C

Crisperwhisper

Developed by nyrahealth
CrisperWhisper is an advanced variant of OpenAI Whisper, designed for fast, precise, and word-for-word speech recognition, providing accurate (clear) word-level timestamps.
Downloads 10.23k
Release Time : 8/29/2024

Model Overview

CrisperWhisper is an advanced variant of OpenAI Whisper, designed for fast, precise, and word-for-word speech recognition, providing accurate (clear) word-level timestamps. Unlike the original Whisper which tends to omit disfluencies and adopts a more paraphrased transcription style, CrisperWhisper aims to precisely transcribe every spoken word, including fillers, pauses, stutters, and false starts.

Model Features

Precise word-level timestamps
By adjusting the tokenizer and using custom attention loss during training, it provides precise timestamps even when handling disfluencies and pauses.
Word-for-word transcription
Accurately transcribes every spoken word, including and distinguishing fillers like 'um' and 'uh'.
Filler word detection
Detects and accurately transcribes filler words.
Reduced hallucinations
Minimizes hallucinations in transcription, improving accuracy.

Model Capabilities

Speech recognition
Word-level timestamp generation
Filler word detection
Multilingual support

Use Cases

Speech transcription
Meeting minutes
Used for accurately recording meeting content, including all disfluencies and filler words.
Provides word-for-word transcription and precise timestamps.
Academic research
Used for transcribing interviews and research data, ensuring all spoken details are accurately recorded.
Highly accurate word-for-word transcription.
Speech analysis
Speech behavior analysis
Analyzes speakers' disfluency patterns and filler word usage.
Provides detailed speech behavior data.
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
© 2025AIbase