Transcribe audio and video with Volcengine ASR and auto rough-cut talking-head footage.
npx skills add https://github.com/mcncarl/yichen-skills --skill yichen-volc-asr
Please help me install this Agent Skill. Skill: yichen-volc-asr Source: https://github.com/mcncarl/yichen-skills/tree/main/yichen-volc-asr Command: npx skills add https://github.com/mcncarl/yichen-skills --skill yichen-volc-asr
Manually transcribing talking-head videos and cutting out filler words, repetitions, and long pauses is slow and error-prone. This Skill automates speech-to-text with timestamps and produces a rough-cut MP4 plus SRT subtitles ready for final editing in Jianying/CapCut.
Ask the assistant to transcribe and rough-cut your video by saying: help me transcribe this video and auto-cut the filler words, repetitions, and pauses into a rough-cut MP4 with subtitles.
High-intent search queries and answers about installing and using this skill.
Run the transcribe.py script on a working copy of your video with the --no-execute flag. It uploads the file to TOS, submits a Volcengine ASR bigmodel task, and saves full text with per-segment timestamps plus a reusable cache file.
The script detects pure filler segments, silence gaps over one second, and adjacent repeated sentences using difflib similarity, then renders a trimmed MP4 with ffmpeg trim and concat filters. Smart margins and smoothing prevent clipped words and flickering frames.
You need TOS_ACCESS_KEY, TOS_SECRET_KEY, and TOS_BUCKET for upload, plus VOLC_ASR_TRIAL_APP_ID and VOLC_ASR_TRIAL_TOKEN for the trial app. Paid usage requires VOLC_ASR_PAID_APP_ID and VOLC_ASR_PAID_TOKEN after marking the trial exhausted.
Yes. After submission it writes an asr_pending.json file with the request_id, and reruns resume the pending query instead of resubmitting. Only passing --force-new uploads and submits a new task.
Yes. The script generates an SRT subtitle file from the kept segments with standard HH:MM:SS,mmm timestamps, which imports directly into Jianying or CapCut alongside the rough-cut MP4.
barkleesandersEdit raw video footage into polished final.mp4 files via conversational AI.
Leonardo-CorteEdit CapCut and JianYing draft JSON files to apply animations, transitions, and subtitle imports.
CeeonTranscribe Chinese talking-head videos and produce a human-reviewed word-deletion ledger.
vahnxuTranscribe audio files via Doubao Seed-ASR 2.0 API with speaker diarization.
