Veo 3.1: what a clip costs and how to use it

# Veo 3.1 prompt card · AIGE

How to write prompts for Google's Veo 3.1 and which settings to pick when you run it in AIGE (aige.ws).

Updated 2026-09-27 · By AIGE · Prompting advice comes from Google's own Veo guides; AIGE settings and credit prices come from the AIGE product. Sources are listed at the end.

## Which Veo to pick

AIGE offers three Veo 3.1 tiers. Google describes them like this: Veo 3.1 is for final production cuts where visual fidelity comes first, Fast is for standard production work, and Lite is the most cost-effective tier, built for high-volume work and fast iteration [5]. In AIGE, "Pro" is Google's standard Veo 3.1.

| Tier in AIGE | Use it for | Lengths | Resolution | Credits, 4 / 6 / 8 s (sound on) |
|---|---|---|---|---|
| **Pro** | The final render of a shot you have already tried | 4, 6, 8 s | 720p, 1080p, 4K | 160 / 240 / 320 · at 4K: 240 / 360 / 480 |
| **Fast** | Most everyday clips | 4, 6, 8 s | 720p, 1080p, 4K | 60 / 90 / 120 · at 4K: 140 / 210 / 280 |
| **Lite** | Trying ideas, framing and timing cheaply | 4, 6, 8 s | 720p; 1080p at 8 s only | 20 / 30 / 40 · 1080p 8 s: 64 |

- A simple way to work: find the shot on Lite or Fast, then render the keeper on Pro.
- 720p and 1080p cost the same on Pro and Fast. Only 4K costs more.
- Aspect ratio: 16:9 (landscape) or 9:16 (vertical).
- Sound is on by default. In the studio, Pro and Fast have an Audio switch. Pro without sound costs less (80 / 120 / 160). Fast at 720p or 1080p costs the same with or without sound. Lite always makes sound.
- Plans: Veo 3.1 Pro at 8 seconds comes with the Premier and Max plans. Veo is not included on the Free plan.
- Where you set this: the studio prompt bar lets you pick tier, length, resolution and audio. Clips made inside the Canvas video editor or timeline run on Fast at 720p with sound (60 / 90 / 120).
- The studio shows the credit cost before you run. If a generation fails, AIGE returns its credits.

## The prompt formula

Google's Veo 3.1 guide suggests a five-part formula [4]:

**[Cinematography] + [Subject] + [Action] + [Context] + [Style & ambiance]**, then the sound in separate sentences [2].

| Element | What to write | Example |
|---|---|---|
| Cinematography | Shot size, camera angle, camera move | `Slow dolly in, low angle, medium close-up.` |
| Subject | Who or what, with specifics | `A woman in her sixties with cropped silver hair and a yellow raincoat` |
| Action | What happens in this one moment | `unfolds a paper map and frowns at it` |
| Context | Where and when | `at a windy bus stop on a coastal road, late afternoon` |
| Style & ambiance | Look, light, color, mood | `Overcast light, muted teal tones, 35mm film grain, quiet and wistful.` |
| Focus and lens | Depth of field, lens type | `Shallow depth of field, background softly blurred.` |
| Audio | Dialogue, sound effects, ambient noise | `She sighs and says: Wrong bus again. Ambient noise: wind and distant waves.` |

Google lists subject, action and style as the core elements and marks camera, composition, focus/lens and ambiance as optional [1]. Put together:

> Medium close-up, slow dolly in at eye level. A woman in her sixties with cropped silver hair and a yellow raincoat unfolds a paper map at a windy bus stop on a coastal road, late afternoon. Overcast light, muted teal tones, 35mm film grain, shallow depth of field. She sighs and says: Wrong bus again. Ambient noise: wind gusts and distant waves.

