The mistake in budgeting for Pencil Scribble is treating a subscription price as the cost of a finished brief. Our verdict, based on documented capabilities: confirm access, estimate the entire production job, and obtain the allowance and overage terms in writing before approving spend. A cheap entry plan and an “unlimited” headline answer neither the eligibility question nor the final invoice question.
By AI Marketing Picks
Updated October 6, 2026. All prices and features below are as of October 2026, checked October 6. Vendor illustrations are labelled; our budget calculations use explicit assumptions.
Pencil plans and Scribble access: what is available?
| Plan | Monthly billing | Annual billing, monthly equivalent | Advertised allowance | Scribble access |
|---|---|---|---|---|
| Core | $14/month | $11/month | 50 generations/month | Current Scribble documentation does not identify Core as eligible |
| Growth | $55/month | $44/month | 250 generations/month | Current Scribble documentation does not identify Growth as eligible |
| Pro | Custom quote | Custom quote | Advertised unlimited generations; contractual allowance applies | Pro-only open beta |
As of October 6, 2026, Pencil lists Core at $14 monthly or $11 per month billed annually, and Growth at $55 monthly or $44 billed annually. These are subscription prices. Pencil’s pricing page and Scribble access documentation identify Scribble as a Pro-only open beta with custom pricing. Confirm eligibility before buying.
The plan cards advertise AI agents more broadly. That wording is insufficient evidence that a specific named beta is included. If Scribble is the reason you are buying, ask Pencil to confirm the subscription and workspaces that qualify before committing to Core or Growth.
As of October 2026, a Workspace Admin enables Scribble under Settings → Beta features, separately in each workspace. The help article says activation cannot be undone. Put that decision into your rollout approval, alongside the commercial terms. Source: Pencil’s Scribble setup guide.
What changed in September—and what remains beta?
Two dated vendor updates matter:
- September 23, 2026: Pencil updated granular model rates and introduced seven agentic text buckets. Source: generation accounting.
- September 28, 2026: Pencil announced Scribble’s public launch, describing an assistant that routes briefs through agents, models and skills, with marketer review. Source: Scribble launch announcement.
Public launch and continuing open-beta status can coexist. For purchasing, the current access documentation is the useful eligibility check.
The announcement also promises additional functionality linking creative through to media buying towards the end of 2026. As of October 2026, treat that statement as roadmap information. Do not include its expected savings in a budget that depends on functionality already available. Source: Pencil’s launch announcement.
The budgeting consequence is straightforward: an old estimate based on one flat charge per prompt needs rebuilding around the work requested.
Agent charges: seven buckets, Chat models, Auto and Deep
As of October 2026, Scribble draws from your existing Pencil generation allowance, with no separate Scribble fee or add-on. A turn can involve several model calls. Its bucket reflects the higher tier indicated by call count or tokens processed; the selected Chat model sets the agent rate, including planning and reasoning. Source: Scribble pricing documentation.
Representative published rates, as of October 2026, in generations per unit of agent work:
| Bucket | GPT-5.5 | GPT-6 Sol |
|---|---|---|
| XS | 0.2 | 0.1 |
| S | 3.3 | 1.3 |
| M | 6.6 | 2.7 |
| L | 16.2 | 6.4 |
| XL | 29.6 | 11.7 |
| XXL | 64.6 | 25.5 |
| XXXL | 139.8 | 55.2 |
Pencil rounds fractional charges up to one decimal place. Source: Pencil’s model and bucket schedule.
Choosing a named model establishes a rate schedule; it does not establish the final bucket. A job with few calls can still enter a larger bucket because of its token volume. The cheap-looking XS price is therefore a poor basis for approving a campaign brief.
As of October 2026, Auto is the default Chat setting and lets Pencil select a model. Adaptive is the default effort mode; Deep forces high effort and costs more on average. Pencil publishes no fixed Deep multiplier in this guide. Image, video and audio models are selected separately. Source: Scribble pricing.
For an initial forecast, we favor a named Chat model and Adaptive. The tradeoff is giving up automatic Chat model selection for a known schedule. Reserve Deep for jobs where your team can justify the additional effort.
After completion, the turn’s cost menu shows its bucket, charge and models. Ask for those records during an evaluation, then estimate from representative briefs rather than a generic “messages per month” assumption. Source: Pencil’s usage-tracking guidance.
