Poko Pulse and Flow: AI Models for Video Work

Poko Motion now has its own managed AI model choices: Poko Pulse and Poko Flow.

They are not new buttons for the sake of it. They answer a practical problem: a video agent can spend a lot of time planning, writing, editing, checking assets, and revising a composition. The model you use affects the speed of that loop and the credit cost behind it.

Pulse is the quality-first default for a full video build or more complex multi-step work. Flow is for fast iteration when you want to make changes, test directions, or move quickly through a short-turnaround task. Both live in Poko Motion, use your Poko credit balance, and appear in the model picker alongside other supported choices.

This is a managed model layer, not a claim that Poko trained a new foundation model from scratch. Poko gives you named choices, routes each choice through its configured managed runtime, meters the work, and keeps the pricing visible in your account. The point is simple: choose the video-work priority you need without setting up another provider account.

Why a video workflow needs more than one model choice

A video build is not one prompt. A typical Poko Motion project can involve a source review, a creative plan, scene creation, narration, asset checks, motion work, and revisions. Some moments reward a quality-first model. Others are about trying another direction quickly.

Poko Pulse is the default when you want to make the main build carefully. It is intended for full video builds and complex multi-step work. Use it when you are turning a website, repository, presentation, or document into a video and want the agent to carry the work through several connected decisions.

Poko Flow is the faster option. Use it when you already have a direction and want to iterate: rewrite a scene, change a visual treatment, adjust a callout, or make a quick follow-up edit.

The decision rule is short: start with Pulse for the build; use Flow when the work is mostly iteration.

What Pulse and Flow cost in Poko credits

Poko Engine pricing is published per one million tokens. That matters because a video agent can make many small decisions during a project, not just generate a single block of text.

The price of one finished video still varies. Source quality, video length, the number of revisions, narration, music, and the amount of agent work all affect the final credit use. The useful comparison is not a made-up fixed cost per video. It is whether you can see the model rate, choose the trade-off, and track the work as it happens.

How the published rates compare with premium API models

For a direct token-rate comparison, Poko Pulse's published standard rate is $0.90 per million input tokens and $2.90 per million output tokens. Anthropic lists Claude Opus 5 at $5 input and $25 output per million tokens, while Claude Fable 5 is listed at $10 input and $50 output. OpenAI lists GPT-5.6 Sol at $5 input and $30 output for standard short-context processing.

That means Pulse's published input and output rates are lower than those premium API rates. It does not prove that every task, prompt, or finished video will have the same quality or cost. Different models have different strengths, and a long, revision-heavy project will use more tokens than a short one.

Use premium models when their particular strengths matter to your work. Use Pulse when you want Poko's quality-first managed default for video builds, and use Flow when iteration speed is the priority. Pick the model for the job instead of paying the most expensive rate by habit.

Sources for the comparison: Anthropic's Claude Opus 5 announcement, Anthropic's Fable 5 model documentation, and OpenAI's API pricing page.

How the 7% Lifetime Deal founder rate works

Poko is offering eligible Lifetime Deal members a permanent 7% discount on Poko Engine usage. The discount applies to the Poko Engine models, not to every third-party provider model a user may connect.

Existing eligible LTD members receive the discount automatically. The founder offer currently says that buying a Lifetime Deal by August 27, 2026 locks in the same rate. Poko confirms eligibility at billing time, so the rate stays tied to the account rather than a coupon that can be lost.

For example, a 7% discount turns Pulse's published rates from $0.90 to about $0.837 per million input tokens, $0.30 to about $0.279 cached input tokens, and $2.90 to about $2.697 output tokens. The exact amount charged is calculated from actual usage.

The takeaway: if you already have an eligible LTD, the discount is automatic; if you are considering one, confirm the current offer before the stated deadline.

Where to find Poko Engine in Poko Motion

Open the model picker in the Poko Motion composer. Under Poko Engine, you will see Poko Pulse, Poko Pulse Priority, Poko Flow, and Poko Flow Priority.

Start with Pulse unless you have a specific reason to optimize for faster iteration. If the project is urgent, choose a Priority option. You can also open the Poko pricing page from the app to see the current published rates.

The launch goal is not to make model selection more complicated. It is to make the cost and speed trade-off visible when you need it—and out of the way when you do not.

A practical first project to try

Use Poko Pulse on one source-driven project this week. Pick a real asset you already have: a landing page, product update, sales deck, or repository. Ask Poko Motion for a short product video, review the first cut, then use Flow for one or two fast creative changes.

That workflow gives you a useful comparison inside your own product and creative process. You will see the output, the revision speed, and the credit use in the same place.

Choose the model deliberately. Then make the next version.