Lite is the default
Lite opens first with text or one start frame, plus an optional end frame in image mode. Choose 720p/1080p, 4/6/8 seconds, and 16:9/9:16. Audio and Reference mode remain unavailable.
Use the Veo 3.1 AI Video Generator to prepare text, image, first-and-last-frame, or multi-reference video tasks with version-aware controls.
Loading generator…
Model overview
The Veo 3.1 AI Video Generator in ZMS AI prepares one short task from text, a start frame, optional end frame, or 1–3 references. It keeps Lite, Fast, Pro, mode, resolution, timing, and audio boundaries visible so a reviewer can distinguish a configured request from a delivered clip.

Google documents broader Veo 3.1 capabilities, but this page follows the controls in the current independent ZMS adapter. Use the Veo model family to compare versions, then return here only when one of these four input paths fits the approved brief. Signed-in task completion and final debit still require production verification.
Version boundaries
The Veo 3.1 AI Video Generator changes its available controls with the selected version and mode. Match the workflow to the input you have, then keep the same settings while comparing prompt revisions.
Lite opens first with text or one start frame, plus an optional end frame in image mode. Choose 720p/1080p, 4/6/8 seconds, and 16:9/9:16. Audio and Reference mode remain unavailable.
Fast keeps text, one start frame, and an optional end frame. It adds 4K, retains 4/6/8 seconds, and shows audio in text or image mode; live speed and delivery remain unverified.
Pro exposes text or image input, an optional end frame, two ratios, three durations, 720p–4K, and audio. It is a selectable adapter path, not verified quality or cost evidence.
Upload 1–3 references and describe the action. This Pro-only path submits resolution and optional negative direction, but not aspect ratio, duration, or audio.
Suggested task recipes
These are original, suggested preparation recipes—not prompts recovered from the four unledgered Veo files in public assets. Try injects only the visible prompt; select the named Lite, Fast, Pro, or Reference path and confirm its active controls before submitting.
Veo 3.1 task paths
4 starter recipesUse Lite for a short silent composition test where one camera move and one ending frame carry the assignment.
Wide view inside a whitewashed coastal studio at early morning. A ceramic artist opens one tall shutter and a band of cool daylight moves across a row of unfinished bowls. The camera makes one slow, level dolly forward from the doorway, keeping the window frame straight and every bowl in the same position. No dialogue, no extra people, no cut, no sudden exposure shift. End with the artist and the brightest bowl sharing the center of the frame.
Check straight architecture, bowl count, the single light change, and whether the dolly reaches the specified ending without a late speed jump.
Use Fast audio only when a small set of audible events can be matched to visible actions during review.
Medium close shot of a bicycle mechanic at a clean workbench. The mechanic spins the front wheel once, listens, then presses the brake lever and the wheel settles. Keep the camera locked at hub height. Time three sounds to the picture: a soft freewheel click during the spin, one brief pad rub as the lever closes, then quiet room tone. Preserve the same hands, spokes, tools, and wheel geometry. No speech, music, camera move, cut, or second action.
Verify wheel geometry first, then compare the click, brake rub, and stop against the exact visible timing; reject invented speech or music.
Use Pro image mode when an approved opening and closing frame define a physically plausible transition that the prompt must protect.
Move continuously from the supplied opening frame of an empty museum corridor to the supplied closing frame with the same visitor standing beneath the far skylight. Preserve the corridor width, stone joints, skylight geometry, visitor clothing, and cool daylight direction. Make one slow forward tracking move while the visitor enters naturally from the right, crosses once, and stops at the final position. No cut, morph, new artwork, camera roll, lighting change, or alteration to either anchor composition.
Compare the first and last frames directly, then inspect the crossing for identity drift, wall warping, extra artwork, or an implausible position jump.
Use Reference mode only when 1–3 approved images have distinct, non-conflicting jobs for subject, material, or environment.
Create one restrained product shot using the supplied references: keep the first image's compact radio silhouette and control layout, the second image's brushed aluminum material, and the third image's pale limestone studio environment. Place one radio on a low plinth as the camera performs a slow quarter orbit at product height. Preserve all buttons, grille spacing, proportions, and surface direction. No hands, text, logo changes, duplicate products, transformation, cut, or invented accessories. End on a clean three-quarter view with wide negative space.
Audit each reference job separately; reject blended geometry, changed controls, extra objects, material drift, or a composition that loses the requested negative space.
Archived motion reference
Watch foreground parallax, the lateral reveal, water reflection, tree detail, mountain stability, and whether the camera move resolves cleanly. These checks are more useful than judging one attractive frame in isolation.
Prior internal archived Veo 3.1 example, not a current ZMS AI live-generation test. The source project does not include a separate license or rights-clearance ledger.Practical workflow
Build one reviewable shot before planning a sequence. A consistent setup makes it easier to separate prompt changes from differences caused by version, mode, framing, or resolution.
Name the subject, setting, action, camera behavior, and final beat. Split unrelated locations or actions so each clip has one readable progression.
Use text for an open brief, image mode for an anchored composition, an end frame for a planned transition, or Reference mode for 1–3 guiding assets.
Confirm only the version, resolution, ratio, duration, audio, and negative direction shown by the active UI; hidden controls are not submitted implicitly.
After submission, check status, media, motion, framing, continuity, sound, and delivery fit. Until that path works end to end, call the workspace configured rather than proven live.
Production review
Treat upstream capability, the current ZMS control surface, and published output evidence as three separate claims. This ledger states what can guide a task today, what still needs a signed-in check, and what the archived landscape clip can honestly demonstrate.
| Decision | Current fact | Review action |
|---|---|---|
| Choose a task path | Lite, Fast, Pro, and Reference expose different input, resolution, timing, and audio controls. Upstream Veo documentation does not prove every option is available through ZMS. | Select from the visible interface, keep the path fixed during comparison, and move to the broader AI video generator if another family accepts the approved input more directly. |
| Interpret published evidence | The landscape clip is prior internal archived Veo 3.1 evidence with exact duration and date. The four newer public files lack the required machine-readable generation ledger. | Use the landscape only for motion review. Do not infer current latency, reliability, prompt adherence, commercial clearance, or service delivery from it or from unpublished files. |
| Approve a production run | Client estimates reflect configured version and mode rules, while signed-in creation and final account debit remain unverified. ZMS is independent from Google and DeepMind. | Test one account task, retain prompt and settings, verify picture and audio, confirm reference rights, and check ZMS AI pricing before committing a batch. |
FAQ
Clear answers about local versions, end frames, references, audio, live verification, and the archived example shown on this page.
It is ZMS AI’s independent Veo 3.1 task workspace for Lite, Fast, Pro, text, image, optional end-frame, and multi-reference paths. A configured adapter is not proof of completed live delivery.
Lite offers text or image, optional end frame, 720p/1080p, and 4/6/8 seconds. Fast and Pro add 4K and audio. Quality, speed, delivery, and debit still need signed-in verification.
Yes. In image mode, upload a start frame and optional end frame. Lite, Fast, and Pro then apply their own resolution and audio boundaries.
Reference mode requires 1–3 images and forces Pro. It submits prompt, images, resolution, and optional negative direction, but not ratio, duration, or audio.
Audio appears for Fast and Pro text/image tasks, not Lite or Reference mode. Google documents native audio, but current ZMS audio delivery is not yet verified.
No. It is archived internal Veo 3.1 evidence, not proof of current quality, latency, reliability, delivery, or cleared commercial rights.