Whisper
#6
Whisper is a cutting-edge artificial intelligence developed by OpenAI that is capable of recognizing and transcribing human speech with impressive accuracy and robustness. It has been trained on an enormous dataset of 680,000 hours of multilingual and multitask supervised data collected from the web, which has allowed it to achieve human-level performance on English speech recognition. In addition to its ability to transcribe speech, Whisper is also able to translate spoken words from multiple languages into English. OpenAI is open-sourcing the models and inference code for Whisper, making it available for use in a variety of applications and as a foundation for further research on speech processing.
Highlights
- Introduction of Whisper
A state-of-the-art automatic speech recognition system trained on 680,000 hours of multilingual and multitask supervised data collected from the web. - Architecture of Whisper
An end-to-end approach implemented as an encoder-decoder Transformer, trained on a large and diverse dataset of audio to text. - Robustness of Whisper
Whisper demonstrates improved robustness to accents, background noise, and technical language. It makes 50% fewer errors than other models and is effective at learning speech to text translation. - Multilingual Capability of Whisper
Whisper is capable of transcribing in multiple languages and translating to English from those languages. - Open Source of Whisper
Open sourcing models and inference code to serve as a foundation for building useful applications and for further research on robust speech processing. - ASR Training Data Inputs and Outputs
A third of Whisper's audio dataset is non-English, and it is alternately given the task of transcribing in the original language or translating to English. - ASR Summary of Model Architecture
Whisper's architecture is a simple end-to-end approach implemented as an encoder-decoder Transformer. Input audio is split into 30-second chunks and then converted into a log-Mel spectrogram. - Whisper Examples
Whisper demonstrates its high accuracy and ease of use with examples like speed talking, K-Pop, French accent, and other diverse datasets. - Other ASR Approaches
Other existing approaches frequently use smaller, more closely paired audio-text training datasets, or use broad but unsupervised audio pretraining. - Whisper and Developers
We hope Whisper’s high accuracy and ease of use will allow developers to add voice interfaces to a much wider set of applications. - References
Whisper's research references include Solving Some Formal Math Olympiad Problems, Hierarchical Text Conditional Image Generation With Clip Latents, and more. - OpenAI company
A machine learning research institute that builds and publishes free AI systems that help solve some of the world's most pressing and most difficult problems.
#AI
#OpenAI
#SpeechRecognition
#Multilingual
#Multitask
Top 16 Alternatives To "Whisper"
#1
#GPT3
#ChatGPT
#OpenAI
#ConversationalAI
#NaturalLanguageGeneration
#2
#OpenAICodex
#AI
#NaturalLanguage
#GitHubCopilot
#Programming
#3
#DALLE2
#AI
#AIArt
#ImageGeneration
#ImageManipulation
#4
#Creativity
#Innovation
#Imagination
#Discover
#Share
#5
#ContentCreation
#AI
#RealtimeCollaboration
#AImagicTools
#Inpainting
#7
#GPT3
#OpenAI
#NLP
#AI
#LanguageGeneration
#8
#InstructGPT
#OpenAI
#RLHF
#LanguageProcessing
#GPT3
#9
#BeautifulAi
#PresentationSoftware
#ProfessionalDesign
#RealTimeFormatting
#Branding
#10
#WhisperAI
#AIHearingSystem
#SoundSeparationEngine
#RealTimeProcessing
#CrystalClearHearing
#11
#StableDiffusion
#DigitalArt
#Creative
#OnlineTool
#Privacy
#12
#Ghostwriter
#AI
#CodingAssistant
#CodeFaster
#Refactoring
#13
#NotionAI
#AI
#ContentCreation
#Creativity
#Notion
#14
#CRM
#Dactyl
#LeadsManagement
#LeadGeneration
#SalesAutomation
#15
#Tome
#Storytelling
#Narrative
#AI
#CreativeContent
#16
#Superhuman
#EmailSoftware
#AITriage
#UndoSend
#SocialMediaInsights
#17
#FakeNewsDetection
#NeuralNetworks
#Grover
#StateOfTheArt
#AI
Submit an entry
Is there an entry, which should be added to this list?