What Google recommends on top of the formula:

- **Be specific.** Google's example: "a woman in her twenties with wavy brown hair and light freckles" gets more specific results than "a brown-haired woman" [6]. Specific subjects avoid generic output [2].
- **Be clear and direct.** Replace rambling descriptions with plain shot language, for example "Low-angle close-up shot of a man with a somber expression" [3].
- **One moment per clip.** Chaining "A, then B, then C" in a short clip often gives muddled or incomplete results. Make each part its own clip [3].
- **Timed shots in one clip.** To direct several shots inside one 8-second clip, Google shows timestamp prompting [4]:

```
[00:00-00:03] Close-up of a barista's hands pouring steamed milk into a cup, drawing a leaf. SFX: the hiss of the steam wand.
[00:03-00:05] Medium shot, the barista slides the cup across the counter toward the camera.
[00:05-00:08] Over-the-shoulder shot of a customer lifting the cup and smiling. Ambient noise: café chatter, cups clinking.
```

## Writing sound and dialogue

Veo generates the soundtrack together with the picture, synced to it, from cues in your prompt [1]. Google recommends saying clearly that you want audio and describing it in separate sentences [2].

**Dialogue.** Google documents two ways to write a spoken line:

- With quotation marks: `A woman says, "We have to leave now."` [1][4]
- After a colon, without quotation marks: `A woman says: We have to leave now.` Google's Cloud guidance recommends this form to prevent the model from rendering the words as text in the video [3].

If words or subtitle-like text show up on screen, switch to the colon form. The examples on this card use it.

- Name each speaker before the line (`Man:`, `Woman:`) and add how they say it, as a note in parentheses (`Woman (voice tight with fear):`) or in words (`says in a calm, low voice:`) [1][3].
- Keep lines short. Google publishes no word limit per clip, but every dialogue example in its guides is one or two short sentences.
- To keep a voice the same across clips, reuse the same voice description word for word, e.g. "in a crisp, clear voice with a standard American accent" [3].
- English is fully supported. Google says other languages have not been evaluated, so results can vary [1].

**Sound effects.** Describe each sound clearly. A label helps: `SFX: thunder cracks in the distance.` [4]

**Ambient noise.** Describe the background soundscape: `Ambient noise: the quiet hum of an office.` [2][4]. Google's prompt guide also uses a plain `Audio:` label [6].

**Sound off.** If you will add your own track, switch Audio off in the studio (Pro and Fast). You can also mute or replace sound later in AIGE's video editor.

## Camera and lens vocabulary

These are the terms Google documents for Veo [1][2][4]. Google notes that some advanced camera angles and lenses are not officially supported, so results can vary [2].

| Angle and framing | What it does | Camera movement | What it does |
|---|---|---|---|
| Eye-level shot | Neutral view from human height | Static shot | Camera does not move |
| Low-angle shot | Below, looking up: subject looks powerful | Pan / tilt | Turns left-right / up-down from one spot |
| High-angle shot | Above, looking down: subject looks small | Dolly in / out | Moves toward or away from the subject |
| Bird's-eye view, top-down shot | Straight down, map-like | Truck left / right | Slides sideways, parallel to the subject |
| Worm's-eye view | From the ground straight up | Pedestal up / down | Rises or lowers, staying level |
| Dutch angle | Tilted horizon, unease or energy | Zoom in / out | Lens magnifies; camera stays put |
| Extreme close-up / close-up | A small detail / the face | Crane shot | Rises or sweeps, often for a reveal |
| Medium shot | Waist up; Google notes it is common for dialogue | Aerial / drone shot | Flies high over the scene |
| Full shot | Whole body with some setting | Tracking shot | Follows the subject |
| Wide or establishing shot | Subject small in the setting | Handheld | Less stable; feels real or urgent |
| Over-the-shoulder shot | From behind one person toward another | Whip pan | Very fast pan that blurs |
| POV shot / two-shot | Through a character's eyes / two subjects in frame | Arc shot | Circles around the subject |

**Lens and focus:** wide-angle lens, telephoto lens, macro lens, shallow depth of field (bokeh), deep depth of field, soft focus, rack focus, lens flare, fisheye, vertigo effect (dolly zoom) [1][2][4].

**Time and editing:** slow-motion, time-lapse, fast-paced action; jump cut, match cut, montage [2].

## Negative instructions

AIGE's Veo has **no negative-prompt field**, in the studio or over MCP. Everything goes in the main prompt.

What Google says:

- Describe what the scene should be, including what it lacks. Google's example: write "a desolate landscape with no buildings or roads" instead of just "no man-made structures" [4].
- For its separate negative-prompt field (not available in AIGE), Google advises against instructive words like "no" or "don't" and suggests listing the unwanted things instead ("wall, frame") [2].

