ggml-org/whisper.cpp

Add ability to limit auto-detection to a subset of languages

Open

#1,242 opened on Sep 2, 2023

View on GitHub
 (5 comments) (2 reactions) (0 assignees)C++ (49,693 stars) (5,535 forks)batch import
enhancementgood first issue

Description

From what I can tell, auto-detection simply picks the language with the highest probability score. Sometimes, I know the language could only be 1 out of 5 possible languages. In such cases, it would be useful to be able to specify the possible languages, to improve the likelyhood of the auto-detection picking the correct language.

Contributor guide

Tech stack
cpp
Domain
machine learningaibackend
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
1-3 hours
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
fresh
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
clear
Prerequisites
C++whisper.cpp codebaselanguage detection logic
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
35
Research direction
First, examine the existing auto detection implementation in whisper.cpp, likely in whisper.cpp or common.cpp, to understand how languages are scored and the highest probability selected. Then, propose a modification to accept a user specified subset of languages (e.g., via an API parameter or environment variable) that filters the candidate languages before scoring. Consider adding a new function or modifying the existing whisper full default params to include a language filter. Test the change with a few languages to ensure detection accuracy improves.