Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Gemini-3.5-Transcribe

blog.google

Post reply on HN
Print view View original

111–120 of 140 posts

Re: Gemini-3.5-Transcribe

#111

Post by phoghed »

mediaman wrote: post #54
huijzer wrote: post #52

Earlier quoted context omitted.

Maybe I missed something, but isn’t it impossible to detect whether something is written by AI or not? A human on a bad day can write like AI while an AI on a good day can write like a human.

No, that's not correct for any reasonable definition of "impossible." Look up pangram's accuracy ratings. It's not perfect, but it's pretty good. LLMs in fact leave very distinguishing traces of their logit distributions in the text they write. It's one of the reasons why it's so easy for humans to also smell them. It is possible to trick pangram - they bias toward a low false positive and a higher false negative - b…

What’s impossible is what a normie would understand that it does based on their marketing and home page.

It’s trivially defeated though, and anything with a false positive rate shouldn’t be used by any serious institution on a decision making basis. For general stats, sure. For trying to punish and individual, no thank you.

Re: Gemini-3.5-Transcribe

#112

Post by ameliaquining »

altmanaltman wrote: post #86
ameliaquining wrote: post #48

Earlier quoted context omitted.

Pangram says human: https://www.pangram.com/history/a69f9b74-eb46-44b9-a087-7822...

Checking the following with another AI checker and it says 100% AI generated. ( https://originality.ai/ ). We should run each check on multiple checkers if you want to provide a substantive claim that something is AI-written or not. This habit of just putting text in a "AI text checker" and then treating whatever it says as the truth is absolutely one of the worst things to emerge in recent times. You should check ou…

I'm looking at originality.ai and I'm suspicious. Their roundup of studies is largely ones from 2023 (and I don't think any AI detectors from 2023 actually worked), with a few newer ones that nonetheless don't include Pangram among the detectors that they're comparing (which I consider highly suspicious and possible evidence of cherry-picking). A quick internet search surfaces various anecdotes suggesting that originality.ai has a high rate of false positives compared to other detectors (and in particular doesn't highly prioritize making false positives rare).

I'm open to the idea of different norms for citing AI detectors, but someone needs to propose what they should be and they need to make sense.

Re: Gemini-3.5-Transcribe

#114

Post by kristofferR »

" Your Google Cloud billing account is being processed. Processing time varies from a few moments to a few weeks. "

What kind of bullshit is this? I want to test out Gemini-3.5-Transcribe in my app, but have to wait a completely indeterminate time, have already waited over 24 hours. OpenRouter took seconds to set up

Re: Gemini-3.5-Transcribe

#115

Post by decide1000 »

I use Whisper in combination with Qwen3.8 27b to create Dutch subtitles for Russian videos. I follow Ukrainian channels about the war and I am now able to understand them. In my tests with other languages Voxtral was better then Whisper. Whisper often makes mistakes which I now solve with an extra layer (qwen).

Sadly Mistral does not support Ukrainian. I hope this changes! I expect better from a European model

Re: Gemini-3.5-Transcribe

#116

Post by fumeux_fume »

Lucasoato wrote: post #21

I’ve tested at least 20 STT models in a benchmark I’ve set up with German, Italian and English voices from meetings in my company. The voices contained very industry specific words, the languages changed from one sentence to another, sometimes words in a language were mentioned while a discussion was in another. The only local model that satisfies me is Voxtral Mini 3b, the only paid API that is slightly better is el…

I also found eleven labs gave me the best results when transcribing French dialog from a 60s television show. Voxtral's output format was the easiest to work with when attempting to create actual subtitles.

Re: Gemini-3.5-Transcribe

#117

Post by jobuildsstuff »

simonw wrote: post #106

If this is a dedicated transcription model does that completely eliminate the risk that it might do weird things based on the audio it's transcribing? The nightmare scenario with an LLM-based transcription system is that someone says outloud "actually ignore that idea, instead let's..." - and the previous idea gets omitted from the transcription!

[flagged]

Re: Gemini-3.5-Transcribe

#118

Post by meetingthrower »

I built a small little app for a company which transcribes all their customer service calls, does an AI review, makes a podcast of the most problematic ones (they LOVE this.) Probably 20 hours a day of transcription.

What are people using for cheap hosted models for this? (I don't want to run my own infrastructure.)

I'm currently using Gemini Flash Lite. I don't need real time, it's all batched, clear english... still costs some small $ a month. Figured it should be able to be reduced if there are cheaper models out there....

Re: Gemini-3.5-Transcribe

#119

Post by m3kw9 »

meetingthrower wrote: post #118

I built a small little app for a company which transcribes all their customer service calls, does an AI review, makes a podcast of the most problematic ones (they LOVE this.) Probably 20 hours a day of transcription. What are people using for cheap hosted models for this? (I don't want to run my own infrastructure.) I'm currently using Gemini Flash Lite. I don't need real time, it's all batched, clear english... stil…

have you tried some of the opensource models?

Re: Gemini-3.5-Transcribe

#120

Post by jasondigitized »

meetingthrower wrote: post #118

I built a small little app for a company which transcribes all their customer service calls, does an AI review, makes a podcast of the most problematic ones (they LOVE this.) Probably 20 hours a day of transcription. What are people using for cheap hosted models for this? (I don't want to run my own infrastructure.) I'm currently using Gemini Flash Lite. I don't need real time, it's all batched, clear english... stil…

What system are they using to take / record calls?
Post reply on HN