音视频鸟鸣种类识别（birdsong-recognizer）

分享海报

@user_2f6efe1f/birdsong-recognizer

从音频或视频文件中识别鸟叫声对应的鸟种，基于康奈尔实验室 BirdNET v2.4 声学模型。自动抽取音轨（无需系统 ffmpeg），采用动态地理/物候权重表（非硬性白名单）做地理校正，并内置节律/颤音检测补齐模型对夜鹰、啄木鸟敲击等连续发声鸟类的识别盲区。支持对整个文件夹的批量识别（如声纹监测设备连续采样数据），输出标准化优先级分级报告，可直接用于指导人工标注排序。可生成语谱图核验。首次使用需联网下载约50MB模型。 Identify bird species from an audio OR video file using the BirdNET v2.4 acoustic model. Use when the user provides a media file (mp4/mov/wav/mp3/m4a/etc.) and asks "what bird is this", "识别鸟叫/鸟种", "分辨这是什么鸟", "音视频鸟鸣识别", or wants acoustic bird-call identification. Auto-extracts the audio track (no system ffmpeg required), applies a dynamic geographic/phenological genus-weighting table (not a hard whitelist), and includes a rhythm/churring detector covering BirdNET's blind spots for continuous vocalizations (nightjars, woodpecker drumming). Supports batch processing of an entire folder of recordings with a standardized priority-tier report for annotation triage. Can generate a spectrogram for verification. Fully portable — no hardcoded machine paths.