Skip to content
AtheronLABS

You're visiting from the United States. Prices are shown in US dollars. Not right?

labs@atheron:~/industries/marketing/ai$ train agent --grounded

Brand voice models

A model fine-tuned on your own approved copy.

General models write fluent copy that sounds like everyone. A model fine-tuned on your own approved copy writes like you, which is why a custom model is the right route here. Brand managers approve every piece, and their edits teach the next version.

// what matters here

What is different in this industry.

01

Trained on your best work

The model learns from copy your brand has already approved, plus your style guide and product data.

02

Brand managers approve

Nothing is published without a person. Edits and rejections are recorded and become the next round of training data.

03

Unreleased campaigns stay private

The model runs on GPUs you control, so product launches and campaign plans never go to an outside AI provider.

04

Two languages as originals

English and French versions are each written in the brand's voice, not translated word for word.

// example projects

Priced examples for this service.

Each one is hypothetical, labelled as such, and priced live from our rate card. Open any of them in the estimator and make it yours.

Example projectWeb appAICustom LLM

Brand voice model

Hypothetical. Not a client, not a result.

// The problem

A brand team uses general AI tools for product copy, emails and social posts, and every piece is rewritten because it sounds nothing like the brand.

// What we would build

A custom model fine-tuned on the brand's own approved copy, style guide and product data, so first versions read like the brand, with a review queue where brand managers approve, edit or reject and their edits teach the next version. It runs on GPUs the client controls, so unreleased campaigns stay private.

// What is in it

  • A model fine-tuned on your approved copy, guide and product data
  • Product descriptions, emails and social posts in your voice
  • English and French versions written as originals, not translated word for word
  • A review queue for brand managers, with edits recorded
  • Scored against held-out approved copy before every release

// Stack

  • Fine-tuned open-weight model
  • vLLM
  • Python
  • Next.js
  • DigitalOcean GPU

// estimate

Build
≈ US$138,000 to US$210,000, delivered within 31 weeksCAD 196,100 to 299,200
Hosting
≈ US$4,360 a monthCAD 6,210 a month
Support
≈ US$2,500 a monthCAD 3,565 a month

Prices in your currency are estimates from today's Bank of Canada rate. All invoicing is in CAD or USD.

AI route
Custom LLM
Model running cost
≈ US$4,190 a monthCAD 5,960 a month

Timeline by milestone

Discovery
2.9 to 3.4 weeks
Data and evaluation plan
0.4 weeks
Design approved
2.4 to 3.9 weeks
Core features
1.9 to 2.9 weeks
Full build
1.4 to 1.9 weeks
Training
6.4 to 11.4 weeks
Testing and fixes
0.4 weeks
Evaluation and red-teaming
3.9 to 4.4 weeks
Launch
0.9 to 1.4 weeks
Production
0.9 to 1.4 weeks

How it is paid

Deposit 20%
CAD 38,980 to 59,600
Discovery 1.5%
CAD 2,923.50 to 4,470.00
Data and evaluation plan 7.9%
CAD 15,397.10 to 23,542.00
Design approved 1.5%
CAD 2,923.50 to 4,470.00
Core features 4.6%
CAD 8,965.40 to 13,708.00
Full build 3.1%
CAD 6,041.90 to 9,238.00
Training 23.9%
CAD 46,581.10 to 71,222.00
Testing and fixes 1.5%
CAD 2,923.50 to 4,470.00
Evaluation and red-teaming 15.9%
CAD 30,989.10 to 47,382.00
Launch 1.5%
CAD 2,923.50 to 4,470.00
Production 8.6%
CAD 16,761.40 to 25,628.00
Holdback, 30 days after launch (10%)
CAD 19,490 to 29,800
GPU time for training and testing, at cost
CAD 1,200. Billed up front, at cost, outside the milestones

// questions

Questions we get about this.

How much approved copy does it need?

Enough to show the range of what you write: product pages, emails, posts. We review what you have and tell you whether it is enough before we train.

Why fine-tune rather than prompt a frontier model?

Prompts and examples get a frontier model part of the way and are cheaper to start. A fine-tuned model holds the voice more consistently at volume and keeps unreleased work on your own GPUs; we price both.

Who owns a fine-tuned brand voice model?

You do, once the work is paid for, including the trained weights.

// next

Not quite your project?

Tell us what you have in mind. We will come back to you with a range and the questions that would narrow it. Or book a call and talk it through.

Custom AI brand voice models for marketing teams | Atheron Network Labs