Complete brief costs: adaptations, images, video takes and scores
Our budget worksheet uses this structure:
Brief consumption = agent work + generated media + applicable scoring + separately charged processing.
Keep each term visible. Otherwise, a single “cost per asset” number can hide discarded takes, intermediate images and repeat score runs.
| Component | Published treatment as of October 2026 | What to put in your estimate |
|---|---|---|
| Agent work | Chat model and completed-work bucket | All production and revision turns |
| Existing-master adaptations, resizes or cutdowns | Agent charge; no new-media charge | Agent work even when reusing assets |
| New images | Model-specific per-image rate; Nano Banana Pro: 2.4 generations/image | Generated candidates, not just selected outputs |
| Generated video | Veo 3.1: 4 generations/second; Kling AI VIDEO 2.5 Turbo: 0.7/second | Generated seconds across takes |
| Default Pencil scores | 0 generations | No additional generation allowance |
| Custom Scores | 1 generation/run, Automatic or Manual | Qualifying saves or requested runs |
| GWI Insight Agent / Share of Model score | 4/request / 2/request | Relevant requests separately |
| Audio or other processing | Confirm applicable model/unit charges | Any voiceover, enhancement or other billable work |
Sources: Scribble’s agent/media treatment, generation accounting, Custom Score documentation.
Automatic Custom Scores run when an eligible work item is saved. Manual gives you control over when a score runs; it still costs one generation. Source: Pencil’s score-trigger explanation.
For example, our calculation for one Nano Banana Pro image with three Custom Scores, each run once, is 2.4 + 3 = 5.4 generations, before agent work. That is a defined scenario, not a universal image price. Ask how many saves and score runs your production process creates.
Pencil says failed generations are not charged; successfully generated candidates belong in the estimate even if your team rejects them. Source: Scribble pricing.
Pencil’s worked video example
The vendor’s film illustration uses two three-second scenes and two takes per scene:
2 scenes × 3 seconds × 2 takes = 12 generated seconds.
With GPT-5.5 L-bucket agent work:
- Veo 3.1: 16.2 + (12 × 4) = 64.2 generations.
- Kling AI VIDEO 2.5 Turbo: 16.2 + (12 × 0.7) = 24.6 generations.
These are vendor illustrations at the published standard rates as of October 2026. Source: Pencil’s example brief costs.
As of October 6, 2026, Pencil’s generation-accounting example budgets 12 generated seconds for a six-second film, reflecting two takes per scene. With GPT-5.5 agent work, it totals 64.2 generations using Veo 3.1 or 24.6 using Kling AI VIDEO 2.5 Turbo. These vendor illustrations show why finished duration alone cannot predict consumption.
Our monthly budget sensitivity calculation
Suppose you commission ten repetitions of that exact brief, holding agent bucket and take count constant:
| Planning scenario | Veo illustration | Kling illustration |
|---|---|---|
| Ten briefs | 642 generations | 246 generations |
| Hypothetical period allowance | 500 generations | 500 generations |
| Excess above that allowance | 142 generations | 0 generations |
| Remaining allowance | 0 generations | 254 generations |
The difference is 396 generations, or 61.7% lower consumption in the Kling scenario. This is our arithmetic using vendor inputs; the 500-generation allowance is hypothetical. It establishes no equivalence in creative quality.
A further three-second take adds 12 video generations with Veo or 2.1 with Kling, before additional agent work or scoring. That is the sensitivity your approver needs: another take has a different budget effect from another finished film.
Before approving a pilot, specify the output count, scene lengths, maximum takes and score runs in the brief. Review the proposed plan, then record completed-turn charges and generated media separately. Include one routine brief and one revision-heavy brief in the evaluation: a forecast built only around an easy first pass leaves the approval budget exposed. Ask Pencil to confirm the applicable resolution tier and any optional capacity charges.
Do not multiply these figures by Core or Growth’s subscription price divided by their allowance. Those plans do not establish your Pro dollar rate. Translate consumption into money using the actual Order Form.
For broader planning, put this worksheet alongside your 2027 AI marketing budget.
Spending caps: allocations, user limits and overlapping roles
As of October 2026, subscription admins manage controls under Settings → Credit limits:
- Workspace or workspace-group allocations are shared.
