glyphcall

n8n nodes for Glyphcall: transcribe recordings to text and subtitles, summarize meetings, text to speech; and turn Word, Excel, PowerPoint and PDF files into text or Markdown and extracts named fields (invoice number, total, due date...) as JSON, from a f

Documentation

n8n-nodes-glyphcall

Glyphcall turns recordings and documents into data, and text into speech. Made by SpellzCode.

Two n8n community nodes.

Glyphcall Voice: recordings in, text and speech out:

  • Transcribe: speech to text (Whisper) for up to 60 minutes of MP3 or WAV in one call, with SRT or WebVTT subtitles and timed segments if you want them. 19+ languages, detected by itself.
  • Summarize: a meeting, call or voice note as a title, short summary, key points and action items (with owner and due date when someone said them), plus the transcript.
  • Text to Speech and Text to Speech (Lite): up to 2,000 characters as one MP3 in a natural voice (Deepgram Aura-2), written to a binary field you can send on to Telegram, Slack, Drive or e-mail.

Glyphcall Documents: files in, text and fields out:

  • Convert to Text: a Word (.docx), Excel (.xlsx), PowerPoint (.pptx) or PDF file as Markdown or plain text. Headings, lists and tables of a Word file are kept, each sheet of a workbook becomes a table, a deck comes slide by slide.
  • Extract Fields: the fields you name, as JSON. Give it an invoice and ask for supplier, invoice_number, due_date, total, currency; a field the file does not state comes back empty instead of guessed.
  • Extract Text From PDF: a PDF page by page, with its page count, title and author.

Every operation takes a file from an earlier node (an e-mail attachment, a download, a form upload) or a public URL. Files are read by the Glyphcall API and are not stored.

Installation

In n8n open Settings > Community Nodes > Install and enter n8n-nodes-glyphcall. See the n8n documentation on installing community nodes.

Credentials

The node needs a Glyphcall API key (it starts with adt_). Get one at https://glyphcall.com/credits: the page shows the key right after checkout. Create a Glyphcall API credential in n8n and paste it. The key is sent in the Authorization header and never in a URL.

What it costs

A call costs a fixed price in credits (1 credit = $0.001). Failed calls are not charged.

Operation Tool Price per call
Convert to Text /document/text $0.005
Extract Fields /document/extract $0.03
Extract Text From PDF /pdf/text $0.005
Transcribe /transcribe/long $0.05 (a whole recording, up to 60 minutes)
Summarize /transcribe/summary $0.05
Text to Speech /text-to-speech $0.08 (up to 2,000 characters)
Text to Speech (Lite) /text-to-speech/lite $0.04 (up to 2,000 characters, English)

Credits come in packs from $1 (1,000 credits) or by monthly plan; both give one key. The $9 monthly plan (15,000 credits) covers about 300 hours of transcription. These are the prices on 16 November 2026; the current ones are at https://glyphcall.com/tools.

Using it

  1. Add a node that outputs a file, for example Gmail Trigger with Download Attachments on, HTTP Request with a file response, or Read/Write Files from Disk.
  2. Add Glyphcall Documents, pick the operation, keep File Source on File From an Earlier Node and name the binary field in Input Binary Field (data by default; Gmail attachments are attachment_0, attachment_1, ...).
  3. For Extract Fields, list the field names you want, separated by commas.

The answer of Extract Fields looks like this:

{
  "type": "pdf",
  "data": { "supplier": "Nordlys Design AB", "invoice_number": "2026-0147", "due_date": "2026-12-03", "total": 14686.25, "currency": "SEK" },
  "missing": [],
  "charsRead": 294,
  "truncated": false
}

A ready workflow is in templates/invoice-attachment-to-sheet.json: every new Gmail message with an invoice attached becomes one Google Sheets row. Import it with Import from File and connect your accounts.

Good to know

  • Recordings for Glyphcall Voice: up to 25 MB by file or URL (MP3/WAV up to 60 minutes; M4A, OGG and WebM up to 5 MB for Summarize). Convert long video to MP3 first.
  • Document files up to 10 MB. The kind of file is read from the file itself, not from its name.
  • Scans and photos are read by OCR: PNG, JPEG and WebP images up to 3 MB, and scanned PDFs (up to three pages a call).
  • Invoice lines and other rows: write the field as line_items[description,quantity,unit_price,amount] and every row comes back with exactly those keys.
  • Old binary Office files (.doc, .xls, .ppt) and password-protected files are refused and not charged; save them as .docx, .xlsx or .pptx first.
  • Excel cells hold the values the file stores: a formula's last result, and dates as Excel's day numbers.
  • Extract Fields is read by an AI model from the first 24,000 characters of the file. Check values that matter, such as amounts, before you act on them.
  • The node can be attached to an n8n AI Agent as a tool.

Compatibility

Built with the n8n nodes API version 1 and tested with n8n 2.x.

License

MIT

Discussion