Alibaba Cloud Bailian Video to Text
Paste the following prompt into your AI chat to install this skill:
Follow https://skillhub.cn/install/skillhub.md to install @user_238e6fa6/aliyun-video2text.
About this skill
Problem
Short videos, screen recordings, or local video files often contain spoken narration that needs to become clean text. Manual dictation and line-by-line review are slow, while generic video analysis can lose speech detail. This skill converts the audio from common local video files into reusable Mandarin transcripts for editing, subtitle prep, and script archiving.
How it works
The skill runs a local transcription pipeline:
- extracts and re-encodes audio with ffmpeg to mp3
- uploads temporary audio to Alibaba Cloud OSS and obtains an accessible URL
- calls Alibaba Cloud Bailian Paraformer-v2 to create a transcription job
- polls the task status and parses the result
- emits JSONL, one JSON object per video, for batch reading
Scope and caveats
It supports common containers such as mp4, avi, mov, mkv, flv, and wmv, with a per-file limit of about 500MB. The transcription target is Mandarin Chinese. Audio with little speech, heavy noise, strong accents, or mixed-language content may produce lower-quality text. Missing credentials, OSS upload failures, or API errors are reported per file.
Use Cases
- A video editor extracts spoken dialogue from dozens of local vlog videos to prepare scripts for revisions.
- A records clerk batch-transcribes Mandarin product training videos into text to build a searchable knowledge base.
- A pipeline developer parses audio from video files and streams the results as JSONL for downstream NLP models.
Best For
- Video editors handling post-production who need to quickly convert spoken audio from local videos into editable text drafts.
- Engineers building multimedia pipelines who need to batch-extract video audio and output it as JSONL for downstream programs.
- Knowledge managers responsible for enterprise archiving who need to batch-transcribe local Mandarin training videos into searchable text.
Related Skills
Creates 10,000-30,000-word Tomato Novel short stories with trend research, topic selection, outlines, segmented writing, pacing control, and UTF-8 txt export.
An AI editorial master skill that manages the full content creation process from vague ideas to publishable drafts, covering topic selection, writing, review, and publication adaptation.
An AI skill that generates original lyrics and synthesizes a custom song based on user-provided stories, occasions, and details.
A prompting skill for Jimeng image-to-video covering single-image, first-last-frame, multi-frame, and multimodal reference modes, focusing on incremental motion, explicit stillness, and camera pacing.