Early preview: draft and save prompts. AI rendering and payments are coming later.
shotmelo.
My drafts
All guidesPROMPT REVISIONS · UPDATED NOVEMBER 9, 2026

AI video negative prompts: check the model before writing the list.

Use a negative prompt when your selected model and interface support an exclusion field. Otherwise, describe the shot you want in positive terms. A list of unwanted elements and an instruction about what should happen are different inputs; copying one between tools can change its meaning.

1. Check the model, mode, and interface

A negative prompt describes content you want the generator to leave out. Before using one, record the model name, generation mode, and the field where the text will go. The name of the website alone is not enough: an app can offer several models with different controls.

Two documented workflows and an unknown interface
WorkflowDocumented guidanceWhat to do
Runway Gen-4 videoUse positive phrasing; negative phrasing is unsupported.Write the desired action and framing in the main prompt.
Veo through the documented Google Cloud text-to-video APIAn optional negativePrompt parameter is documented.Keep the scene prompt and unwanted elements in their respective fields.
An app or model whose controls you have not checkedSupport is unknown.Read that mode's current help; start with a clear positive brief.

These distinctions come from the official Runway Gen-4 prompting guide and Google Cloud's Veo text-to-video API documentation, checked November 9, 2026. They do not establish the controls of every Runway model, every Veo app, or a third-party wrapper. Follow the documentation for the specific route you use.

2. Translate the unwanted result into a visible intention

First decide what a good shot looks like. “Bad movement” is difficult to inspect. A cup that stays centered while steam rises is a concrete intention. The following rewrites are original planning examples, not tested generation results.

Replace a complaint with a shot you can review
Vague or negative instructionPositive shot descriptionReview the result for
Do not zoom.The framing stays fixed on the cup throughout the shot.The cup's apparent size and position.
No frantic motion.A thin curl of steam rises slowly above the cup.Steam pace and movement of other objects.
Do not add writing.A plain, unbranded ceramic cup rests on an undecorated placemat.Marks on the cup and overlays on the frame.
No changing shape.The cup keeps its rounded rim, single handle, and upright position.Rim and handle shape from start to finish.

These descriptions make your intention clearer; they are not enforcement rules. When working from an image, prepare a reference that already matches the intended object and framing. If you need a camera move, use the camera movement guide to choose one rather than mixing fixed framing with an orbit or zoom.

3. Copy a draft for the workflow you actually use

Imagine an original planning exercise: a plain terracotta cup on a pale placemat, with a small curl of steam. No reference image or rendered clip is supplied for this exercise. Write the main brief first:

Scene draft for the documented Veo Cloud text-to-video route

A plain terracotta cup with one handle rests on a pale placemat beside a window. A thin curl of steam rises slowly. The framing stays fixed, with the whole cup and its handle visible. Soft window light remains steady. End with the cup in the same position.

For a supported negative field, choose a short list tied to this shot. For example, if you want fixed framing and a plain image, a draft might list zoom, camera roll, overlay text. These are unwanted elements, not extra scene directions. Google's official negative prompting guidance recommends describing the exclusions themselves instead of adding commands such as “no” or “don't.”

Optional exclusion field for that supported route

zoom, camera roll, overlay text

Copy the unwanted terms into the documented negative field, separately from the scene above. The headings and review notes are for your plan; they are not part of either prompt. This example is not a claim that every interface offering Veo exposes the Cloud API parameter.

Gen-4 image-to-video: a positive motion draft

If your approved input image already shows the cup and placemat, describe the movement from that starting frame. This original draft assumes the full cup and its handle are visible and the image matches the planned shot. No image or rendered result is supplied here.

A thin curl of steam rises slowly above the cup. The camera holds the opening framing. The cup stays in place and ends in the same position.

Do not paste that exclusion list into the Gen-4 main prompt as though it were a supported negative field. Use the positive draft and adapt it to that model's guidance. If your selected app exposes no negative control, do not assume a hidden one exists because another API supports it.

