• AIGC
  • B2B2C
  • Interior Design

Taimer.ai

An AIGC platform that pairs generative models with designer fine-tuning to turn a renovation brief into client-ready interior renders — in minutes, not weeks.

Role
Product Builder
Team
Taimer.ai · team of 5
Timeline
7 months · 2023
Tools
Figma

Overview

A brief to client-ready renders, in minutes not weeks.

China's home-interior industry runs on slow, manual work: long back-and-forth with homeowners, renders drawn by hand, and lead-to-deal conversion that rarely pays back the cost of acquiring the lead. Taimer.ai set out to collapse that loop.

Over seven months, working with a team of five (a PM, me, a marketer, and two engineers), I designed a multi-modal AIGC platform — text-to-image, image-to-image, recognition, and 3D — that lets a designer fine-tune a model to their own house style, then generate client-ready interior renders in minutes. The work won the Outstanding Project award at the Tsinghua University AIGC Application Innovation Challenge and drew investment from Beijing institutions and angel backers.

The Taimer.ai workspace — a sidebar of creation and training tools, a grid of generated interior renders, and a right-hand panel of style and generation controls
The workspace — custom style models, an image library, and generation controls in one place.

The problem

A stack of small frictions, plus four structural barriers.

I started by mapping where designers and homeowners actually lose time. The friction wasn't one big thing; it was a stack of small ones.

High communication cost

Endless back-and-forth with homeowners drains time and energy.

Slow manual production

Producing renders by hand is painstaking and hard to scale.

Low conversion

Leads rarely turn into signed projects.

Blurry references

Stock imagery is unclear and low quality.

No acquisition channels

Designers lack reliable ways to reach new clients.

Lack of inspiration

Fresh creative starting points are hard to find.

On top of the day-to-day friction sit four structural barriers: the expertise the work demands, the difficulty of guaranteeing detail quality, cross-disciplinary tech that's gated abroad, and a delivery experience that keeps cost and timelines high.

Market opportunity

The upside of removing that friction is large.

Removing the friction pays off for the people doing the work and for the market around it.

1wk → 1h

Revision turnaround

+45%

Design-to-deal conversion

1 min

Brief → first proposal

$500B

N. American market

9.5M+

Designers in China

20,000+

Beta renovation users

The product

A generation surface designers actually control.

Taimer is a community-plus-creation platform. The core is a generation surface designers actually control, wrapped in tools that make a personal model worth building.

Interior generation

Upload a room, choose a type, then generate material styles, sketches, and floor plans — and expand, redraw, or annotate any region.

Creator tools

Custom style models, a style plaza, a personal image library, and asset management.

Fine-tuning controls

Image count, clarity, steps, learning rate, text intensity, sampler, seed, and copyright protection.

A horizontal strip of six AI-generated interiors — bedrooms in varied styles and a set of bar stools
Sample interiors generated through the platform.

How it works

Fine-tuning is the feature — a designer’s style, not a generic one.

Under the surface, Taimer runs a proprietary deep-learning fine-tuning solution built on ControlNet and diffusion models, with multi-modal support and compute backed by AWS and NVIDIA. The point of the fine-tuning layer is control: a designer's model should reproduce their style, not a generic one.

Diagram of the Stable Diffusion and ControlNet-LITE pipeline, from input conditioning through the model to a generated interior render
The Stable Diffusion + ControlNet-LITE pipeline, from input conditioning to output.

Model training

Five stages to a style model

  1. 01Preprocess & label the designer's reference images.
  2. 02Prepare the datasets and configure the run.
  3. 03Train on the fine-tuning stack.
  4. 04Compare & select from candidate results.
  5. 05Tune weights until the output matches the target style.
Screenshots of the model-training process — data preprocessing and labeling, a training run in progress, and a grid for comparing and selecting results
Data preprocessing, training, and result selection.

Who it’s for

Built for independent designers and small studios.

The design centered on independent designers and small studios — people who need volume and consistency without a render farm.

Alex

33 · Changsha · Interior Designer

Motivation

  • Serve more clients and finish projects faster
  • Generate large volumes of low-cost drafts with AI
  • Offer varied styles for different clients and markets

Needs

  • Models that match the firm's house style
  • Convenient fine-tuning tools
  • Less manual back-and-forth with homeowners

Business model

Designers subscribe; product brands pay per usage.

The tiers map to how many drafts a studio actually ships in a month.

Basic

¥59/mo

10 orders / month

Standard

¥399/mo

100 orders / month

Pro

¥1999/mo

400 orders / month

Inventory listing

¥699

Upload 10 SKUs · pay-per-usage

Behind the subscription, a designer requests a customized model, a training expert builds and tracks it, candidate outputs are reviewed together, and the result is delivered — with evaluation and after-sales support if it isn't right the first time.

Impact

From a week to an hour, and +45% conversion.

1wk → 1h

Revision turnaround, down from a week to an hour

+45%

Lift in design-to-deal conversion

20,000+

Renovation users reached in beta

Beyond the numbers, Taimer opened a new income stream for independent designers and gave consumers a one-minute path from brief to first proposal. The lesson that stuck with me: in an AIGC tool, control is the feature — the fine-tuning that makes a model feel like the designer's own is exactly what makes the output worth paying for.