What is Google's documented Veo 3.1 prompt structure, and what do its first/last-frame, ingredients-to-video and timestamp workflows give for shot continuity?
Corrected 2026-10-24 - see C3-B955-07; the original statements below are kept for history and are superseded where the correction says so. Corrected 2026-10-24 - see C3-B825-04; the original statements below are kept for history and are superseded where the correction says so. Problem. Getting controlled, multi-shot output from one video model. Conditions. Veo 3.1 on Vertex AI per a Google Cloud blog dated October 15, 2025; clips of 4, 6 or 8 s, 720p or 1080p, 16:9 or 9:16. Method. (1) Five-part prompt: [Cinematography] + [Subject] + [Action] + [Context] + [Style & ambiance]. (2) Dialogue goes in quotation marks; SFX and ambient noise labelled (SFX:, Ambient noise:). (3) Negative guidance is phrased as the thing wanted ('a desolate landscape with no buildings or roads' instead of 'no man-made structures'). (4) First-and-last-frame: generate a start still and an end still (the guide uses Gemini 2.5 Flash Image), then describe the transition, e.g. 'smooth 180-degree arc shot'. (5) Ing...
Problem. Getting controlled, multi-shot output from one video model. Conditions. Veo 3.1 on Vertex AI per a Google Cloud blog dated October 15, 2025; clips of 4, 6 or 8 s, 720p or 1080p, 16:9 or 9:16. Method. (1) Five-part prompt: [Cinematography] + [Subject] + [Action] + [Context] + [Style & ambiance]. (2) Dialogue goes in quotation marks; SFX and ambient noise labelled (SFX:, Ambient noise:). (3) Negative guidance is phrased as the thing wanted ('a desolate landscape with no buildings or roads' instead of 'no man-made structures'). (4) First-and-last-frame: generate a start still and an end still (the guide uses Gemini 2.5 Flash Image), then describe the transition, e.g. 'smooth 180-degree arc shot'. (5) Ingredients to video: supply reference images of characters and the setting, then write a separate prompt per shot ('Using the provided images for the detective, the woman, and the office setting, create a medium shot...') to build a dialogue scene. (6) Timestamp prompting: one prompt with [00:00-00:02], [00:02-00:04]... segments each with its own shot (medium, reverse shot, tracking, crane). Craft note (inference): the reverse-shot segment in the example only keeps a consistent person if the same character description is repeated in each segment; the guide does not show an eyeline or screen-direction check. Failure modes. Guide says the features are experimental and actively improving; add/remove object runs on Veo 2 without audio. Limitations. Marketing-style vendor guide; no failure rates; features and model availability may have changed. Version/date. Oct 15, 2025.
Curator correction 2026-10-24 (discovery audit record C3-B825-04; sources re-fetched 2026-10-24, excerpts re-located by curator GET): Problem: older recipes cite veo-3.1-*-preview ids. Conditions: Gemini API (ai.google.dev) callers. Method: the docs page states the preview models were shut down on 22 October 2026 and points to Gemini Omni Flash or GA models on the Gemini Enterprise Agent Platform; the Agent Platform page lists veo-3.1-generate-001 and fast variants for first/last-frame work (veo-3.1-lite-generate-001 is labelled preview there, per page text). Failure modes: scripts with preview ids will fail (inference). Limitations: whether Omni Flash accepts the same two-frame conditioning is shown only on the Agent Platform page; I did not test. Version/date: page fetched 2026-10-24.
Curator correction 2026-10-24 (discovery audit record C3-B955-07; sources re-fetched 2026-10-24, excerpts re-located by curator GET): C3-B576-11 says add/remove object runs on Veo 2 and that the guide pairs Veo with Gemini 2.5 Flash Image, with a hedge that model availability may have changed. The current lifecycle page lists Veo 2 as retired and gives dates for the others. The Veo blog now also calls Veo 3.1 generally available (its later text still says preview). Veo 3.1 and Gemini 2.5 Flash Image have announced retirement dates, so a reader should not treat them as open-ended.
Curator qualifications: All cited documentation excerpts located on the live pages by curator GET; no video sources.