Check for conflicts before generating. Excluding all movement fights a request for moving steam. Excluding text can conflict with a shot whose purpose is to show a printed label. Write down what must remain before deciding what to exclude.

Check the destination before you paste

  1. Put the scene or motion brief in the main prompt field. Supply an image through the actual reference control when the mode needs one.
  2. Put exclusions in a separate field only when the selected route documents that field. A heading such as “negative prompt:” inside the main text does not create a separate control.
  3. Save the provider, exact model, mode, field labels and text you used. Keep cut timing and acceptance notes in your production record.

4. Check what an exclusion might remove from the plan

Review each term against the action, source and intended edit. These are planning conflicts to resolve, not predictions of how a model will respond to a word. Narrow the exclusion or change the brief when the same element is both required and unwanted.

Original checks before reusing an exclusion list
The shot needsA conflicting exclusionDecision to make
Steam moving above a still cupAll movementKeep the steam action; describe fixed framing separately.
A readable product labelAll textDistinguish the required label from unwanted overlay text; plan label review or editing.
A deliberate camera orbitCamera movementDecide whether the shot should move or hold its framing.
A person making a visible gesturePeople or handsPreserve the intended subject; inspect the gesture rather than excluding its essential parts.

All examples are unrendered planning material. Scroll wide tables sideways on a narrow screen; keyboard users can focus a table region and use the arrow keys.

Words such as “perfect,” a longer defect list, or excluding an entire subject do not establish a successful take. For a difficult hand gesture, keep the hand in the shot and use the hand and contact checklist to record the visible issue.

5. Change one observed problem, then compare

A long generic defect list does not tell you which decision helped. Our suggested review process starts with a small, understandable brief and changes one item at a time. Treat it as an editing method, not a promise that a particular term will fix an artifact.

  1. Keep a baseline: save the main prompt, reference, model, mode, and available settings.
  2. Name the visible problem: “The handle changes shape halfway through” is more useful than “low quality.”
  3. Locate the problem: is it in the starting image, generated motion, or a cut between clips?
  4. Choose one revision: correct the reference, simplify the action, rewrite a positive instruction, or adjust a supported exclusion field.
  5. Review the whole clip: check the intended action still happens and the revision did not damage something you wanted.

A small revision log

Provider / model / mode: ___
Main prompt version: ___
Optional negative field and text, or left blank: ___
Observed problem and clip time: ___
Desired visible behavior: ___
One change and the field used: ___
Result: kept / revised / discarded, because ___

Generation can vary even with the same wording. One better clip does not prove a phrase always works. If the issue is a character changing between shots, review the references and continuity plan using the character consistency checklist instead of expecting an exclusion list to establish identity.

Common negative prompt questions

Can I use one negative prompt for every AI video model?

Keep a reusable note of the problems you want to prevent, then translate that note for the current model and input fields. The two documented workflows above already require different handling.

Why does my Veo app not show the field used in this example?

The parameter cited here belongs to a specific documented Cloud API route. A hosted app or another integration can expose different controls. Check that interface's current help rather than assuming the field is hidden or that copying an exclusion list into the main prompt has the same effect.

Will a negative prompt repair my input photo?

Do not rely on it to repair a reference. Inspect the starting image and use a corrected version when an unwanted mark, shape, or crop is already present. Then decide how much motion the image can sensibly support.

Does Shotmelo send negative prompts to a video model?

The current Shotmelo preview lets you edit a prompt plan, preview a reference locally, and save or export text drafts. It does not connect to video generation or automatically test model-specific fields. Keep your optional exclusions in your planning notes and enter them in the appropriate supported field at your chosen provider.

Write the shot before the exclusions.

Use Shotmelo's editable prompt planner to organize the subject, action, and camera intention. Save the draft, then adapt it to the generation workflow you have checked.

Open the prompt planner

First published October 13, 2026. Updated November 9, 2026 with separate workflow examples, a field-placement check and an exclusion conflict audit. Linked Runway and Google Cloud guidance was rechecked on that date. Prompt drafts have not been rendered or tested.