Veo3o1 Prompt Troubleshooting Checklist

Use this checklist when a Veo3o1 video is complete but the subject, action, transition or composition differs from what you wanted. Pick one change for the next attempt. For a full prompt-writing method, read the Veo 3.1 prompt guide.

For pending tasks, confirmed failures or an unknown submission, check My Works and the generation settings guide before starting another generation.

Start with the task recordCopy section link

Before editing the prompt, record the settings used for the result:

• generation mode;
• model preset: Lite, Fast or Quality;
• duration and aspect ratio;
• output resolution;
• input images and their order;
• seed, if one was set;
• the exact prompt.

Changing the model, duration, images or seed can change the result even when the prompt stays the same. Keep those settings with your notes when comparing attempts. The same settings and seed do not guarantee an identical video.

Check whether the mode matches the problemCopy section link

Text to VideoCopy section link

Use text-only generation to describe a scene without supplying images. If the shot depends on a particular face, package, costume or opening composition, consider an image-guided mode.

First & Last FramesCopy section link

A starting image is required; an ending image is optional. If the opening looks right but the subject changes during the shot, simplify the requested movement. Check that the images show compatible subjects, locations and lighting. See the First & Last Frames guide for the setup.

Reference ImagesCopy section link

Use 1–3 reference images to guide a subject's appearance without specifying exact opening or ending frames. This mode is available with Fast and Lite, requires 8 seconds and uses 16:9 or 9:16. Auto is unavailable in this mode. Choose images that agree on the person, product version, wardrobe, materials and palette, and remove conflicting references. See the Reference Images guide.

Diagnose the visible symptomCopy section link

The main action is missingCopy section link

Name the subject and give it one clear action, using a direct verb.

Less specific:

A cyclist in a city, energetic, cinematic, dramatic light.

More specific:

A cyclist accelerates through one wet intersection while the camera tracks beside the bicycle at street level.

If the prompt contains several actions, locations, or time jumps, split it into separate shots.

The camera movement is wrongCopy section link

Separate subject movement from camera movement. For example: "The dancer turns clockwise while the camera stays still." Choose one camera instruction, such as a locked shot, slow push-in, lateral tracking, gentle orbit, tilt or pull-back.

A person or product changes appearanceCopy section link

Choose a clear reference and remove contradictory images. Name only the appearance details you want to keep. Avoid hiding the subject for long periods, fast rotations, heavy motion blur, costume changes or changes in lighting within the same short shot.

The start or end frame is not respectedCopy section link

Check that the starting and ending images show a compatible subject and a plausible change in composition. Describe how the subject and camera move between them. If you only need a starting image, leave the ending image empty. The images guide the request; they do not guarantee an exact frame match.

The result contains extra objects or textCopy section link

Describe the composition you want, then add a short exclusion aimed at a problem you can see, such as "clean background, no added text, no extra objects." Keep the exclusions brief so the main subject and action stay clear.

Vertical composition feels croppedCopy section link

Select 9:16 in the generator before submitting. Keep the subject near the center and leave space at the top or bottom for platform overlays. Describe a vertical shot rather than relying on a later crop of a horizontal composition.

Change one variable per retryCopy section link

Choose one change for the next generation and keep the other settings and prompt sections the same. For a prompt change, keep the model, seed, duration and images unchanged. For an image or model change, keep the prompt unchanged. This makes it easier to compare the attempts.

A practical order is:

1. Remove contradictions.
2. Reduce the scene to one main action.
3. Clarify subject movement.
4. Clarify camera movement.
5. Simplify or replace image inputs.
6. Change the model or duration only after clarifying the prompt, and stay within the selected mode's limits.

Changing the model preset can also switch the input mode and remove a selected image. Check the mode and image previews before submitting.

Review cost and delivery before another runCopy section link

Each new generation creates a new task and reserves the credits shown for its settings. Extending or enhancing a completed video is a separate result action with its own displayed credit requirement. See Extend and upscale results for eligible sources and the separate action workflow. Processing time can vary; the progress display is not a completion estimate.

New video records appear in My Works for up to 14 days from task creation. Result links can stop working sooner. In the generator, Download video opens the result link; save the versions you want to compare to your device. An extension or enhancement does not restart the original video's record window.

Pre-submit checklistCopy section link

• The selected mode fits the shot I want to make.
• The prompt names one primary subject and one main action.
• Subject motion and camera motion are separate.
• If I use reference images, their subject details, product versions, lighting and colors do not conflict.
• If I use an ending image, the two frames suggest a plausible transition.
• The aspect ratio suits where I will use the video.
• Exclusions are short and tied to a visible problem.
• I checked the credits required before submitting.
• I changed only one setting or prompt detail from the previous attempt.