n8n-nodes-videocaptioner-asr

This is an n8n community node. It lets you transcribe audio into subtitles inside your n8n
workflows using the two free speech recognition interfaces that ship with
VideoCaptioner:

Both engines are free and require no credentials — they are the "免费在线" engines from
VideoCaptioner's ASR configuration docs. Source parity is intentional: the request flows,
headers, signing, polling and response parsing are ported one-to-one from
videocaptioner/core/asr/bcut.py and videocaptioner/core/asr/jianying.py.

Because this is a derivative work of GPL-3.0 code, the package is licensed under the
GNU General Public License v3.0 — see License and NOTICE.md.

n8n is a fair-code licensed workflow automation platform.

Installation
Operations
Credentials
Compatibility
Usage
Resources
Version history

Installation

Follow the installation guide in the n8n community nodes documentation.

npm install n8n-nodes-videocaptioner-asr

Operations

A single node, ASR Transcriber, with one operation: Transcribe audio.

Input

• Input Source — Binary File (read the audio from an input binary property) or File URL
  (download the audio first).
• Engine — B Interface (Bcut / Bilibili) or J Interface (JianYing / CapCut).
• Word Level Timestamps — return one entry per word instead of one entry per sentence.

Output

• Output Mode — how the result is emitted:

  • Single Item (default) — one item per input item. Combine with Output Format below.
  • Split Into Segments — one output item per segment, ready to be written to a sheet,
    database or one file per line. Output Format does not apply in this mode.

• Output Format — only shown for Single Item. Multi-select, so you get exactly what you
  need and nothing else (all five are selected by default, matching earlier behaviour):

  Format	Field	Description
  Plain Text (One Line per Segment)	txt	Plain transcript, one segment per line — a clean .txt script
  Plain Text (Single Line)	text	Whole transcript joined without separators — handy as an LLM prompt
  SRT Subtitle	srt	SubRip document with timestamps, ready to be written as a .srt file
  Subtitle Segments (JSON)	segments	Array of { index, text, startTime, endTime, start, end, words? } (startTime / endTime in milliseconds)
  WebVTT Subtitle	vtt	WebVTT document with timestamps, ready to be written as a .vtt file

  Every item always carries engine, engineLabel and segmentCount in addition to the
  selected formats; raw is added when Include Raw Response is enabled. Selecting only
  Plain Text (One Line per Segment) gives you just txt, selecting only SRT Subtitle
  gives you just srt.

Options

Credentials

No credentials are required. The B and J interfaces are public free services; the node simply
talks to them directly. See the notes on rate limits below.

Compatibility

• Requires n8n 1.x and Node.js 22 or higher.
• n8nNodesApiVersion: 1
• No runtime dependencies: CRC32, AWS Signature V4 and the device fingerprint are implemented
  locally on top of Node's built-in crypto module, and waiting is done with the sleep helper
  exported by n8n-workflow.

Known limitations

• The free services only accept audio containers (mp3, wav, m4a, flac). A video file
  must be converted to MP3 first (for example with n8n's Execute Command node and ffmpeg).
• Long audio is uploaded in one piece. VideoCaptioner splits audio into 10-minute chunks with
  10-second overlap before calling these same services — for very long files you should chunk
  upstream (or reuse VideoCaptioner) rather than relying on this node.
• Both services are shared community endpoints with rate limits (VideoCaptioner caps them at
  100 calls / 360 minutes per 12 hours) and no availability guarantee. The J interface additionally
  depends on a third-party signing service, which is configurable via Sign Service URL.
• Input audio always leaves your n8n instance and is uploaded to Bilibili / ByteDance. Do not use
  these engines for confidential material.
• The B and J endpoints are unofficial and may change without notice. If they do, the node
  needs a matching update — the request flow lives in nodes/AsrTranscriber/shared/.

License

GPL-3.0-only. See LICENSE.md and NOTICE.md.

This package is a TypeScript port of GPL-3.0 code, so it inherits that license. Two practical
consequences:

• It can be published to npm and installed in self-hosted n8n without any problem.
• It is not eligible for n8n's verified-node program (the nodes panel in n8n Cloud), because
  n8n's verification guidelines
  require verified community nodes to be MIT licensed. n8n additionally states that a package
  should integrate a single third-party service; this one wraps two.

Accordingly, n8n.strict is false in package.json and eslint.config.mjs switches off the
single rule that demands an MIT declaration (community-package-json-license-not-default), with
the reason documented inline. Every other rule of n8n's preset remains active, and both
npm run lint and npm run build pass.

Usage

Minimal workflow:

Read Binary File  →  ASR Transcriber (Binary File, B Interface)
                  →  Convert to File (name: transcript.srt, content: {{ $json.srt }})

Using a remote URL:

HTTP Request (download audio)  →  ASR Transcriber (File URL, J Interface, Word Level Timestamps)
                               →  Split Into Segments  →  Google Sheets

The srt and vtt fields are plain strings, so they can be handed straight to Convert to File
or Move Binary Data to produce subtitle files.

Development

npm install          # installs @n8n/node-cli, eslint, prettier, typescript
npm run dev          # start n8n with this node loaded (http://localhost:5678)
npm run lint         # n8n-node lint
npm run build        # n8n-node build -> dist/

The scripts/ folder contains the verification harness used while porting the two engines:

npm run build
<python> scripts/reference_values.py > ref.json
node scripts/compare_reference.js ref.json
node scripts/smoke.js ./sample.mp3 both

Before publishing, replace the placeholder author, homepage and repository values in
package.json (they currently read your-github-username / dev@example.com) with your own
details — n8n's linter rejects unresolved template placeholders and empty author fields.
Publishing must go through the bundled GitHub Actions workflow
(.github/workflows/publish.yml) so the package carries an npm provenance statement.

Resources

• n8n community nodes documentation
• VideoCaptioner — the upstream project these
  interfaces were extracted from
• VideoCaptioner ASR configuration

Version history

• 0.1.0 — Initial release. B interface (Bcut) and J interface (JianYing) transcription, binary
  and URL input, word-level timestamps, single-item and split-segment output, SRT/TXT/VTT
  generation.