Overview
In most cases, a generation is every time a customer interacts with an AI model to generate something. Knowing how many generations you've used is important as most subscriptions include a set number of generations.
Below are specific details pertaining to:
Text Generation
Image Generation
Video Generation
Feed Variation
Text Generation
Each message sent to a text model counts as 1 generation. This includes both the user’s prompt and the AI’s response, treated as a single interaction.
Model Name | Cost per use |
Gemini 2.0 Flash | 1 Generation / Message |
Gemini 2.5 Flash | 1 Generation / Message |
Google Gemini 2.5 Pro | 1 Generation / Message |
Gemini 3 Flash instant | 1 Generation / Message |
Gemini 3 Flash thinking | 1 Generation / Message |
Gemini 3 Pro Instant | 1 Generation / Message |
Gemini 3 Pro Thinking | 1 Generation / Message |
Gemini 3.1 Pro Preview Thinking | 1 Generation / Message |
Gemini 3.1 Pro Preview | 1 Generation / Message |
Gemini 3.5 Flash | 1 Generation / Message |
Google Gemini 3.1 Flash Lite Preview | 1 Generation / Message |
Google Gemini 3.1 Flash Lite Preview Thinking | 1 Generation / Message |
GPT 5.1 -Thinking | 1 Generation / Message |
GPT 5.1 – High-Thinking | 1 Generation / Message |
GPT 5.1 – Minimal-Thinking | 1 Generation / Message |
OpenAI GPT-5.1 instant | 1 Generation / Message |
OpenAI GPT-5.2 Instant | 1 Generation / Message |
OpenAI GPT-5.2 Thinking | 1 Generation / Message |
OpenAI GPT-4o | 1 Generation / Message |
OpenAI GPT-4.1 | 1 Generation / Message |
OpenAI GPT-5.6 Sol | 1 Generation / Message |
OpenAI GPT-5.6 Terra | 1 Generation / Message |
OpenAI GPT-5.6 Luna | 1 Generation / Message |
OpenAI GPT-5.5 | 1 Generation / Message |
OpenAI GPT-5.4 | 1 Generation / Message |
OpenAI GPT-5.4 Mini | 1 Generation / Message |
OpenAI GPT-5.4 Nano | 1 Generation / Message |
OpenAI GPT-5 | 1 Generation / Message |
OpenAI GPT-5 Mini | 1 Generation / Message |
OpenAI GPT-5 Nano | 1 Generation / Message |
OpenAI o3 | 1 Generation / Message |
OpenAI o3 mini | 1 Generation / Message |
OpenAI o4 mini | 1 Generation / Message |
Claude Fable 5 | 1 Generation / Message |
Claude Opus 4.7 | 1 Generation / Message |
Claude Sonnet 4.6 | 1 Generation / Message |
Claude Sonnet 5 | 1 Generation / Message |
Claude Haiku 4.5 | 1 Generation / Message |
Whether your message is short or long, each prompt-response pair is billed once.
Image Generation
For image models, every image generated counts as 1 generation, regardless of resolution, quality, or model type. This ensures consistency across providers.
Model Name | Cost per use |
Bria v2.3 | 1 Generation / Image |
Adobe Firefly v4 | 1 Generation / Image |
ChatGPT Image 1.5 | 1 Generation / Image |
ChatGPT Image 2 | 1 Generation / Image |
Google Imagen 4 Ultra | 1 Generation / Image |
Google Imagen 3 | 1 Generation / Image |
Nano Banana 2 Lite: Gemini 3.1 Flash Lite Image | 1 Generation / Image |
Nano Banana: Gemini 2.5 Flash Image | 1 Generation / Image |
Nano Banana Pro - (Gemini Image 3) | 1 Generation / Image |
Nano Banana 2 (Gemini 3.1 Flash Image) | 1 Generation / Image |
Kling AI IMAGE 2 | 1 Generation / Image |
Kling AI IMAGE 2.1 | 1 Generation / Image |
Kling AI IMAGE 3 | 1 Generation / Image |
Kling AI IMAGE 3 OMNI | 1 Generation / Image |
Note: Post generation actions will have separate cost, e.g. 1 edit = 1 generation
Video Generation
Video generation is priced per second of generated video, but with an important rule:
In cases where the cost is a fraction, the generations are rounded up to the nearest full generation.
