n8n-nodes-tripo

An n8n community node for Tripo, the AI 3D generation platform. Turn text prompts, product photos or multiview shots into textured 3D models, retexture, convert, retopologize, segment, rig and animate them, generate reference images and turntable views, and poll long-running tasks — from your workflows and as tools for AI agents.

Package: @amdrautomate/n8n-nodes-tripo · Author: AMDR Automate · License: MIT

Installation

Follow the community nodes installation guide.

Self-hosted: Settings → Community nodes → Install and enter @amdrautomate/n8n-nodes-tripo.

To let AI agents use the node as a tool on self-hosted n8n, set N8N_COMMUNITY_PACKAGES_ALLOW_TOOL_USAGE=true.

Credentials

The node authenticates with a Tripo API key (sent as Authorization: Bearer).

1. Create an account in the Tripo developer console: developers.tripo3d.ai for international accounts (card billing) or developers.tripo3d.com for China mainland.
2. Open API Keys and create a key (it starts with tsk_). Copy it; it is shown once.
3. In n8n create a Tripo API credential, pick the Region that matches the site you signed up on, paste the key and click Test. The test reads your credit balance. A key used on the wrong region fails with Invalid API key.

Every task is billed in credits to the account that owns the key. Credits are frozen when a task starts and released again if it fails or is cancelled. Prices per task type are on the Tripo pricing page.

Tripo node

Models

Text/Image/Multiview to Model pick the geometry model from a dropdown: the H series (v3.1-20260211 default, v3.0, v2.5) for the highest fidelity and the P series (P2-20260801 preview with native quads, P1-20260311) for clean low-poly topology. Options cover face limit, texture and PBR, texture quality and version, geometry quality, quad output, smart low poly, part generation, auto size, orientation and seeds. Texture has its own texture model list (v3.5, v3.0, v2.5); Rig its rigging versions (v1.0, v2.5, v3.0); Retopologize its algorithm tiers (v1.0, v2.0, v3.0 with smart UV); Text to Image and Image to Image the Seedream, Banana and Chat Image models.

Inputs

Every image or model input takes one of:

• Binary Data of the incoming item (default property data) — the node uploads it to Tripo for you and passes the returned file_token.
• URL — a public http(s) link Tripo downloads.
• File Token — a file_token from File → Upload, handy when one file feeds several tasks.
• Task ID — the output of an earlier Tripo task, so you can chain Text to Image → Image to Model → Texture → Rig → Retarget → Convert without downloading anything in between.

Multiview inputs (Multiview to Model, Texture with multiview references) take one field per view (front required, at least two) that accepts a binary property name, URL or file token, or the task ID of an Image to Multiview task.

Waiting and output

All generation and processing calls create an asynchronous Tripo task. With Wait for Completion on (the default) the node polls GET /v3/tasks/{id} every three seconds until the task succeeds or Max Wait runs out (default 10 minutes; 3D generation usually takes 1–5 minutes). Turn it off to get the task_id back immediately and collect the result later with Task → Get, or let Tripo call an n8n Webhook node (configure the endpoint under Settings → Webhooks in the Tripo console).

A finished task is returned as one item: the task JSON (task_id, type, status, progress, output, credits_consumed, ...) plus, with Download Files on (the default), every result file as binary data named after its output key — model (GLB/FBX), pbr_model, base_model, rendered_image, uv_render, generated_image, front_view … Result URLs expire, so keep downloads on unless you only need the JSON. Rig Check returns riggable and the recommended rig_type in output.

Example: product photo → game-ready animated character

1. Image → Image to Image with template T-Pose on the product or character photo (Chat Image 2.5 Flare).
2. Model → Image to Model with input Task ID from step 1, model P2-20260801, options Quad on and Face Limit 20000.
3. Animation → Rig Check, then Animation → Rig (rigging model v3.0, spec Unity Humanoid) on the task ID from step 2.
4. Animation → Retarget with preset:idle, preset:walk, preset:run and output format FBX.
5. Hand the FBX binary to a storage node, or Model → Convert to USDZ for AR Quick Look.

Using the node as an AI-agent tool

The node is marked usableAsTool. Add it to an AI Agent, pick a resource and operation (for example Model → Text to Model), and let the agent fill the prompt with $fromAI(). URL and task-ID inputs work well for agents; binary inputs need the file on the incoming item.

Notes

• Tripo limits concurrent tasks per category (10 for H-series generation, 5 for P-series and model processing, 1 for image generation, 10 for animation and mesh operations) and rate-limits API calls per key. Over the limit the node fails with Tripo's 2000/1007 message; run items sequentially or add a Wait node.
• Errors carry Tripo's own message and suggestion plus a hint, for example 2010 → top up credits in the console.
• Smart UV needs an input of at most 80,000 triangles; run Retopologize first for denser meshes.
• Export Orientation on generation applies to that export only; when the model will be textured, rigged or converted afterwards, leave it at the default and set the orientation in the final Convert.
• The node has no runtime dependencies; the file upload is encoded by the node itself.

Compatibility

Built and tested with n8n 2.43 (Node.js 20 and 24). Uses the Tripo API v3 (https://openapi.tripo3d.ai/v3 for international accounts, https://openapi.tripo3d.com/v3 for China mainland).

Resources

• Tripo API documentation
• Tripo models and versions
• n8n community nodes documentation