Adding Voice and Audio to NSFW Video

Nureta generates audio with your video by default — you do not have to ask for it. Moans, dialogue, breathing, room tone: the soundtrack is rendered alongside the picture. You can also swap in your own audio track.

This guide covers both, how to prompt for the sound you want, and how to keep voice from fighting the motion.

Audio is on by default#

generate_audio defaults to true. Every clip comes back with sound unless you turn it off. There is no separate audio step and no extra call — the model scores the clip as it renders it, and audio is included in the per-second price either way.

What gets generated#

• Vocal reactions — moans, breathing, gasps, matched to the on-screen motion.
• Dialogue — short spoken lines when your prompt calls for them.
• Ambience — room tone, bed sounds, and environment that fits the setting.

Prompt for the sound you want#

Audio follows the prompt. If you want a specific vocalization or a line of dialogue, put it in the text: she moans softly, building or he says "don't stop" gives the model something to voice. Leave it out and it fills in reactions that match the motion. Keep spoken lines short — long speeches drift out of sync with the mouth.

Bring your own audio#

Prefer a specific track — a voiceover you recorded, licensed music, a moan bed? Attach it as an audio_url content item. At most one audio item per task, hosted at an HTTPS or HTTP URL. If it is not already hosted, upload it the same way as an image via the presign endpoint.

curl -s -X POST "$TOKENSTORE_URL/api/v3/contents/generations/tasks" \
  -H "Authorization: Bearer sk-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seahorse-720p",
    "duration": 10,
    "content": [
      { "type": "text", "text": "a woman on top, slow grinding, nude, warm light" },
      { "type": "audio_url", "audio_url": { "url": "https://.../voiceover.mp3" } }
    ]
  }'

Keeping lip and motion sane#

• Short dialogue lines sync better than long ones — the mouth only has a few seconds of screen time.
• When you supply your own audio_url, the picture is not driven by that track, so match its length to the clip so it does not cut off or run long.
• Heavy dialogue over an explicit shot pulls focus — let the motion carry the shot and keep speech sparse.

When to turn audio off#

Set generate_audio: false for a silent clip — useful when you will score it yourself in an editor, or when you only want the picture. You are not billed differently; audio is included in the per-second rate whether it is on or off.

curl -s -X POST "$TOKENSTORE_URL/api/v3/contents/generations/tasks" \
  -H "Authorization: Bearer sk-..." \
  -H "Content-Type: application/json" \
  -d '{
    "model": "seahorse-480p",
    "duration": 8,
    "generate_audio": false,
    "content": [{ "type": "text", "text": "a nude couple, cowgirl position, slow motion" }]
  }'

The full content shape and every field are in the generation API docs. New to the endpoint? Start with how to create an NSFW video from a prompt.