Media billing
How media generation is metered in AnyBot Tokens, images per image, video per second, presenter per minute, each on its own pool, with no double-charge on a failed retry.
Media doesn't cost the same as a chat reply, so it isn't metered the same way. Text work is priced from the provider's language tokens; media is priced from the output you actually produce. AnyBot converts that into the same unit as everything else, an AnyBot Token (ABT), and debits it from the pool that matches what you made. This page covers only the media dimensions; for what an ABT is and how pools reset, start with the ABT overview.
Metered per unit of output
Media rides three of AnyBot's 6, and each pool is billed by the natural unit of its medium:
| Media | Billed | Pool |
|---|---|---|
| Images | per image | image generation (monthly) |
| Video | per second | video generation (monthly) |
| Avatar / presenter | per minute | avatar / presenter (monthly, dedicated) |
The math is deliberately simple: units × the model's rate = ABT, rounded and debited to the matching pool. Ask for four images and you're metered for four; a twelve-second clip bills twelve seconds; a ninety-second presenter bills a minute and a half. There's no per-request flat fee and no language-token accounting on media, only the output that landed.
Because each medium meters into its own pool, heavy use of one never starves another. Generating a batch of images draws down image generation and leaves your video and presenter allowances untouched.
The dedicated presenter pool
Avatar and presenter video get their own pool rather than sharing the video pool. Talking-head minutes are far heavier per unit than a plain generated clip, so keeping them separate means a run of presenter videos can't quietly drain the quota you'd rather spend on other video, and vice versa. It's the same "one pool per medium" principle, applied where the cost profiles differ most.
No double-charge on failure
Metering is winner-only: the pool is debited only when a finished file is delivered.
- A generation that fails costs nothing, no partial charge for the attempt.
- Retrying a failed generation replays your stored brief as a fresh attempt and bills only if that run succeeds. You never pay twice for one result.
- A partial batch meters only the parts that landed, four images requested, one failed to store, three billed.
This is enforced in the shared pipeline, so it holds for every media type. See how generation works for the request → progress → finalize flow that decides when the meter ticks.
What this page won't tell you
Next
- AnyBot Tokens (ABT), the unit, the six pools, and reset windows.
- How generation works, the pipeline that decides when the meter ticks.
- Plans, how media allowances scale by tier.