Limitations: Vendor blog; inference on continuity is mine. No numbers on adherence.
Evidence status: statements located in the cited sources by a correlated desk review (Claude); not executed; not independently reviewed.
Structured fields
{
"citations": [
{
"url": "https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1",
"note": "A formula for effective prompts; model capabilities; Mastering negative prompts; Workflow 1; Workflow 1 prompt; Workflow 2; Workflow 3; continuity note"
}
]
}Evidence claims
- [sourced_fact] Curator correction 2026-10-24 (discovery audit record C3-B825-04; sources re-fetched 2026-10-24, excerpts re-located by curator GET): Problem: older recipes cite veo-3.1-*-preview ids. Conditions: Gemini API (ai.google.dev) callers. Method: the docs page states the preview models were shut down on 22 October 2026 and points to Gemini Omni Flash or GA models on the Gemini Enterprise Agent Platform; the Agent Platform page lists veo-3.1-generate-001 and fast variants for first/last-frame work (veo-3.1-lite-generate-001 is labelled preview there, per page text). Failure modes: scripts with preview ids will fail (inference). Limitations: whether Omni Flash accepts the same two-frame conditioning is shown only on the Agent Platform page; I did not test. Version/date: page fetched 2026-10-24. · locator: curator correction paragraph in this record body (C3-B825-04, C3-B955-07) source
- [sourced_fact] Curator correction 2026-10-24 (discovery audit record C3-B955-07; sources re-fetched 2026-10-24, excerpts re-located by curator GET): C3-B576-11 says add/remove object runs on Veo 2 and that the guide pairs Veo with Gemini 2.5 Flash Image, with a hedge that model availability may have changed. The current lifecycle page lists Veo 2 as retired and gives dates for the others. The Veo blog now also calls Veo 3.1 generally available (its later text still says preview). Veo 3.1 and Gemini 2.5 Flash Image have announced retirement dates, so a reader should not treat them as open-ended. · locator: curator correction paragraph in this record body (C3-B825-04, C3-B955-07) source
- [sourced_fact] SUPERSEDED IN PART 2026-10-24 by curator correction C3-B825-04, C3-B955-07 (see the correction claim above); original kept for history: Problem. Getting controlled, multi-shot output from one video model. Conditions. Veo 3.1 on Vertex AI per a Google Cloud blog dated October 15, 2025; clips of 4, 6 or 8 s, 720p or 1080p, 16:9 or 9:16. Method. (1) Five-part prompt: [Cinematography] + [Subject] + [Action] + [Context] + [Style & ambiance]. (2) Dialogue goes in quotation marks; SFX and ambient noise labelled (SFX:, Ambient noise:). (3) Negative guidance is phrased as the thing wanted ('a desolate landscape with no buildings or roads' instead of 'no man-made structures'). (4) First-and-last-frame: generate a start still and an end still (the guide uses Gemini 2.5 Flash Image), then describe the transition, e.g. 'smooth 180-degree arc shot'. (5) Ingredients to video: supply reference images of characters and the setting, then write a separate prompt per shot ('Using the provided images for the detective, the woman, and the office setting, create a medium shot...') to build a dialogue scene. (6) Timestamp prompting: one prompt with [00:00-00:02], [00:02-00:04]... segments each with its own shot (medium, reverse shot, tracking, crane). Craft note (inference): the reverse-shot segment in the example only keeps a consistent person if the same character description is repeated in each segment; the guide does not show an eyeline or screen-direction check. Failure modes. Guide says the features are experimental and actively improving; add/remove object runs on Veo 2 without audio. Limitations. Marketing-style vendor guide; no failure rates; features and model availability may have changed. Version/date. Oct 15, 2025. · locator: A formula for effective prompts; model capabilities; Mastering negative prompts; Workflow 1; Workflow 1 prompt; Workflow 2; Workflow 3; continuity note source
Rights
| component | license | hosting | redistribution | modification | training | Evidence and notes |
|---|---|---|---|---|---|---|
| documentation | unknown | permitted | permitted | unknown | unknown | Applies to this original summary only, not to the cited sources. |