There are no partial charges—even 0.1 over will round up.
Model Name | Cost per use |
Runway Gen3 Alpha Turbo | 0.33 Generations / sec (Rounded off to nearest higher based on total seconds duration) |
Runway Gen4 Alpha Turbo | 0.33 Generations / sec (Rounded off to nearest higher based on total seconds duration) |
Google Veo2 | 3 Generations / sec |
Google Veo3 | 2 Generations / sec |
Google Veo3.1 | 2 Generations / sec |
Gemini Omni Flash* | 2 Generations / sec |
Adobe Video1 Standard | 4 Generations / sec |
Sora 2 | 1 Generations / sec |
Sora 2 Pro | 3 Generations / sec |
Kling AI Video 2.1 | 2 Generations / sec |
Kling AI Video 2.5 | 2 Generations / sec |
Kling v3 | 2 Generations / sec |
Kling v3 Omni | 2 Generations / sec |
Examples:
A 5-second Runway Gen 4 output = 5 × 0.33 = 1.65 → 2 generations
A 3-second Google Veo3 output = 3 × 2 = 6 → 6 generations
Note: You will not be charged for failed generations
Note: Kling models must be enabled by a Workspace Admin before use. See: How can I use Kling 3 and Kling 3 Omni in Pencil?
*Gemini Omni Flash is in preview from Google, so quality and behaviour may change. It is enabled automatically in any workspace that already has a Google video model enabled; otherwise an admin enables it under AI Governance, then Settings, then Video Generation.
Feed Variations
We count one generation per every variation generated. If 20 variations are generated from a master creative, that would count as 20 generations.
Note: If 1 creative has 5 formats, we still only count as 1 generation.
AI Tools inside feeds
Using an AI tool inside the feed also counts as one generation. For example, translating 5 cells to a different language using AI will count as 5 generations.
Feed Sync
We count one generation per ten variations generated using Feed Sync. For example, if 20 variations are generated using the feed sync feature, that would count as 2 generations.
Adobe After Effects Imports
You can import After Effects projects to Pencil and link them to Feed Variations to scale video content. Imports are priced per duration of the rendered output video.
Software format for import | Cost per use |
Adobe After Effects | 1 Generation / 10 sec duration |
For more information on preparing and uploading After Effects projects and connecting them to your Pencil Feed Variations, read: What Are The After Effects Upload Requirements?
AI Voiceovers
With any of these models, one generated voiceover counts as a single generation
Model name | Cost per use |
ElevenLabs v2.5 Flash | 1 Generation / audio |
ElevenLabs v2 Multilingual | 1 Generation / audio |
ElevenLabs v3 | 1 Generation / audio |
Google Chirp 3 HD | 1 Generation / audio |
Google Gemini 2.5 TTS Pro | 1 Generation / audio |
Scores
Custom Scores consume generations. A Custom Score set to Automatic runs when a work item that meets its criteria is saved, and each run counts as one generation. Default Pencil scores do not consume generations.
For example, if a generated image meets the criteria of three Custom Scores set to Automatic, it typically consumes four generations: one for the image itself and one for each Custom Score.
You can set each Custom Score’s Score generation method to Automatic or Manual when creating or editing it. A Manual score runs only when you request it.
Workspace Admins can choose which scores are activated in a workspace. See: How Can I Set Up Custom Scores in Pencil
Scoring Agents
Agents in Pencil can also call additional models or external tools as part of their work. Scoring agents like the Share of Model (SoM) Agent is an example. These agents may cost more than a single generation.
Agent / Score System | Cost per use |
GWI Insight Agent (or any other consumer insight agents powered by the GWI plugin) | 4 Generation / request |
Share of Model score | 2 Generations / request* |
The Share of Model Agent is the easiest way to run Share of Model analyses in Pencil. Share of Model is also available as a standalone product, with consumption that varies by module and model. Read more abut Share of Model here.
Tracking Generations
If your subscription has a limited generation count, you can view your remaining generations in the bottom right corner of your workspaces using the counter shown below:
You can also view generations used per user in any workspace's Manage Team menu:
