AI Usage
AI generation in AI Boost Post is billed in USD against your AI balance. Distributing content to channels is always free — your balance is only charged when you generate text, images, or videos in the Generation Studio. Manage your balance at Credits in the top navigation.
Screenshot pending
/help/credits/open_credits_menu.png
Workspace sidebar with the Credits menu item highlighted alongside the balance badge visible in the top-right nav
See docs/testing/help-screenshot-manifest.md for capture instructions.
Screenshot pending
/help/credits/credits_page_overview.png
AI Usage page showing balance card at top, top-up form on the right, and transaction history table below
See docs/testing/help-screenshot-manifest.md for capture instructions.
The balance badge
Your current AI balance is always visible in the top-right corner of the navigation bar. The badge changes colour based on your balance health:
- Green — healthy balance.
- Yellow — low balance (less than 50% of your average monthly spend).
- Red — negative balance. All generation is blocked until you top up.
Click the badge to go directly to the AI Usage page.
Screenshot pending
/help/credits/credit_badge_states.png
Top navigation bar close-up showing the balance badge in three colour variants side-by-side: green (healthy), yellow (low), red (negative)
See docs/testing/help-screenshot-manifest.md for capture instructions.
How usage is charged
Before any generation starts, a pre-flight check confirms your balance is not negative. If it is negative, the request is rejected immediately with no provider call made. If your balance is zero or positive, the generation proceeds — after it completes, the actual USD cost (based on real provider usage, not the estimate) is deducted and logged in your transaction history.
A single generation can push your balance below zero. That is allowed. The next generation will be blocked until you top up.
Every USD amount you see is the final charged cost, not the raw provider price. AI Boost Post applies a per-model markup on top of the underlying provider cost (typically 1.5×–4× depending on the model tier — fast models like Haiku and Flux Schnell carry the highest multiplier because their per-call cost is tiny). The numbers in the cost reference table below already include this markup.
Cost reference table
Numbers are approximate — actual charges depend on exact token counts, image sizes, video durations, and provider pricing at the time of generation.
| Operation | Approx cost |
|---|---|
| Text — Claude Haiku 4.5 (1K in + 1K out tokens) | ~$0.024 |
| Text — Claude Sonnet 4.5 (1K in + 1K out tokens) | ~$0.054 |
| Text — Claude Opus 4.7 (1K in + 1K out tokens) | ~$0.135 |
| Image — Flux Schnell (1024×1024) | ~$0.01 |
| Image — Flux Pro 1.1 (1024×1024) | ~$0.12 |
| Image — DALL-E 3 (1024×1024) | ~$0.12 |
| Video — Kling v1.6 (5 s) | ~$0.60 |
| Video — Pika 2.0 (5 s) | ~$0.90 |
| Video — Luma Dream Machine (5 s) | ~$1.20 |
| Video — Runway Gen-3α Turbo (5 s) | ~$1.50 |
| Video — Runway Gen-3α (5 s) | ~$3.75 |
| Distribute to any channel | Free |
Requesting a top-up
Go to AI Usage
Fill in the top-up form
Wait for admin approval
Screenshot pending
/help/credits/topup_request_form.png
Top-up request form with USD amount input, payment reference field, submit button, and a list of prior pending requests below
See docs/testing/help-screenshot-manifest.md for capture instructions.
Reading the transaction history
The transaction history on the AI Usage page shows every balance movement with a timestamp, entry type, amount, and running balance. Use the entry type filter to narrow the view:
- Top-up — amount added by a top-up approval.
- Consumption — amount deducted after a generation completes.
- Refund — amount returned if a generation fails after being charged.
- Adjustment — manual grant or deduction by a Ubalitics admin.
Screenshot pending
/help/credits/credit_ledger_view.png
Transaction history table with columns for timestamp, entry type badge, description, amount, and running balance; entry-type filter dropdown visible above the table
See docs/testing/help-screenshot-manifest.md for capture instructions.
When credits are NOT deducted
The platform is careful not to charge you for AI work that didn't actually happen or couldn't be used. You will NOT see a consumption ledger entry in any of these situations:
- Automation skipped generation because no channel supports the media type. If you have Video: Auto but every selected channel is text/image-only (LinkedIn, Dev.to, Medium, etc.), the generation is skipped entirely — no provider call, no charge. Same for Image: Auto with only text-body platforms (Medium, WriteFreely).
- Generation failed before completion. If a provider fails a generation (Replicate / Runway / OpenAI / Anthropic returns an error), any credits deducted for that attempt are automatically refunded as a refund ledger entry. For composite jobs (text + image + video), each sub-job is charged and refunded independently.
- Insufficient balance skipped an Automation media step. If your balance is too low to fund an image or video generation, the AM fire logs the skip, notifies the workspace owner once every 24 hours, and completes text-only. No partial charge is applied.
- Cache hit on per-channel adaptation. If Claude has already adapted this exact source text for this exact channel recently, the cached result is reused and no new Claude call runs — no charge for the cache hit.
- Sandbox mode. API calls made with a
pk_test_*key exercise the full pipeline with mock outputs and no real provider calls. No credits are deducted for sandbox traffic.
All of the above show up in the promotion detail page's amber "Media generation note" banner (except cache hits, which are silent — they're the happy path).