Foggy forest morning clip with ambient sound: Gemini Omni, 10s

A foggy forest is one of the easiest things to ask Gemini Omni for: no faces, no hands, soft light, slow motion, and a soundscape that is mostly texture. Put the fog, the light shafts and three ambient sounds in the prompt and ask for a single unbroken scene. A 10-second 720p clip is $1.25 on Sume.

Mood clips like this are useful as a title-card background or a calm opener for a longer edit.

Why a people-free scene is a good first test

Landscapes hide the usual video-model errors. Nobody has fingers or lips to get wrong, and slow drifting fog is hard to judge as a mistake. It is a good scene for checking the model's audio, since the sound is the main content. Google's guide says to describe the soundscape you want (read 2026-10-07), so be concrete.

• Visual: 'low fog between tall pines, soft shafts of morning light, slow drift'.
• Camera: 'slow forward glide at knee height, single continuous shot'.
• Sound: 'distant birdsong, occasional water drips, a faint breeze in the needles'.
• Exclusions: 'no music, no dialogue, no human voices'.

Request and polling

The script below submits a 10-second 720p job and polls with a 20-second pause, which suits a job that normally runs for tens of seconds to minutes.

import os, time, requests

H = {'Authorization': 'Bearer ' + os.environ['SUME_API_KEY']}
prompt = ('Low fog between tall pines, soft shafts of morning light, slow drift. '
          'Slow forward glide at knee height, single continuous shot. '
          'Audio: distant birdsong, occasional water drips, a faint breeze in the needles. '
          'No music. No dialogue.')
body = {'model': 'gemini-omni-flash-1.1', 'prompt': prompt, 'duration': 10,
        'resolution': '720p', 'aspect_ratio': '16:9'}
r = requests.post('https://api.sume.com/v1/videos', json=body, timeout=60,
                  headers={**H, 'Idempotency-Key': 'fog-forest-1'})
r.raise_for_status()
url = r.json()['polling_url']
while True:
    time.sleep(20)
    s = requests.get(url, headers=H, timeout=60).json()
    if s['status'] in ('completed', 'failed', 'cancelled'):
        break
print(s['status'], s.get('unsigned_urls'), s.get('error'))

Cost

Sume bills list times 1.25 per second, rounded up to the cent per job; the Video Router doc lists $0.03, $0.10, $0.15 and $0.30 per second at 360p, 720p, 1080p and 4K.

What to watch for

Check that fog moves in one direction, that light shafts stay anchored to the trees, and that the birdsong does not stop abruptly at the cut-off. If the clip drifts into something unnatural at second eight, shorten to 6 seconds. For a webhook instead of polling, send an HTTPS callback_url on the request, which Sume signs with x-sume-webhook-signature; verify it with a non-empty secret before you trust the payload.

A calmer variant set

Mood clips are good for a set. Keep the same sound line and change one thing per clip: dusk fog, snow fog, rain fog. Ten seconds each, three variants, cost $3.75 at 720p. Use them as quiet intercuts between faster clips, with a short fade at each join. Because the audio is generated, the birdsong in one clip will not continue into the next; add a shared bed under the full edit if you want a continuous soundscape.