AI meditation video generator: voice, calm loop, music bed

An AI meditation video generator builds a guided session from three parts: a slow narration voiced from your script, a calm visual loop, and a soft instrumental bed that dips under the voice. The video runs as long as the narration, so the script sets the length; on Sume one render runs up to 30 minutes.

Facts come from Sume's Video generation, Music Router, Music 1.0 and Timeline 1.0 docs and the Sume API reference, read on 2026-09-29. In the Agents tab you can ask for the session in plain words; the agent asks before it spends. Nothing here is a health or wellbeing claim.

How do I voice a guided meditation with AI?

Send the script to POST /v1/tts-1.0/generate as transcript, with a voice: voice.id, or the avatar_id of an avatar whose voice is ready. Then these settings matter for a meditation:

• generation_config.speed is a multiplier from 0.6 to 1.5; the low end gives the unhurried pace a session needs. Listen before you commit.
• language names the language of the script. Set it for any non-English script; left out, it defaults to English.
• One call takes up to 20,000 characters, and audio longer than 1,200 seconds fails with tts_duration_exceeded. For a longer script, voice it in passages and list the files in the render's audio.parts[] (up to 20 slices, joined without gaps).

How do I make the calm visual loop?

Make one short clip that ends on the frame it starts on, from a single calm still such as a lake at dusk or a candle flame, with small motion and a still camera. The render then replays it for the whole session with render.pad_mode: "loop". AI lofi video generator walks through the first-and-last-frame request for that clip, and Seamless loop AI video shows how to check the two ends for a jump.

What differs for a meditation is the length: a lofi loop plays under music alone, while here the narration decides how long the loop has to run, so render the voice first and read its length.

How do I put music under the voice?

Generate the bed with POST /v1/music-router/generate. Length is asked for in the prompt, since duration is rejected. Close the brief with “Instrumental, no vocals.” and, under narration, add “no spoken word”. AI meditation music generator has calm briefs to adapt.

Then one Timeline 1.0 render joins everything: the narration is the audio spine, the music is the soundtrack, and the loop fills one video[] slot as long as the session. In current code the clip's own sound is dropped, so viewers hear only the voice and the bed.

curl -X POST https://api.sume.com/v1/timeline-1.0/render \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: meditation-001" \
  -d '{
    "audio": {
      "url": "https://media.sume.com/artifacts/artf_demo/narration.wav",
      "duration_seconds": 1200
    },
    "soundtrack": {
      "url": "https://media.sume.com/artifacts/artf_demo/calm-bed.mp3",
      "loop": true, "duck_db": 6, "fade_out_seconds": 10
    },
    "output": { "width": 1920, "height": 1080 },
    "render": { "pad_mode": "loop" },
    "video": [
      { "source_url": "https://media.sume.com/artifacts/artf_demo/lake-loop.mp4", "start": 0, "duration": 1200 }
    ]
  }'

How much does an AI meditation video cost?

A 20-minute session, with an assumed script length and two music takes:

What are the limits?

• One render is at most 30 minutes; for a longer session, render parts and join them in your own editor.
• Every URL in the render must be this workspace's media.sume.com files, such as the narration, music and clip Sume returned.
• The default output is 1080×1920, a vertical frame; set output.width and output.height for landscape.
• duck_db needs a real spine: it is rejected when audio.mode is "silence".
• Music is generated per brief, not picked from a library, and has no seed or duration setting, so listen to each take.