How to make a cooking tutorial short without filming

Last updated 2026-08-16 · 5 min read

What is the short version?

Recipe shorts are simple to watch and awkward to make: you need a kitchen that looks good, hands that stay consistent, and a way to show four steps in thirty seconds. Generating them removes the kitchen problem entirely, and creates a new one — keeping the cook the same person from shot to shot.

1. Go to Vizard Agent and describe the dish and its steps.
2. Say who is cooking and what the register is.
3. Say the length, the language, and the feed it is going to.

What do you need before you start?

The recipe as steps, and a decision about who is on screen. Vizard Agent generates the cook rather than filming one, so what it needs from you is enough character to build a reference from — a home cook in their own kitchen is a different video from a chef in a studio.

• The dish and its steps, in order. Four steps is about right for thirty seconds. Six is too many.
• Who is cooking. Age, setting, register. "A home cook in their own kitchen" is enough to start.
• The ingredients that must be visible. Vizard Agent builds a reference sheet from these, so name the ones that matter.
• The language. The narration is generated, so any language works, and it should match your audience rather than your prompt.
• The length and shape. Vertical and thirty seconds for a feed.

What do you type into Vizard Agent?

Describe the video you want rather than the technique. Vizard Agent works out the shot list, the reference sheet and the voice from a plain description, and the details it genuinely needs from you are the dish, the person and the length.

Prompt

Variants worth knowing:

• Your own kitchen footage. Upload it and Vizard Agent uses your material for the steps and generates only what is missing.
• Your voice. Upload a voice sample and the narration is yours rather than generated.
• No narration. "Text on screen only, no voice." Common for feeds watched muted.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real recipe short. Two things stand out: it checked what the tools cost before planning, and it built a reference sheet before generating a single scene, which is what keeps the cook the same person throughout.

1. Looks at the available tools and what they cost, so the plan fits the job.
2. Loads its film-making guidance rather than improvising.
3. Checks the video generation options, then the possible durations and voices.
4. Checks the image generation options.
5. Creates the cook's voice before anything is filmed, so the narration can drive the pacing.
6. Creates a reference sheet of the cook and the ingredients — the single most important step for consistency.
7. Generates the four scenes in parallel rather than one after another.
8. Checks the results of those generations before assembling.

Step 6 is what separates a coherent recipe short from four clips of four different kitchens. Step 7 is why it does not take four times as long as one scene.

What does the result look like?

From the run this page is written from, probed on the delivered file: 1080x1920, H.264, 24fps, 30.84 seconds, AAC audio. Vertical and just over thirty seconds, which is what a Shorts feed wants, and the narration is mixed in rather than delivered separately.

Ask for a longer version with more steps and Vizard Agent extends from the same reference sheet, so the cook and the kitchen stay the same across both cuts.

How long it takes was not measured separately here. Comparable work runs a median of 28 to 38 minutes end to end, and the upper quartile is several times the median. Across all projects the median cost by tier is Flash 47, Pro 55, Max 242, Ultra 263 credits.

When does this not work well?

Generated cooking video is convincing at a glance and unreliable in the details, and food is a subject where the details are the content. Vizard Agent builds what you described rather than what is true of your kitchen, so this is a category to watch through once before posting rather than to trust on sight.

• Hands doing precise things are the hard case. Chopping, folding, pouring. A wide shot of a step reads better than a close-up of fingers.
• The recipe is yours to verify. Vizard Agent builds what you described. Temperatures, times and quantities are your claims, and someone will cook from them.
• The food may not look like your food. If the dish has a specific appearance that matters, upload a photo and say to match it.
• Four steps is the practical ceiling in thirty seconds. More than that and each step becomes too short to follow.
• Ingredient counts and quantities are approximate on screen. Vizard Agent generates a plausible picture of "three eggs", and plausible is not the same as three. If the count is part of the instruction, say it in the narration rather than relying on the visual.
• Continuity across many scenes drifts. The reference sheet holds the cook and the kitchen well; a dozen scenes stretch it further than four.

How do you fix a result that came back wrong?

Name the step that is wrong. Vizard Agent keeps the reference sheet, the generated voice and every scene it made, so replacing one scene keeps the cook and the kitchen identical to the others rather than regenerating the short from the beginning — the reference sheet is the thing that makes a targeted fix possible.

• "Step three looks wrong." That scene is regenerated against the same reference sheet.
• "The narration is too fast." Re-voiced without touching the picture.
• "Make it thirty seconds exactly." A timing change across the existing scenes.

How does Vizard Agent compare to doing it yourself?

Filming a recipe short means a clean kitchen, decent light, two hands you can see, and enough takes that the sauce is cold by the end. It looks better when it goes well and it is a genuine production. Generating it trades that for speed and for never having to clean up.

Common questions

Can it use footage I filmed? Yes. Upload it and Vizard Agent uses yours for the steps, generating only what is missing.

Can the narration be in another language? Yes. Say which, and Vizard Agent generates the voice in that language rather than translating captions.

Will the cook look the same in every shot? That is what the reference sheet is for. It holds well across a few scenes and drifts as the count grows.

Does Vizard Agent write the script? Yes. Give it the dish and the steps and Vizard Agent writes the narration, though you can paste your own wording if the recipe has to be exact.

Can I get a longer version with more steps? Yes. Vizard Agent extends from the same reference sheet so the two versions match.