Back to Nodes

Project

Last updated Sep 23, 2026

Rendley node for n8n. Create, edit, and automate video with the Rendley AI agent, AI video, image, music and speech generation, transcription, dubbing, lip sync, media uploads, brand kits and MP4 rendering.

18 Weekly Downloads
438 Monthly Downloads

Included Nodes

Project

Description

Rendley

n8n-nodes-rendley

![npm](https://www.npmjs.com/package/n8n-nodes-rendley) ![CI](https://github.com/rendleyhq/n8n-nodes-rendley/actions/workflows/ci.yml) ![License: MIT](LICENSE)

Rendley lets you create, edit, and automate video. This node
puts Rendley in your n8n workflows. The AI Video Agent turns a plain-language prompt and
your footage into a finished video on a real editing timeline. Around it, AI media
operations generate video, images, music and speech and transform existing media, and
utility operations upload files, read brand kits, export MP4s and follow jobs.

This is an n8n community node. It
has no runtime dependencies, talks only to the Rendley API through n8n’s request
helpers, and is marked usableAsTool, so an n8n AI Agent can call any of its
operations.

Installation · Credentials · AI Video Agent
· Other operations · Following long jobs
· Templates · Development · Release

Installation

  • n8n Cloud and self-hosted 1.94+: open the nodes panel, search for Rendley, and
  • install it from the More from the community section (available once the package is
    verified by n8n).

  • Self-hosted, any version: Settings → Community Nodes → Install and enter
  • n8n-nodes-rendley.

    Requires an n8n version that supports community nodes and Node.js 20 or newer on
    self-hosted instances.

    Credentials

    Create a Rendley API credential and paste an API key from
    app.rendley.com/settings. The API Base URL is
    pre-filled with https://api.rendley.com/v1; leave it unless Rendley gave you another
    host. The credential test calls GET /workspaces.

    AI actions, agent runs and exports use Rendley credits. When an operation cannot run,
    Rendley answers with HTTP 402 and the node reports the message it sends.

    AI Video Agent

    The AI Video Agent resource is the heart of the node. Give Run a prompt and,
    optionally, some files, and the agent edits like a person would, on a real timeline it can
    keep editing later.

  • Turn raw footage into a social clip, reframed to 9:16 with captions and music.
  • Cut bad takes, silences and filler words out of an interview or a talking-head recording.
  • Add styled captions, titles and b-roll.
  • Build a video from scratch from generated clips, images, voiceover and music.
  • Reframe one video for several platforms.
  • Apply any edit you can describe, and continue the conversation on the same project with
  • a thread ID.

    | Operation | What it does |
    | — | — |
    | Run | Sends the prompt, optional Files (public URLs, each imported into the project before the agent starts), an optional Project (leave it empty to create one) and an optional Thread ID. With Export Video After Edit on, the node exports the project once the edit lands and returns url. |
    | Get Job | Reads an agent job without waiting, for workflows that poll on their own. |
    | Cancel Job | Stops a running agent job. |

    The output carries jobid, projectid, threadid, status, lastmessage and
    commandsapplied, plus url and exportjob_id when exported. The agent runs
    non-interactively. If it pauses to ask a question the node fails with the question in
    the error, so rephrase the prompt to remove the ambiguity.

    The Edit resource offers the same agent behind curated prompts, so results are
    repeatable. Every operation takes the project, optional files, **Export Video After
    Edit, and an Additional Instructions** option. Reframe takes an aspect ratio (9:16,
    1:1, 4:5, 16:9), Create Shorts takes a count and length bounds, Remove Filler Words a
    minimum silence.

    Other operations

    Most operations are asynchronous. The node starts a job and, when **Wait for
    Completion is on (the default), polls until it finishes. Poll Interval (Seconds)**
    (default 10, minimum 5) and Timeout (Minutes) (default 60) are set per operation.

    | Resource | Operations |
    | — | — |
    | Edit | Remove Filler Words, Auto Edit, Reframe, Add Captions, Create Shorts, Custom Prompt |
    | Video | Generate, Transcribe, Dub, Lip Sync, Isolate Voice, Change Voice, Remove Background, Upscale |
    | Image | Generate, Upscale, Remove Background |
    | Audio | Text to Speech, Generate Music, Generate Sound Effect |
    | Export | Export Video, Estimate Cost, Get Job |
    | Media | Upload, Get Download URL, List |
    | Brand Kit | Get, Import From Website |
    | Project | Create, Get, List, Delete |

    Video, Image and Audio

    Each AI operation takes an optional Project (leave it empty to save the result to the
    workspace library, with a Workspace picker for accounts that have several), an
    optional Model picked from the models that action supports, its typed options, an
    Additional Parameters (JSON) escape hatch for model-specific fields documented in the
    model catalog, and an Estimate Cost Only toggle
    that returns the credit price without running anything.

    Source files (Transcribe, Dub, Lip Sync, Isolate Voice, Change Voice, Remove Background,
    Upscale) accept a public URL, a media ID, or the file hash of an upload in the
    project. Rendley resolves the reference and probes duration itself.

    When the node waits for completion the output carries url (a fresh signed
    URL), urlexpiresat, mediaid and filehash. Transcribe returns the transcript in
    result_data instead of a file.

    Export

    Export Video exports a project with Export Settings (codec, quality, target
    resolution) and returns url when waiting. Estimate Cost returns the credit
    price. Get Job reads any Rendley job by ID.

    Media

  • Upload puts a file into a project. Source is either Binary Data (a file from
  • an earlier node such as Google Drive or an email attachment, up to 50 MB) or a public
    URL, which Rendley fetches server-side with no size limit through n8n. Returns
    mediaid and filehash, both usable as the file reference of AI operations.

  • Get Download URL resolves a media ID or file hash to a fresh signed URL.
  • List returns every upload in a project with a fresh url.

Signed URLs expire after a few hours. Copy files to your own storage in the same
workflow rather than saving a Rendley URL for later.

Brand Kit and Project

Read a workspace brand kit or import one from a public website. Create, get (with a
Simplify toggle that leaves out the editor document), list and delete projects.
Leaving the workspace empty uses the account’s first workspace.

Following long jobs

For jobs that can run for many minutes, or for large batches, turn **Wait for
Completion off and poll yourself with an n8n Wait** node of 70 seconds or more, then
AI Video Agent > Get Job or Export > Get Job, then an If that loops until status is
completed. n8n offloads waits of that length, so they cost no execution time. Rendley
sends no outbound webhooks.

Templates

Ready-to-import workflows live in templates/. Each carries sticky notes
covering the rough credit cost and the link-expiry warning. See
templates/README.md.

Development

npm install
npm run typecheck   # tsc
npm run lint        # n8n-node lint (the ruleset n8n verifies with)
npm run build       # n8n-node build -> dist/
npm test            # drives the compiled node against a stub of the Rendley API
npm run scan        # @n8n/scan-community-package on the built package
npm run dev         # runs n8n locally with this node linked

npm run test:live runs the compiled node against a real Rendley API. It needs
RENDLEYAPIKEY (it spends a few credits) and optionally
RENDLEYAPIBASE_URL.

Release

Publishing goes through GitHub Actions with npm provenance, as n8n requires for
verified community nodes.

License

MIT