In practice, write the clean version of the shot: `An empty beach at dawn, smooth untouched sand` rather than `no people, no footprints`.

## Starting from an image / first and last frame

AIGE supports both, on Pro, Fast and Lite.

- **One image → image-to-video.** Veo uses your image as the first frame [1]. In the studio, attach one image.
- **Two images → first and last frame.** Veo makes the shot move from the first image to the last [1]. In the studio, select two images (the second becomes the end frame) or set an end frame. On Lite, a first-and-last-frame clip is always 8 seconds.

How Google says to prompt from an image [3]:

- **Use a sharp, well-composed source image.** Veo takes character detail, light and style from it. The image-to-video endpoints AIGE calls ask for 720p or higher in 16:9 or 9:16; on Pro and Fast, other shapes are cropped to fit [7].
- **Prompt for motion only.** Describing the character, background or light again confuses the model. Describe the camera move, the subject's movement and what changes in the scene.
- **Refer to people generally:** "the woman", "he", "they".
- **Pick the movement type:** camera motion (the simplest and most reliable, e.g. `Slow dolly in on the subject.`), subject animation (`Her hair and scarf flutter in the wind.`) or environmental animation (`Fog rolls in slowly across the field.`).

For first and last frame, describe the transition and the sound you want [4]:

> The camera makes a slow 180-degree arc shot around the chef, from her face lit by the stove to the busy dining room behind her. SFX: a sizzling pan, then restaurant chatter rising as the room comes into view.

Not available in AIGE for Veo: reference images ("ingredients to video"), Google's video extension, and a seed setting. The Canvas "Extend" makes a new clip from the last frame. It does not use Google's extension feature.

## Common problems and fixes

| Problem | Documented fix |
|---|---|
| Words or subtitle-like text appear when someone speaks | Write speech after a colon without quotation marks: `She says: We have to leave now.` [3] |
| The clip is muddled or skips part of the action | One moment per clip. Split a sequence into separate clips, or direct timed shots with timestamps [3][4] |
| The result looks generic | Add specifics: age, hair, clothing, place, time of day, light [2][6] |
| Image-to-video drifts from your picture | Prompt only the motion, and call the subject "the woman" or "he" [3] |
| A character looks or sounds different in each clip | Paste the same full character and voice description into every prompt. Change only the action and setting [3] |
| Something unwanted keeps appearing | Describe the scene without it (see Negative instructions) [4] |
| An unusual angle or lens is ignored | Some advanced angles and lenses are not officially supported. Use a standard term from the lists above [2] |
| A generation is blocked or fails | Veo runs safety filters and sometimes blocks a video because of problems with the audio [1]. Rephrase and run it again. AIGE returns the credits for a failed generation |
| The source image gets cropped | Use a 16:9 or 9:16 image, 720p or larger [7] |

Note on text-to-video: if a prompt fails a content or validation check, the provider may rewrite it instead of rejecting it. Its `auto_fix` setting is on by default for text-to-video, and AIGE leaves it that way [7].

## 8 example prompts

Credits are for sound on, from AIGE's price table. Examples 2 and 4 use Pro at 8 s, which comes with the Premier and Max plans.

**1. Talking-head UGC, one line of dialogue**
> Medium close-up, handheld selfie-style shot at eye level. A woman in her late twenties with curly shoulder-length hair and a grey hoodie sits cross-legged on her bedroom floor and holds a small amber skincare bottle up to the lens. Soft daylight from a window on the left, warm tones, shallow depth of field. She smiles at the camera and says in a relaxed, friendly voice: I stopped buying ten products. This one does the job. Ambient noise: quiet room tone, faint traffic outside the window.

Fast · 8 s · 9:16 · 720p · **120 credits**

**2. Product reveal with sound design**
> Slow dolly in at a low angle. A matte black wireless speaker stands on a wet slate surface in a dark studio. A thin rim light sweeps across it from behind and picks out the fabric texture of its grille, while water droplets bead on the slate. Shallow depth of field, cool blue and silver palette, high-end commercial look. SFX: a low bass hum swells, then a crisp click as a ring of white light turns on around the top of the speaker. Ambient noise: soft studio room tone.

