AI video with sound vs voiceover plus lip sync: which Sume route

It depends on whether a face speaks. For B-roll, product motion and ambient sound, use a video model's own audio: on Sume, Gemini Omni Flash 1.1 has native synced audio always on, and the API rejects generate_audio: false. For a person speaking on camera, generate the voice with TTS first and then run a lip sync model on a posed still; Sume's docs state that video models do not lip-sync to generated TTS or to a later voiceover.

From the Video Router docs and the models index on origin/main b28f3b0, read 2026-10-13.

What does native audio cost?

For Omni, credits scale with the output seconds and the resolution (3 to 10 seconds, 360p to 4K, 16:9 or 9:16). The router has no bitrate_mode and no reference_audio_urls. Read capabilities from GET /v1/video-router/models because each model has its own envelope.

How do I add a music bed?

Music Router generates a track from a prompt. There is no duration field; Sume rejects duration and duration_seconds, so write the length in the prompt. Put exclusions in the positive prompt, for example Instrumental, no vocals. Every Music Router model uses the same flat credits per generation.

curl -X POST https://api.sume.com/v1/music-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: bed-001" \
  -d '{
    "model": "sume/music-auto",
    "prompt": "Warm lo-fi, 84 BPM, C minor. A 20-second track. Instrumental, no vocals."
  }'