Veo 3.1 tier & pipeline comparison
Veo 3.1 Capabilities: Standard, Fast & Lite
Same pocket-watch brief across tiers — text-to-video and image-to-video from one Nano Banana 2 still
What each tier can do with the same brief — not a ranked winner pick.
Workflow
Veo 3.1 does not generate still images — so we used Nano Banana 2 (Google Gemini 3.1 Flash Image) to create one reference photograph, then fed that same JPEG into each Veo tier's image-to-video endpoint. Separately, we also ran text-to-video on each tier with the video prompt (no input image). We generated one shared still, not three — a per-tier still (Lite / Standard / Fast image each) is the next iteration if you want stricter apples-to-apples on i2v.
Image model (all stills): Nano Banana 2 via fal.ai (fal-ai/nano-banana-2)
Prompts used
These are not identical — the image prompt describes a frozen photograph; the video prompt adds motion and audio.
Image prompt (Nano Banana 2)
Extreme close-up photograph of a hand holding an ornate antique brass pocket watch under a single warm desk lamp in a dim study at night. Intricate hand-engraved floral patterns on the watch case catch and scatter warm lamplight, fine dust motes visible in the light beam, shallow depth of field with softly blurred bookshelves in the background. Cinematic macro photography, rich shadow detail, warm color grade, photorealistic, 16:9 composition, no text, no watermarks.
Video prompt (Veo 3.1 — t2v and i2v)
Extreme close-up of a hand slowly rotating an ornate antique brass pocket watch under a single warm desk lamp in a dim study at night. Intricate hand-engraved patterns on the case catch and scatter the lamplight as it turns, fine dust motes drifting through the light beam, soft shallow depth of field. Ambient audio: a delicate ticking from the watch mechanism and a faint grandfather clock chiming twice in another room. Cinematic macro lens, rich shadow detail, warm color grade, no text overlays, no watermarks.
Reference image
One shared still from Nano Banana 2 — input for all image-to-video runs below.
Shared reference still
Nano Banana 2 · fal-ai/nano-banana-2
Used as image_url for all i2v runs. Seed 42, 1K JPEG, 16:9.
Comparison table
| Tier | Pipeline | Image input | fal.ai endpoint | Output file | Cost | Notes |
|---|---|---|---|---|---|---|
| Standard | t2v | — (none) | fal-ai/veo3.1 | veo-standard-v1.mp4 | €2.95 | Sharpest engraving detail in the t2v round. |
| Fast | t2v | — (none) | fal-ai/veo3.1/fast | veo-fast-v1.mp4 | €1.1 | Strong t2v result but visible temporal glitch mid-clip — surprising for a tier that often wins on value. |
| Lite | t2v | — (none) | fal-ai/veo3.1/lite | veo-lite-t2v-v1.mp4 | €0.59 | Extra run — Lite t2v was generated in addition to Lite i2v, which is why you see two Lite videos. |
| Lite | i2v | pocket-watch-reference.jpg (Nano Banana 2) | fal-ai/veo3.1/lite/image-to-video | veo-lite-i2v-v1.mp4 | €0.59 | Cheapest i2v tier off the shared still. |
| Fast | i2v | pocket-watch-reference.jpg (Nano Banana 2) | fal-ai/veo3.1/fast/image-to-video | veo-fast-i2v-v1.mp4 | €1.1 | i2v from shared still — compare to veo-fast-v1.mp4 (t2v) above; same tier, different pipeline. |
Text-to-video results
Prompt goes straight to Veo — no input image. Same video prompt on each tier.
Veo 3.1 Standard
T2V · €2.95
Text-to-video. Best fine detail on the watch engraving.
Veo 3.1 Fast
T2V · €1.1
Text-to-video (veo-fast-v1.mp4). Visible glitch mid-clip — re-test with another seed before shipping.
Veo 3.1 Lite
T2V · €0.59
Text-to-video — second Lite clip exists because this pipeline was run separately from Lite i2v.
Image-to-video results
Same Nano Banana 2 still + same video prompt animated on each tier.
Veo 3.1 Lite
I2V · €0.59
Image-to-video from shared Nano Banana 2 still.
Veo 3.1 Fast
I2V · €1.1
Image-to-video from shared Nano Banana 2 still. Different file from veo-fast-v1.mp4 (t2v).
Takeaways
- Image and video prompts are related but not the same: the still is a frozen photograph; the video prompt adds rotation, drifting dust, and audio cues.
- Only one Nano Banana 2 image was created — not one per Veo tier. All i2v clips animate that single JPEG.
- You have three t2v videos (Standard, Fast, Lite) and two i2v videos (Lite, Fast). Fast t2v is veo-fast-v1.mp4 — it is on the page under Text-to-video, not missing.
- Lite appears twice because we ran both text-to-video and image-to-video for Lite. Standard i2v was cancelled but a file may exist on disk from a partial run — it is not published here.
- Fast t2v showed a noticeable glitch despite being the usual value pick — worth re-running with a different seed before trusting it on hero content.
- Next iteration: generate three Nano Banana 2 stills (or Imagen) — one per tier brief — then i2v Standard, Fast, and Lite each from its own image.
Want the raw prompt behind this battle?
We'll email you the exact prompt and settings from this test, plus new battles as we publish them.