Ask HN: How to transcribe a couple thousand calls per day?
The user finds Microsoft Speech Service complicated and Azure OpenAI whisper has a low quota. They seek a method to batch transcribe short calls using a powerful server with RTX 4090 cards.
The user has evaluated various speech transcription services, finding Microsoft Speech Service overly complicated and noting that the Azure OpenAI whisper deployment has a low quota. They have had some success using ggeranov's whisper on a Mac, but this solution is not integrated into their corporate network. The user is seeking a method to batch transcribe calls, as they are currently behind schedule. They have access to a server equipped with two RTX 4090 graphics cards, which is ready for use with the necessary Nvidia drivers. The calls that need transcription are relatively short, averaging 90 seconds each.
- Microsoft Speech Service is considered too complicated for user needs.
- Azure OpenAI whisper has a low quota, limiting its usability.
- ggeranov's whisper works well on a Mac but is not part of the corporate network.
- The user needs to batch transcribe calls and is currently behind schedule.
- The available server has powerful hardware (2x RTX 4090) ready for transcription tasks.
Related
OpenAI is set to lose $5B this year
OpenAI's projected costs for 2024 are $7 billion, with a potential $5 billion loss. Revenue from ChatGPT is about $2 billion annually, indicating a significant financial shortfall.
OpenAI rolls out voice mode after delaying it for safety reasons
OpenAI is launching a new voice mode for ChatGPT, capable of detecting tones and processing audio directly. It will be available to paying customers by fall, starting with limited users.
OTranscribe: A free and open tool for transcribing audio interviews
oTranscribe is a free web application for transcribing audio recordings, allowing seamless navigation, automatic saving, and local file storage. It supports multiple export formats and is open source under the MIT license.
Show HN: LLM Aided Transcription Improvement
The LLM-Aided Transcription Improvement Project on GitHub enhances audio transcription quality using a multi-stage pipeline, supporting local and cloud-based models, requiring Python 3.12 for installation and execution.
Benchmarks show even an old Nvidia RTX 3090 is enough to serve LLMs to thousands
An analysis by Backprop shows the Nvidia RTX 3090 can effectively serve large language models to thousands of users, achieving 12.88 tokens per second for 100 concurrent requests.
6 commentsIf it doesn't, there are a bunch of other speech recognition APIs. Most of them use older techs but might be good enough: https://www.gladia.io/blog/openai-whisper-vs-google-speech-t...
Personally I found Otter.ai works really well for the transcription part, but they don't have an API: https://otter.ai
You can also just upload them all to YouTube in a private playlist and it'll automatically transcribe them for you.
Related
OpenAI is set to lose $5B this year
OpenAI's projected costs for 2024 are $7 billion, with a potential $5 billion loss. Revenue from ChatGPT is about $2 billion annually, indicating a significant financial shortfall.
OpenAI rolls out voice mode after delaying it for safety reasons
OpenAI is launching a new voice mode for ChatGPT, capable of detecting tones and processing audio directly. It will be available to paying customers by fall, starting with limited users.
OTranscribe: A free and open tool for transcribing audio interviews
oTranscribe is a free web application for transcribing audio recordings, allowing seamless navigation, automatic saving, and local file storage. It supports multiple export formats and is open source under the MIT license.
Show HN: LLM Aided Transcription Improvement
The LLM-Aided Transcription Improvement Project on GitHub enhances audio transcription quality using a multi-stage pipeline, supporting local and cloud-based models, requiring Python 3.12 for installation and execution.
Benchmarks show even an old Nvidia RTX 3090 is enough to serve LLMs to thousands
An analysis by Backprop shows the Nvidia RTX 3090 can effectively serve large language models to thousands of users, achieving 12.88 tokens per second for 100 concurrent requests.