- Member, user-group or role limits count each person’s own usage.
- Personal limits and workspace allocations apply together.
Members without their own limit inherit a group or role limit. For overlapping limits, the most recent change wins; Pencil gives the example of a role limit set after a user-group limit. A member’s “all workspaces” scope includes workspaces created later. Source: Pencil’s credit-limit documentation.
Do not assume the strictest overlapping limit wins. In a hypothetical policy, setting a 100-credit group limit and then a 250-credit role limit can select the later 250-credit limit. Review the sequence of changes whenever you adjust team permissions.
A warning is a notification; blocking is the enforcement setting. Configuring an early warning does not by itself create a stop rule.
As of October 6, 2026, Pencil documents two layers of credit controls: workspace allocations and personal limits. Blocking defaults on; users at the limit cannot start credit-consuming actions, including chats. Admins can disable blocking and configure Inbox warnings. Pencil’s credit-limit documentation does not explain whether already-running jobs stop; clarify this before assuming an absolute spending ceiling.
For budget approval, request written answers about running jobs, queued work and simultaneous requests near the limit. Establish who can disable blocking or raise allocations. A control that stops new work is useful; an absolute dollar ceiling also requires agreement about charges already being incurred.
Pro quote checklist: what must be in the Order Form?
As of October 2026, Pro’s advertised unlimited generations coexist with contractual consumption terms. Sections 4.4.1–4.4.5 specify an allowance for a defined period, model-dependent credit usage, and excess-credit pricing through the Order Form. Excess charges go on the next invoice. Source: Pencil Pro terms.
Before approval, obtain:
- Access and scope: written Scribble eligibility for the intended workspaces, plus any beta restrictions that affect the rollout.
- Committed consumption: base fees, included credits, period length, and whether unused allowance expires or carries forward.
- Complete rates: agent/model schedule, media charges, scoring and optional processing. Ask how public “generations” reconcile with contractual Generation Credits.
- Overages: the applicable excess-credit rate, invoice timing, and whether additional spend requires approval.
- Billing: payment currency, advance commitments, taxes and any services outside the consumption allowance.
- Renewal: initial and renewal term lengths, notice deadlines, price changes and how to exit.
The published terms provide advance payment, a 30-day invoice deadline unless otherwise agreed, and fees excluding taxes. They provide 90 days’ notice for renewal fee increases and automatic renewal unless notice is given at least 60 days before term end. Check your Order Form for controlling provisions. Source: Pro payment and renewal terms.
Our planning formula is:
Period cost = agreed base fees + max(0, period consumption − included credits) × applicable excess-credit rate + separately agreed charges and taxes.
Use it only after confirming the unit mapping and rate schedule. In the hypothetical Veo case above, excess consumption is 142 credits if the contract confirms that mapping; the excess charge is 142 × your agreed rate, not a publicly established Pencil price.
Buying verdict: when is Scribble worth custom pricing?
Consider Pro for repeated adaptation work. Based on documented capabilities, Scribble can coordinate resizes, market adaptations and video cutdowns from existing masters. That makes orchestration a plausible purchase when coordination consumes substantial team time. The tradeoff is custom commercial terms and beta adoption; your team must establish the savings through its own evaluation. Source: Scribble use cases.
For original video, approve a scoped evaluation before a recurring commitment. Define scenes, take counts, models and scoring upfront. Lower-consumption models improve the forecast, but the tradeoff is that suitability and output quality still need validation against your brand brief.
For a small team seeking predictable spend, postpone Scribble if access or excess terms remain unclear. You give up integrated brief orchestration, but avoid committing to a budget you cannot explain. Core and Growth can be evaluated for their own offering; their headline prices do not settle a Scribble decision.
The approval standard should be simple: confirmed access, a complete brief estimate, and written spending terms. If one is missing, the quote is not ready for approval.
Request Pencil Pro pricing
These links go to vendor product or pricing pages. Check the vendor’s current terms before purchasing.
Request a Pro quote and confirm Scribble access. Bring your representative brief, expected generation consumption and spending-control questions.
Freshness note: Eligibility, beta status, generation rates, credit controls and Pro consumption terms may have changed since publication, and the announced creative-to-media-buying expansion had not shipped when sources were checked. See Pencil’s release notes and the launch announcement for current details.