Pro · 8 s · 16:9 · 1080p · **320 credits** (draft it on Fast first: 120)

**3. Food ASMR close-up**
> Extreme close-up with a macro lens, static shot. A bread knife saws slowly through a golden sourdough loaf on a worn wooden board; the crust cracks and a wisp of steam rises from the soft crumb. Warm side light from a kitchen window, shallow depth of field, rich brown and cream tones. Audio: only the loud crackle of the crust and the scrape of the knife on the board.

Fast · 6 s · 9:16 · 720p · **90 credits**

**4. Cinematic landscape**
> Sweeping aerial drone shot at sunrise, flying low over a still fjord between steep green cliffs and rising slowly to reveal a small red wooden cabin on the shore. Thin mist drifts over the water, golden hour light, deep depth of field, shot on 35mm film. Ambient noise: wind over open water, distant seabirds.

Pro · 8 s · 16:9 · 4K · **480 credits** (at 1080p: 320)

**5. Interview or podcast shot**
> Medium shot, static camera at eye level. A man in his forties with a short grey beard and a navy sweater sits at a wooden desk in a small podcast studio, a black microphone on a boom arm in front of him. Warm practical lamps, acoustic foam panels behind him in soft focus. He leans toward the microphone and says in a calm, low voice: Our first year, we did everything wrong. Ambient noise: the quiet hum of the studio.

Fast · 8 s · 16:9 · 1080p · **120 credits**

**6. Comedic skit beat**
> Claymation style, medium shot of a cramped office kitchen. A clay office worker in a blue tie opens the fridge and finds a clay pigeon eating his sandwich on the middle shelf. Worker (shocked): That is my lunch. Pigeon (calm, deadpan): It was. Ambient noise: the hum of the fridge. SFX: a crumb drops onto the shelf with a soft tap.

Fast · 8 s · 9:16 · 720p · **120 credits** (try the timing on Lite first: 40)

**7. Travel vlog**
> Handheld POV shot walking slowly through a narrow market street in Marrakech at dusk. Stalls of spice pyramids and hanging brass lanterns glow on both sides as people pass close to the camera. Warm orange tones, soft lens flare from a lantern. Ambient noise: busy market chatter, a moped passing, music from a distant stall.

Lite · 8 s · 9:16 · 1080p · **64 credits**

**8. Ad hook, vertical**
> Low-angle close-up in slow motion. A runner's neon-green shoe slams into a puddle on a city street at night and throws a sheet of water toward the lens. Neon signs reflect in the wet asphalt, high contrast, cool blue and magenta tones. SFX: a sharp splash, then a deep bass hit.

Pro · 4 s · 9:16 · 1080p · **160 credits** (draft on Fast: 60)

## Settings and prices in AIGE

Credits per clip, with the exact price keys AIGE charges.

| Tier | Resolution | Sound | 4 s | 6 s | 8 s | Price keys |
|---|---|---|---|---|---|---|
| Pro | 720p / 1080p | on | 160 | 240 | 320 | `generateVideoVeoFalAudio4s` / `6s` / `8s` |
| Pro | 4K | on | 240 | 360 | 480 | `generateVideoVeoFalAudio4s4k` / `6s4k` / `8s4k` |
| Pro | 720p / 1080p | off | 80 | 120 | 160 | `generateVideoVeoFal4s` / `6s` / `8s` |
| Pro | 4K | off | 160 | 240 | 320 | `generateVideoVeoFal4s4k` / `6s4k` / `8s4k` |
| Fast | 720p / 1080p | on | 60 | 90 | 120 | `generateVideoVeoFalFastAudio4s` / `6s` / `8s` |
| Fast | 4K | on | 140 | 210 | 280 | `generateVideoVeoFalFastAudio4s4k` / `6s4k` / `8s4k` |
| Fast | 720p / 1080p | off | 60 | 90 | 120 | `generateVideoVeoFalFast4s` / `6s` / `8s` |
| Fast | 4K | off | 120 | 180 | 240 | `generateVideoVeoFalFast4s4k` / `6s4k` / `8s4k` |
| Lite | 720p | always on | 20 | 30 | 40 | `generateVideoVeoFalLight4s` / `6s` / `8s` |
| Lite | 1080p | always on | – | – | 64 | `generateVideoVeoFalLight1080p8s` |

- Lengths: 4, 6 or 8 seconds only.
- Aspect ratio: 16:9 or 9:16.
- Lite: no 4K. 1080p only at 8 s (a 1080p pick at 4 or 6 s renders at 720p). First-and-last-frame clips are 8 s. Sound is always on.
- Inputs: a prompt, plus an optional start image, or a start image and an end image.
- Not exposed in AIGE: negative prompt, seed, reference images, Google's video extension, provider safety or auto-fix settings.

