Tripo H3.1

v3.1-20260211Default

A new class of intelligence for 3D generation and professional work.

Quick Start

Quality

Highest

Speed

~ 40s

No Texture

Speed

~ 120s

With Texture

Face Limit

Up to 2,000,000

Input

Text, Image, Multiview

Output

Mesh + PBR

Tripo v3.1 is our newest flagship model for the most complex 3D generation work. It delivers top-tier geometry with PBR textures, with face counts reaching up to 2 million, making it the best choice when fidelity matters more than seconds. It supports text prompts, single images, and multiview inputs natively.

Best-in-class geometry and PBR textures
Supports quad mesh output (FBX)
Supports smart low-poly topology

Pricing

Pricing is based on the task type and specific parameters. Generating standard textures or using premium topology options incurs additional costs.

Task Type	No Texture	Standard Texture	Detailed Texture	8K Texture
Text to 3D	¥0.7010 credits	¥1.4020 credits	¥2.1030 credits	¥2.8040 credits
Image to 3D	¥1.4020 credits	¥2.1030 credits	¥2.8040 credits	¥3.5050 credits
Multiview to 3D	¥1.4020 credits	¥2.1030 credits	¥2.8040 credits	¥3.5050 credits

Detailed geometry quality: +20 credits (¥1.40)

Quad mesh: +5 credits (¥0.35)

Smart low-poly: +10 credits (¥0.70)

Generate in parts: +20 credits (¥1.40)

Endpoints

Click any of the endpoint cards below to jump directly to the detailed API documentation and code examples.

Text to 3D Model

v3/generation/text-to-model

Image to 3D Model

v3/generation/image-to-model

Multiview to 3D Model

v3/generation/multiview-to-model

Parameters

model

Model version: v3.1-20260211 for H3.1; v3.0-20250812 and v2.5-20250123 remain available

face_limit

Up to 1,500,000, or 2,000,000 with geometry_quality=detailed; up to 150,000 with quad=true; adaptive if omitted

geometry_quality

standard (default) or detailed (Ultra, finer geometry)

quad

Output a quad mesh in FBX (default: false); 10,000 faces if face_limit is omitted

smart_low_poly

Clean, hand-crafted low-poly topology for simple inputs (default: false); 500 to 20,000 faces, or 500 to 10,000 with quad=true

generate_parts

Generate editable segmented parts (default: false); requires texture=false and pbr=false

model_seed

Geometry seed; the same seed and input produce the same mesh

texture

Enable texture maps (default: true)

pbr

Enable PBR materials (default: true); forces texture to true

texture_quality

standard (default), detailed, extreme (8K), or fast (requires texture v3.5-20260815)

texture_version

Texture model; v3.x defaults to v3.0-20250812, and fast or delight require v3.5-20260815

texture_seed

Texture seed; keep model_seed and change texture_seed for new textures on the same geometry

delight

Remove lighting from the reference image before texturing (default: true; v3.5-20260815 only)

auto_size

Scale the model to real-world size in meters (default: false)

export_uv

UV unwrapping (default: true); set to false for faster generation and smaller files

export_orientation

Export forward axis: +x, -x, -y or +y (default: +x); leave unset if you plan post-processing

compress

Set to geometry for meshopt compression and smaller files

Rate limits

Rate limits ensure fair and reliable access to the API by placing specific caps on parallel requests. Your usage tier determines how high these limits are set.

Category	Concurrency Limit
3D Generation — H Series	10 parallel tasks

Snapshots

Snapshots let you lock in a specific version of the model so that performance and behavior remain consistent.

v3.1-20260211

Current stable release