed to have a new model, particularly on devices with 12GB of RAM or more. I have not actually seen the option to enable that new model yet, though, despite being on the beta on a device that meets the requirements. Maybe a benchmark would reveal that it is already active?

Why use relatively ancient models like whisper and not e. g. parakeet?

Why do you not use Whisper large models when on macOS? They're still fast even when streaming and yield a much lower WER.

Also, just out of curiosity, seems like everyone and their mother is making Whisper wrappers, how is your app different?

Please run your benchmark on this new and very impressive model https://huggingface.co/OpenMOSS-Team/MOSS-Transcribe-Diarize in my testing it outperforms all mentioned especially on noisy audio

MOSS-Transcribe-Diarize

If this isn't open source/weights and can't run locally, I don't see how this is a replacement for Whisper or other open models, e.g. within Home Assistant.

It's a local model so it's essentially open weight such that you could feasibly export it somehow since it's already on the laptop somewhere. Apfel is a wrapper app like ChatGPT but using Apple Foundation Models, I assume something similar will happen with this transcription model.

https://apfel.franzai.com/

It's not open weight, but the point is to be an on device (and thus local, privacy preserving) option. The article mentions that as the caveat

> What this means if you just want good transcription

> If you are on a current iPhone or Mac, the best on-device transcription engine for English is already in the operating system, and the private option is no longer the compromise option

> It's not open weight, but the point is to be an on device (and thus local, privacy preserving) option.

How can you be sure this isn't leaking data or metadata to Apple? Can Apple really be trusted?

Test it! Does it make network requests? Unplug the internet and see if it still works!

You are being naive. An Apple device makes dozens of network requests every minute or so to Apple. It is neigh on impossible to verify what is being requested or sent. Also unplugging the Internet and verifying that something still works does not mean the app won't phone home behind your back when it can. These things are designed to fail silently.

Download use Little Snitch or

https://alternativeto.net/software/little-snitch/

https://www.g2.com/products/little-snitch/competitors/altern...

There are many alternatives for trying to find out what’s going on. If you don’t want to bother, and most people don’t, well, what else is there to say?

It is generally a good idea to know what software is phoning home, if you can pinpoint it.

If you have any software recommendations, I’d be happy to know.

It's more complex than that. LittleSnitch or other application firewalls won't help when programs like package managers legitimately need to access the Internet. And figuring what what is being sent would require decrypting the traffic. There is no easy way to figure out if Apple or some rogue app is siphoning off your home directory.

This is Hacker News. Computers are knowable, tractable things. It is trivial to know if a process is making network requests.

It's definitely not trivial to figure out what for, and what is being sent or received.

The article states:

> If you are on a current iPhone or Mac

Presumably if you don't trust apple you wouldn't purchase their products and even if you were for example forced to use it via work or something you wouldn't use this feature ... so it doesn't really change the calculus as presented by this article - IF you ALREADY HAVE a MODERN Mac (and trust apple) this is your best option

If already have current Mac or iPhone, and thereby implictly "trust Apple", then why would one need a "a local, privacy-preserving" option

If trust Apple, then no need for privacy from Apple

It is local

The appeal is that users only have to download it once across all apps that use it. Instead of convincing a user to give a couple gigs for just your one app

Reply on news.ycombinator.com