Sonilo Video-to-Sound (Music + SFX Combined)

Generate a music bed and sound effects for a video in one call, balanced against each other and mixed by the backend — one charge instead of two separate generations. Use this whenever a video needs a full soundtrack (score + SFX), not just one or the other.

Setup: See the [setup-api-key](../setup-api-key) skill.

⚠️ Cost: makes one API call that may incur charges (billed once, not twice, even though it produces both layers). Only call when explicitly requested.

Transport: MCP or CLI

Pick one at the start of the session and stay on it. Do not mix the two inside a single job, and do not announce the choice.

1. Sonilo MCP tools visible in this session (videotosound and friends) — use them. This is the preferred path: it needs no shell, and it is the only one that survives a very long generation. If a call fails to authenticate — rather than failing on its inputs — this transport is not usable in this session: go to 2 instead of retrying it.
2. No usable Sonilo MCP tools, but sonilo account exits 0 — use the CLI commands below. Same API, same account, same credential file. Probe with sonilo account, not sonilo whoami: whoami exits 0 even when signed out, so it cannot tell the two states apart.
3. Neither — stop and run the [setup-api-key](../setup-api-key) skill. Do not call api.sonilo.com with curl to work around it; both transports handle uploads, polling and retries that a bare request does not.

Quick Start

MCP tool call (recommended)

video_to_sound(
    video_path="~/Desktop/trailer.mp4",
    music_prompt="Cinematic, building tension",
    sfx_prompt="Footsteps, wind, distant thunder"
)

video_to_video_sound(
    video_path="~/Desktop/trailer.mp4",
    music_prompt="Cinematic, building tension"
)

Python (pip install sonilo)

from sonilo import Sonilo

client = Sonilo()  # reads SONILO_API_KEY

mix = client.video_to_sound.generate(
    video="trailer.mp4",
    music_prompt="Cinematic, building tension",
    sfx_prompt="Footsteps, wind, distant thunder",
)
mix.save("soundtrack.wav")

video = client.video_to_video_sound.generate(video="trailer.mp4", music_prompt="Cinematic, building tension")
video.save("scored.mp4")

JavaScript / TypeScript (npm install sonilo)

import { SoniloClient, download } from "sonilo";
import { writeFile } from "node:fs/promises";

const client = new SoniloClient(); // reads SONILO_API_KEY

const mix = await client.videoToSound.generate({
  video: "./trailer.mp4",
  musicPrompt: "Cinematic, building tension",
  sfxPrompt: "Footsteps, wind, distant thunder",
});
await writeFile("soundtrack.wav", await download(mix.output_url));

const video = await client.videoToVideoSound.generate({
  video: "./trailer.mp4",
  musicPrompt: "Cinematic, building tension",
});
await writeFile("scored.mp4", await download(video.output_url));

CLI (npm install -g sonilo-cli or pip install sonilo-cli)

sonilo video-to-sound --video trailer.mp4 \
  --music-prompt "Cinematic, building tension" --sfx-prompt "Footsteps, wind, distant thunder" \
  --output soundtrack.wav

sonilo video-to-video-sound --video trailer.mp4 --music-prompt "Cinematic, building tension"

Unlike the music/sound-effects skills, both tools here have CLI commands. --stem music/--stem sfx (repeatable) additionally saves the individual layers next to the combined output.

cURL (raw REST API, no MCP host)

curl -X POST "https://api.sonilo.com/v1/video-to-sound" \
  -H "Authorization: Bearer $SONILO_API_KEY" \
  -F "[email protected]" \
  -F "music_prompt=Cinematic, building tension" \
  -F "sfx_prompt=Footsteps, wind, distant thunder"
# -> {"task_id": "..."}  poll GET /v1/tasks/{task_id}

Both endpoints are task-based (202 + poll), same as the sound-effects tools — the MCP tool waits for you.

Tools

Parameters

Prompting

No prompt is required — the model reads the cut. A short structured brief adds your intent on top. Since this endpoint generates music and SFX in one balanced call, both crafts apply:

• Pre-flight (inspect the video, caps, credits, verification): [references/preflight.md](../references/preflight.md)
• Music brief craft: [references/music-prompting.md](../references/music-prompting.md) · SFX action-map craft: [references/sfx-prompting.md](../references/sfx-prompting.md)

Workflow Tips

• Use this instead of chaining videotomusic + videotosfx. The two layers are balanced against each other by the backend (so the SFX doesn't fight the score), and it's one charge, not two.
• Both musicprompt and sfxprompt are optional — you can leave both unset and let Sonilo interpret the whole scene, or set just one to steer that layer while leaving the other automatic.
• ducking is off by default — turn it on for anything with a voice. Left off, the source speech is not in the mix at all: the output is generated music and effects only. That is the right default for a silent or music-only clip and the wrong one for a talking head, so check the source audio before calling (see the pre-flight reference) rather than after the user tells you the narration is gone.
• For videotovideosound, the source audio is dropped unless you say otherwise. keeporiginalsound=true keeps the whole original track under the generated mix; preservespeech=true keeps only the isolated speech. If a user reports "my dialogue disappeared", this is the fix.
• Only the combined mixed result is saved. The individual music/SFX/processed stems exist in the task body on the backend but are deliberately not downloaded — four files per call would bury the one the user actually wants. If stems are needed, call the REST API directly and inspect the task body.
• Want the video back with the soundtrack baked in? Use videotovideosound instead of videoto_sound.
• Don't know what it should sound like? Run [video-analysis](../video-analysis) first: one call returns a section plan plus ready-to-use generation prompts read off the footage, which beats guessing a prompt and rerolling. It is a paid call that generates nothing, so use it when the brief is genuinely unclear — not when the user already told you what they want.

Recovering a Timed-Out Call

Both tools are async; on timeout the error carries a taskid and the job keeps running (already charged). Call getsfxtask(taskid), or getgenerationtask(task_id) on the hosted server, later — see [task-recovery](../task-recovery).

Output Files

• videotosound: a single .wav, named from musicprompt (falling back to sfxprompt, then sound-<first 8 chars of the task id>).
• videotovideo_sound: a single .mp4 with the mix muxed in, named the same way (fallback v2v-sound-<first 8 chars of the task id>).

Error Handling

Common errors: 401 invalid key, 402 insufficient balance / trial exhausted, 413 file too large, 422 invalid parameters or malformed segments, 429 rate limit. See the [account](../account) skill.