Description

n8n-nodes-speechall

Speechall is a unified speech-to-text API for transcribing audio with multiple providers and models. This package adds a verified-ready n8n community node for Speechall batch transcription workflows.

Installation

Install n8n-nodes-speechall as a community node in n8n Cloud or self-hosted n8n.

For local development:

nvm use
npm ci --ignore-scripts
npm run generate:openapi-types
npm run lint
npm run test
npm run build

This repository keeps runtime dependencies empty for n8n Cloud verification readiness. Development tooling is installed only as devDependencies.

Use Node 22 LTS for local n8n tooling. Newer non-LTS Node releases can fail while installing n8n development dependencies with native modules such as isolated-vm.

Releases

All npm releases are published by the GitHub Actions publish workflow through npm Trusted Publishing with provenance. Do not add an NPM_TOKEN repository secret and do not run npm publish locally.

To create a release:

git switch main
git pull --ff-only
npm ci
npm test
npm run release

Select the appropriate version increment and approve the release commit, version tag, and push. Tags use the version without a v prefix, such as 0.1.4. The pushed tag triggers GitHub Actions, which runs lint and build checks before publishing to npm.

After the workflow succeeds, confirm the new version and provenance on the npm package page. The package’s npm publishing access must remain configured for the Speechall/speechall-n8n-node repository and publish.yml workflow.

Before submitting a published version to the n8n Creator Portal, run the official community package scanner:

npx @n8n/scan-community-package n8n-nodes-speechall

Local n8n UI Testing

Use this flow when you want to test the node in a real local n8n editor before publishing it.

Recommended: install the packed node into a local n8n user folder

This is the closest local equivalent to installing the package as an n8n community node. It avoids development hot-reload edge cases and uses the stable node type name n8n-nodes-speechall.speechall.

nvm use
npm ci --ignore-scripts
npm run build
PACKAGE_TARBALL=$(npm pack --silent)
REPO_ROOT=$(pwd)
mkdir -p /tmp/speechall-n8n-ui/.n8n/nodes
cd /tmp/speechall-n8n-ui/.n8n/nodes
npm install "$REPOROOT/$PACKAGETARBALL" --ignore-scripts
N8NUSERFOLDER=/tmp/speechall-n8n-ui npx n8n@latest

Then open:

http://localhost:5678

Important details:

• N8NUSERFOLDER should be /tmp/speechall-n8n-ui, not /tmp/speechall-n8n-ui/.n8n. n8n creates and reads its .n8n directory inside the user folder.
• The installed community node package lives under /tmp/speechall-n8n-ui/.n8n/nodes.
• After code changes, repeat npm run build, npm pack --silent, and the npm install ...tgz command, then restart n8n.
• Stop n8n with Ctrl+C in the terminal running npx n8n@latest.
• The temporary user folder keeps local credentials and workflows between restarts. Remove /tmp/speechall-n8n-ui only when you want to reset the local n8n instance.

If you prefer to keep the repository path explicit:

REPO_ROOT=/path/to/speechall-n8n-node
PACKAGETARBALL=$(cd "$REPOROOT" && npm pack --silent)
cd /tmp/speechall-n8n-ui/.n8n/nodes
npm install "$REPOROOT/$PACKAGETARBALL" --ignore-scripts

Development hot-reload mode

The package also supports n8n-node dev:

nvm use
npm run dev -- --custom-user-folder /tmp/speechall-n8n-ui

This mode is useful while editing node descriptions or operation code, but it may register a temporary development node type such as CUSTOM.speechall. If n8n reports Unrecognized node type: CUSTOM.speechall, stop the dev server and use the packed-node flow above.

Quick smoke test

1. Create a Speechall API credential and run the credential test.
2. Create Manual Trigger -> Speechall.
3. Run List Speech-to-Text Models and confirm one item is returned per model.
4. Run Transcribe a Remote URL with a public MP3 URL and confirm output includes text, outputFormat, model, source: "remote_url", and sourceUrlSanitized.
5. Keep Include Source URL in Output disabled unless you intentionally want the full URL in output.

Credentials

Create a Speechall API credential and paste your Speechall API key into the masked API Key field. The credential sends requests to https://api.speechall.com/v1 with a Bearer token and tests access by calling /speech-to-text-models.

Operations

Transcribe a File

Transcribes audio from n8n binary data using POST /transcribe.

Important fields:

• Binary Property: Defaults to data.
• Model Selection: Choose a model from Speechall or enter a model ID manually.
• Language: Defaults to en; use auto where supported.
• Output Format: text, json_text, json, srt, or vtt.
• Also Return as Binary File: Adds .txt, .srt, or .vtt transcript binary output.

Transcribe a Remote URL

Transcribes a public HTTP or HTTPS audio URL using POST /transcribe-remote.

Remote URL output includes source: "remote_url" and sourceUrlSanitized by default. The full fileUrl is included only when Include Source URL in Output is enabled, because signed storage URLs can contain secrets.

Remote URL transcription also supports inline replacement rules:

• Exact match
• Regex
• Regex group replacements

List Speech-to-Text Models

Calls GET /speech-to-text-models, applies optional client-side filters, and returns one n8n item per model. The dynamic model dropdown uses the same endpoint, while manual model ID mode remains available if model loading fails.

Output Formats

• text: Plain transcript text.
• json_text: JSON response with transcript text and metadata.
• json: Detailed JSON, preserving segments, words, timestamps, speaker labels, and provider metadata.
• srt: SRT subtitles.
• vtt: WebVTT subtitles.

Text-like responses are normalized into item JSON with text, outputFormat, model, language, and source. Detailed JSON responses preserve unknown Speechall/provider fields.

Advanced Options

The transcription operations expose punctuation, diarization, initial prompt, temperature, expected speaker count, custom vocabulary terms, reusable ruleset_id, and request timeout. Provider and model support varies; unsupported combinations are surfaced as Speechall API errors.

The default request timeout is 300 seconds and can be set from 1 to 900 seconds. n8n instance-level workflow execution timeouts may still end long-running executions. For large hosted files, prefer Transcribe a Remote URL to avoid moving large binary data through n8n memory.

OpenAI-Compatible Endpoint

Speechall also offers OpenAI-compatible transcription APIs for generic HTTP/OpenAI-shaped workflows. This dedicated Speechall node v1 intentionally exposes Speechall-native operations only.

Troubleshooting

• 401: Invalid or missing Speechall API key.
• 402: API key may be valid, but the Speechall account has a billing or credit issue.
• 429: Rate limit exceeded; retry guidance is preserved when Speechall sends Retry-After.
• 504: Upstream transcription provider timed out.
• Missing binary data: Check the incoming item has the configured binary property.
• Remote URL errors: Confirm the URL is public and starts with http:// or https://.

Example Workflows

• Google Drive audio to transcript
• Webhook audio upload to JSON transcript
• Remote podcast URL to SRT
• Meeting recording to diarized transcript

Support

Speechall documentation: https://docs.speechall.com

Repository: https://github.com/Speechall/speechall-n8n-node

License

MIT