Omni Flash prompts in English, captions in your language

Google's Omni documentation says English is fully supported and that other languages have not been evaluated, so write Omni Flash prompts in English. If the audience speaks Korean or another language, add the localized text afterwards with Sume's video captions endpoint.

What Google says

The Omni docs state the language position above, and list system instructions, temperature, top_p and stop sequences as unsupported.

How Sume's route matches

Sume routes gemini-omni-flash-1.1 by request shape. The documented fields are prompt, image_url, end_image_url, reference_image_urls, reference_video_urls and video_url, with resolution, and aspect_ratio and duration on the generation modes. The docs list no system-instruction or sampling fields. Native audio is always on, so the clip arrives with sound; do not rely on the audio being in your target language.

Keep the prompt in English even if the final video is for a Korean audience.

Add captions in your language

POST /v1/video-captions takes a public HTTPS video_url. For text you author yourself, pass cues with text, start and end in seconds; that path skips speech-to-text, burns exactly your copy, and also works for silent clips. Set style to korean-ad for Korean karaoke-style captions, or omit it and Korean wording resolves to black-outline. script_text, words, cues and segments are mutually exclusive.

curl -X POST https://api.sume.com/v1/video-captions \
  -H "x-api-key: $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: omni-ko-captions-001" \
  -d '{
    "video_url": "https://example.com/omni-clip.mp4",
    "style": "black-outline",
    "cues": [
      {"text": "새로운 컬렉션", "start": 0.0, "end": 2.0},
      {"text": "지금 만나보세요", "start": 2.0, "end": 4.0}
    ]
  }'

Have a Korean speaker review the cue text before burning it in. Captions are a render job, so a fix means another job.