Transcribe video to text | OpenAI Whisper | Lightning AI | Insanely Fast Whisper
#openapi #whisper #huggingface #lightningai #clip #transcribe
https://gist.github.com/techforum-rep...
What is OpenAI Whisper?
Whisper is a machine learning model for speech recognition and transcription, created by OpenAI and first released as open-source software in September 2022. It is capable of transcribing speech in English and several other languages,[3] and is also capable of translating several non-English languages into English. The Whisper architecture uses an encoder-decoder transformer. It processes 30-second audio chunks converted into Mel-frequency cepstrum. The encoder handles audio, and the decoder predicts text captions, including phrase-level timestamps using special tokens.
Whisper v3, the newest version of OpenAI’s speech recognition model, offers improved accuracy and commercial applicability
[https://github.com/openai/whisper]
You can directly use whisper CLI or python module to transcribe and transalte
Also Open AI API supports Whisper model, cost associated to use the API - [https://platform.openai.com/docs/guid...]
The use of the models/python package is free under the MIT license.
What is Lightning AI?
Lightning AI,the vendor behind the popular open source Python library Pytorch Lightning, has launched Lightning AI Studios, a cloud-based platform for building and deploying AI products at scale. Lightning Studios has apps that do specialized work, such as code on the cloud, multi-node training, distributed data preparation, or hosting and sharing AI web apps. It integrates popular machine learning tools into a single interface, so you don't have to context switch between different tools. This allows for building scalable AI apps and endpoints more easily.
You can create a free account at lightning.ai/sign-up. To get instant verification, make sure to use an official company or .edu email.
There is no environment setup required. You can code in the browser or connect your local IDE (VSCode or PyCharm). You can also easily switch between CPU and GPU with no environment changes.
It allows hosting and sharing AI apps built with Streamlit, Gradio, React JS, etc.
It also enables multi-user collaboration by coding together.
It provides unlimited storage and the ability to upload and share files as well as connect S3 buckets.
It enables training models at a massive scale using thousands of GPUs (Paid option). You can run hyperparameter sweeps, data preprocessing, and model deployment massively in parallel.
Your Jupyter Notebook / VSCode on the cloud provides scalable hardware for training large language models and running fast inference.
The Lighting AI platform offers Studios that are project templates designed by users. These templates include code, environment settings, and data to help you commence your project. You can find various Studios on the platforms such as training, fine-tuning, preprocessing, inference, and hyperparameters sweep templates. You can easily search and scroll through these templates to find the one that suits your requirements.
[https://lightning.ai/pricing]
[https://github.com/openai/CLIP]
pip install git+https://github.com/openai/CLIP.git
python generate.py
pip install pipx
pipx install insanely-fast-whisper --force --pip-args="--ignore-requires-python"
sudo apt install ffmpeg
insanely-fast-whisper --file-name filename or URL