You made one video—so why is it hard to explain what it cost?
Creating a video from an idea or script has become simple. The challenge comes when you prepare the next brief. From public information alone, it is hard to confidently calculate whether you can estimate costs from video length alone or whether changing a single line triggers the full cost again.
It is especially confusing if you know Vizard’s existing “1 credit per minute of processed video” rule. That rule applies to AI clipping and transcription; it has not been confirmed as the pricing formula for Vizard Agent, which also handles generation, composition, and revisions.
Do not lock in an Agent budget based only on video length or published medians. Instead, record the initial generation cost for identical input and the revision cost incurred within the same conversation yourself.
How the “Claude Code for video” expands the input scope
Vizard Agent is a product that takes not only source video but also URLs, scripts, images, and ideas, then handles video generation, editing, repurposing, localization, and revisions through conversation. This is why founder Gary Zhang described it as applying to video what “Claude Code does for programmers.” That said, this is an analogy describing the product’s direction, not validation that the two products perform equally.
The official AI presenter guide instructs you to enter a script or key points along with tone, publishing channel, and target length, then choose an execution tier from Flash, Pro, Max, and Ultra. Once the result is ready, you can review the speech and on-screen text and revise it in the same conversation.
Tier medians are not a quote for an individual video
The official guide lists median credits aggregated across all projects by tier: 47 for Flash, 55 for Pro, 242 for Max, and 263 for Ultra. However, it does not disclose whether this is a paired-sample comparison that repeats the same projects across all four tiers, or which projects were included in each tier. It is also not a measurement limited to AI presenter work.
| Execution tier | Median credits by tier across all projects | Interpretation caveat |
|---|---|---|
| Flash | 47 | Not confirmed as a comparison of the same task |
| Pro | 55 | It is not disclosed whether the project mix matches Flash |
| Max | 242 | Not an estimated charge for one short video |
| Ultra | 263 | Sample size and mix by project type are not disclosed |
So you cannot look at the difference between 47 and 242 and calculate, “Making the same video with Max instead of Flash costs 195 more credits.” The sample size, measurement period, project mix, paired-sample status, and per-task pricing formula have not been disclosed. These numbers are best treated not as tier prices but only as a reference range to compare with your own pilot results.
Vizard’s standard pricing page shows 60 credits per month for the free plan and at least 600 credits per month for Creator and Business plans. The free plan also has limits of 720p, exports up to 10 minutes, and three days of storage. But public documentation does not confirm whether this balance and Agent’s tier-based credits are deducted from the same wallet.
Don’t carry over savings tactics from the existing clipper unchanged. AI clipping and transcription cost 1 credit per minute of processed video, and using both features on the same project is charged only once. Agent’s separate official guide presents medians aggregated across all projects by tier. Therefore, there is no basis for calculating that cutting silent sections from the source will reduce Agent costs by the same proportion.
Measure generation and revision costs with one 30-second video
The goal of your first run is not to produce a polished final cut. It is to find out how many credits one generation and one revision cost in your workflow. No installation is needed, but you will need a Vizard account and somewhere to record the test results.
- Log in with your existing Vizard account at the official Vizard Agent page. Before running anything, record the credit balance shown on screen and the available tiers.
- Set your first task as a 30-second AI presenter video. As in the official guide, specify a finished script along with the tone, publishing channel, and length. The date and amount in the text below are fictional values used to check speech-and-caption alignment; they are not actual sales terms.
- Paste the following prompt exactly and select Flash.
“Have an AI presenter read the following completed script without changing a single word. Make it a vertical social video with a professional yet friendly tone, a target length of 30 seconds, and captions included.
Script: The content below is test copy for checking the video-generation feature and is not actual sales terms. The Vizard Agent sample campaign begins on September 7, 2026. The test usage amount is KRW 12,345, and the target length of the introductory video is 30 seconds. This video is an internal review sample and will not be published.
Display ‘Vizard Agent,’ ‘September 7, 2026,’ ‘KRW 12,345,’ and ‘not actual sales terms’ exactly as written in both the speech and captions.” - Immediately after generation, check the balance difference and the four strings. Note the credits deducted for initial generation, then confirm that ‘Vizard Agent,’ ‘September 7, 2026,’ ‘KRW 12,345,’ and ‘not actual sales terms’ appear exactly in the speech and captions. Record whether the aspect ratio is vertical as well.
- Correct one incorrect string in the same conversation. For example, request: “Change ‘KRW 12,345’ in the captions to ‘KRW 54,321’ and keep all other speech, captions, and scenes.” The official guide says it retains already created assets and charges only for changed parts, but it does not publish an exact formula by change type. Record the balance difference before and after the revision separately.
- Only if Flash fails your quality bar, repeat the same initial input at a higher tier. Keep the input and review criteria identical, and record the tier, initial generation credits, revision credits, and pass/fail status for the four strings and vertical aspect ratio. Judge the cost difference between tiers using the values actually deducted.
Decide whether to adopt it by the “cost to pass per finished video”
If you evaluate Vizard Agent by counting only the videos generated, workflows with many revisions can look favorable. The real decision unit should be total credits per video that passes your publishing standard. Add the credits spent on initial generation and revisions, then exclude outputs rejected for factual or caption errors from the count of passed videos.
Four items to keep before your first purchase
- Selected execution tier
- Credits deducted for initial generation
- Additional credits deducted for revisions in the same conversation
- Pass/fail status for speech, captions, fixed text, and aspect ratio
Current public information does not establish Agent’s paid pricing, feature differences by tier, or its relationship to the standard Vizard balance. Multiply by your monthly production volume only after you have confirmed both the actual terms shown after logging in and records from identical-input pilots.
Standard Vizard credits are valid for two months when granted through a monthly plan and 13 months when granted through an annual plan; unused credits may also not carry over when you change plans. It has not been confirmed whether these rules also apply to Agent-only credits, so check the balance-handling terms again on your account screen before upgrading.
If you want to dig deeper
Vizard Agent — The First Video AGI You can check supported inputs, work scope, and the official sign-in path. agent.vizard.ai
How to make a video with an AI presenter This official guide covers the information to include in a prompt, execution tiers, all-project medians by tier, and the revision approach. agent.vizard.ai
How Credits Are Used for AI Processing? You can distinguish the 1-credit-per-minute rule and duplicate-charge policy that apply to standard Vizard AI clipping and transcription. help.vizard.ai
