> ## Documentation Index
> Fetch the complete documentation index at: https://help.1nspect.app/llms.txt
> Use this file to discover all available pages before exploring further.

# VoiceDNA deep dive

> Upload a handful of your past reports and the AI learns your writing voice. Future AI Generate calls produce narratives that sound like you wrote them. One-time setup, 50 AI credits per extraction.

VoiceDNA is the AI feature that learns your company's writing voice from a handful of past reports and uses it on every future AI Generate call. The AI stops sounding generic and starts sounding like you wrote it. One-time setup, 50 AI credits per extraction, admins and owners only.

## Where to find it

**Narrative Management** top-nav › **Intelligence & Library Hub** › the **AI Persona (Voice DNA)** tab (third tab).

The VoiceDNA page is owner- and admin-only. Inspectors see the tab but it's locked for them.

## What's on the page

The page is a stack of separate cards.

### Include Analysis in narratives

Whether narratives include a dedicated Analysis section between Condition and Recommendation.

### Narrative format in reports

How narratives are laid out when the report is displayed. Choose **Labelled parts (Condition / Analysis / Recommendation)**, **Bullet points, no labels** or **One paragraph**. A preview shows the result. This is a display setting: it changes how narratives appear in reports, not how the AI writes them.

**Hide headline tags from clients** removes the tag from the start of each headline in the PDF and the online report, so the client sees *"Missing GFCI at kitchen counter"* instead of *"Safety Deficiency: Missing GFCI at kitchen counter"*. Only owners and admins can change it. Click **Save format** to apply it. A published report keeps the setting it was published with, so the change applies to reports you publish or republish afterwards.

### Active Voice DNA Profile

Shows your current profile (if you have one). An **Edit Voice DNA Profile** button opens the writing-preferences editor. Empty state (no profile yet) shows: *"No Voice DNA profile set. Upload reports below to extract your writing style."*

### Writing Preferences

Independent of the extracted profile; works even with no profile set. Saved through its own endpoint. Tune:

* **Tone** (how formal or informal the drafts should read)
* **Bolding usage**
* **Narrative length** (tenant-level axis the AI reads at draft time)
* **Regional terms**
* **Style notes**

The writing preferences layer on top of the extracted voice: the profile sets the voice; the preferences shape how it comes out on each draft.

### Upload area

Where you drop sample reports to extract a profile (see below).

## Extracting a profile

**1. Gather sample reports**

Pick up to 8 PDF reports your company has published. More variety (resale, pre-pour, pre-drywall, 11-month warranty) gives the AI a better read on your range.

**2. Upload the PDFs**

Drag the files into the upload zone on the AI Persona tab. The page shows a progress indicator while uploading.

**3. Click Extract Voice DNA**

The button reads **Extract Voice DNA (50 AI credits)**. On click, the AI analyzes the uploads and extracts your writing style.

**4. Wait for extraction**

The progress label reads *"Extracting… (this may take up to 2 minutes)."* The AI reads narrative structure, sentence patterns, hedging conventions, trade terminology, and the quirks of how you word findings.

**5. Profile is active**

Once extraction completes, the Active Voice DNA Profile section updates. Every future AI Generate call reads from this profile.

## Cost

**50 AI credits** per extraction. The one-time cost is justified by the depth of analysis the AI runs over up to 8 full PDF reports.

## When to re-extract

* You hired new inspectors with a markedly different writing voice and want the average
* You changed your SoP (TREC to InterNACHI, for example) and want the voice recalibrated to the new SoP's language
* The AI output started feeling off and you suspect the profile has drifted

Re-extraction replaces the current profile; the old one is not kept.

## How the profile affects AI output

For every subsequent AI Generate, AI Draft, and Polish call:

* The server reads the tenant's active VoiceDNA profile
* The profile becomes part of the system instruction for the AI call
* The returned narrative matches your voice (sentence structure, vocabulary, hedging, length preferences)

Smart Search is unaffected: it pulls narratives you wrote, so it already sounds like you.

## What VoiceDNA is not

* Not a style transfer for existing narratives (it affects new drafts only; it does not re-polish past reports)
* Not a template editor (it does not change your schema or sections)
* Not a per-inspector profile (it is tenant-wide; one profile per company)
* Not a replacement for Polish (you can still Polish any narrative to refine the AI's take)

## Related

* [AI features](https://help.1nspect.app/1nspect/features/ai-features.md)
* [AI Credits](https://help.1nspect.app/1nspect/features/it-tokens.md)
* [Intelligence & Library Hub overview](https://help.1nspect.app/1nspect/admin-operator/intelligence-hub/overview.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.