> ## Documentation Index
> Fetch the complete documentation index at: https://support.humanizing.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Estimate GO Consumption with the Plural Calculator

> Use Plural's built-in GO Consumption Calculator to estimate how many GOs a typical user journey through your flow will spend before you go live.

Before you publish an Avatar, it helps to know roughly how many GOs a typical visitor interaction will consume. Plural includes a built-in **GO Consumption Calculator** that analyses a selected path through your flow and gives you an estimated GO cost per session. This page explains how to use the calculator and what its results mean.

## What the Calculator Does

The calculator inspects the flow elements you select and totals up the expected GO cost based on the [consumption rates for each feature](/en/account/gos-and-billing). It is designed to help you estimate the cost of a realistic "happy path" — the sequence of steps most visitors are likely to follow — so you can size your GO plan appropriately before going live.

<Warning>
  The calculator produces a **rough estimate only**. Real consumption varies depending on how individual visitors interact with your Avatar, how long they speak, and which branches of your flow they take. Never rely on the estimate as a guaranteed per-session cost.
</Warning>

***

## How to Use the Calculator

<Steps>
  <Step title="Open your project">
    Navigate to the project you want to estimate inside Plural's flow builder.
  </Step>

  <Step title="Identify your happy path">
    Decide on the most likely sequence of steps a visitor will take through your flow. This is often called the "happy path" — the route from start to finish with no errors or unexpected detours.
  </Step>

  <Step title="Select the relevant elements">
    Hold down **CTRL** on your keyboard and click each element on the happy path one by one. Selected elements are highlighted. You can select as many elements as you need.
  </Step>

  <Step title="Open the estimation tool">
    Right-click on any of the highlighted elements. In the context menu that appears, select **Estimate consumption**.
  </Step>

  <Step title="Review the results">
    Plural displays a breakdown of the estimated GO cost for the selected elements, grouped by feature type (elements, TTS, STT, Knowledge Base, etc.).
  </Step>

  <Step title="Read the important notes">
    Click **Important Notes** within the results panel to see the full list of limitations and exclusions that apply to the estimate.
  </Step>
</Steps>

***

## Calculator Limitations

The calculator cannot account for every runtime variable. Before acting on the estimate, review the following known limitations:

<AccordionGroup>
  <Accordion title="Limited TTS estimation">
    Only the **first TTS text block**, variation, condition, or thank-you speech in each element is included in the estimate. If an element contains multiple TTS variations and Plural randomly selects a longer one at runtime, the actual cost will be higher.
  </Accordion>

  <Accordion title="Dynamic content is excluded">
    The calculator cannot estimate GOs for content that is generated or fetched at runtime, including:

    * Responses from **Plural Dialogs** or **Chatbots**
    * Text generated from **API responses**, **Plugins**, or **Custom Scripts**
    * Attributes and variables such as `#ATTRI/greeting` used inside Speech elements

    If your flow uses these features, add a manual estimate on top of the calculator result.
  </Accordion>

  <Accordion title="STT duration is unpredictable">
    The calculator assumes a single 10-second STT block per listening element. In practice, a visitor may speak (or pause) for longer, causing STT to run for multiple 10-second windows. Garrulous visitors, background noise, or long pauses all increase actual STT consumption above the estimate.
  </Accordion>

  <Accordion title="Animations are ignored">
    Animation tokens such as `#ANI/wave_1` that appear inside Speech elements are excluded from the character count and do not contribute to the TTS estimate.
  </Accordion>

  <Accordion title="No-interaction TTS blocks are excluded">
    TTS blocks that only play when **no interaction** has occurred (idle/attract-loop speeches) are not included in the estimate, as they fire independently of a visitor session.
  </Accordion>

  <Accordion title="Knowledge Base minimum cost">
    Even the shortest possible KB request consumes a minimum of:

    * **10 GOs for TTS** (at least 1 character spoken by the Avatar in the response)
    * **10–100 GOs for the KB response** itself, depending on whether you use the Standard (GPT-3.5) or Premium (GPT-4o) model

    The calculator uses these minimums. If the actual KB response is longer, real consumption will be higher.
  </Accordion>
</AccordionGroup>

***

## Tips for a Better Estimate

<Tip>
  Run the calculator on two or three different paths — the happy path, a longer detour, and a worst-case scenario — and average the results to get a more realistic cost band for your project.
</Tip>

* **Minimise TTS character counts** by keeping Avatar speeches short. Fewer characters spoken = fewer GOs spent.
* **Disable STT** on elements where visitors are only expected to tap buttons, not speak.
* **Use Standard (GPT-3.5) Knowledge Base** for cost-sensitive deployments; switch to Premium only where response quality justifies the 10× cost increase.
* **Test in Preview mode** freely — no GOs are consumed during preview, so you can run through paths as many times as needed before going live.

***

## Related Resources

<CardGroup cols={2}>
  <Card title="GOs & Billing" icon="coins" href="/en/account/gos-and-billing">
    Full consumption rates, plan examples, and what happens when GOs run out.
  </Card>

  <Card title="Preview vs. Live Mode" icon="play" href="/en/account/gos-and-billing#preview-mode-vs-live-mode">
    Understand the difference before publishing your Avatar.
  </Card>
</CardGroup>