## Use it from Claude or ChatGPT

AIGE connects to Claude and ChatGPT over MCP. Add a custom connector with the server URL `https://aige.ws/mcp` and sign in on the AIGE screen (steps at aige.ws/mcp-info). Clips are paid from your AIGE credits at the same rates as the studio.

The tool is **`antix_generate_video`**. For Veo:

| Parameter | Value for Veo |
|---|---|
| `model` | `"veo"` (always the **Pro** tier; Fast and Lite are studio-only) |
| `prompt` | Up to 2,000 characters |
| `duration` | 4, 6 or 8 (other numbers snap to the nearest: 5 → 6, 7 → 8, 10 → 8) |
| `aspectRatio` | `"9:16"` or `"16:9"` (other vertical ratios become 9:16, everything else 16:9) |
| `resolution` | `"720p"` (default), `"1080p"` or `"4k"` |
| `audio` | On unless you pass `false` (Pro silent: 80 / 120 / 160) |
| `sourceImage` / `endImage` | Start frame; start + end frame for a transition |
| `characterId` | A saved AIGE character; the server uses its latest image as the start frame |
| `count` | 1 to 4 clips, each charged separately |

Veo over MCP rejects `referenceImages`, `referenceVideos`, `referenceAudios` and `sourceVideo`. The task finishes asynchronously: the assistant polls `antix_check_task` with the returned `taskId`. Because MCP always uses the Pro tier, an 8-second Veo clip there needs the Premier or Max plan, same as in the studio.

A realistic ask:

> Use AIGE to make a Veo clip, 8 seconds, vertical 9:16, 1080p, with sound. Prompt: "Medium close-up, handheld selfie-style shot at eye level. A woman in her late twenties with curly shoulder-length hair … She smiles at the camera and says in a relaxed, friendly voice: I stopped buying ten products. This one does the job. Ambient noise: quiet room tone." Check my credits first.

What the assistant sends:

```json
{
  "model": "veo",
  "prompt": "Medium close-up, handheld selfie-style shot at eye level. …",
  "duration": 8,
  "aspectRatio": "9:16",
  "resolution": "1080p",
  "audio": true
}
```

Cost: **320 credits** (`generateVideoVeoFalAudio8s`).

## Sources

1. Google AI for Developers, "Generate videos with Veo 3.1 in Gemini API": https://ai.google.dev/gemini-api/docs/veo
2. Google Cloud, "Video generation prompt guide": https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/video-gen-prompt-guide
3. Google Cloud, "Best practices for generating videos": https://docs.cloud.google.com/gemini-enterprise-agent-platform/models/video/best-practice
4. Google Cloud Blog, "The ultimate prompting guide for Veo 3.1" (Oct 15, 2025): https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1
5. Google Cloud Blog, "Introducing Veo 3.1 Lite and a new Veo upscaling capability on Vertex AI" (Apr 3, 2026): https://cloud.google.com/blog/products/ai-machine-learning/veo-3-1-lite-and-a-new-veo-upscaling-capability-on-vertex-ai
6. Google DeepMind, "How to create effective prompts with Veo 3": https://deepmind.google/models/veo/prompt-guide/
7. fal.ai, Veo 3.1 endpoints AIGE calls, model pages and input schemas: https://fal.ai/models/fal-ai/veo3.1 · https://fal.ai/models/fal-ai/veo3.1/fast · https://fal.ai/models/fal-ai/veo3.1/lite · schemas at https://fal.ai/api/openapi/queue/openapi.json?endpoint_id=fal-ai/veo3.1 (same pattern for `/image-to-video`, `/first-last-frame-to-video`, `/fast/…`, `/lite/…`)
8. AIGE MCP setup: https://aige.ws/mcp-info