This is a simplified guide to an AI model called Force-Align-Wordstamps maintained by Cureau. If you like these kinds of analysis, you should join AImodels.fyi or follow us on Twitter.
Model Overview
force-align-wordstamps provides word-level timestamp alignment between audio files and transcripts. Unlike similar solutions like whisper timestamped or whisperx, this model excels at matching existing transcripts to audio with high precision. Created by cureau, it builds on stable-ts technology to deliver reliable results even with background noise.
Model Inputs and Outputs
The model takes an audio file and reference transcript text to generate precise word-level alignments. This approach differs from pure transcription models by using the provided transcript as ground truth.
Inputs
- Audio File: MP3 format audio input
- Transcript: Text string containing the known transcript
- Show Probabilities: Optional boolean flag to include confidence scores
Outputs
The model returns a JSON object containing an array of words with their corresponding timestamps:
- Word: Individual word from the transcript
- Start Time: Timestamp for word start
- End Time: Timestamp for word end
- Probability: Optional confidence score for each word
Capabilities
The alignment system handles noisy audi...
Top comments (0)