Dify x Sonilo: Give Your Video Workflows a Licensed Soundtrack

Video workflows on Dify can now ship with sound. The Sonilo plugin is live on the Dify Marketplace: tools that compose licensed music and sound effects from the video your workflow just produced.

Until now, the soundtrack lived outside the pipeline. You rendered the video, downloaded it, searched a stock library, checked the license terms, and lined a track up in an editor by hand.

That manual step matters more as your pipelines get more autonomous. When an agent publishes video without a human review step, you want audio that is licensed for commercial use (terms apply).

With the plugin, you add a node, pass it the finished video, and get back an original soundtrack composed for that exact cut. The music is licensed and safe for commercial use (terms apply).

What the Sonilo plugin does

The plugin ships four tools: video-to-music, text-to-music, video-to-sound-effects, and text-to-sound-effects.

Video-to-music

This is Sonilo’s core capability and the reason it fits video workflows so naturally. Sonilo composes from the finished video, not from a text prompt describing it.

Pass it the video (a public or signed URL) and the model reads the cut: scene changes, pacing, where the mood turns. A hard cut at the eight-second mark shows up in the music. So does the slow push-in at the end. The track follows the edit and ends exactly when the video ends.

01From cut to soundtrack: musical moments land on the edit points

In a workflow, that difference is practical. A text-to-music step needs an accurate written description of footage the model never sees, and the resulting track still has to be trimmed and matched by hand afterward. A video-to-music node works from the artifact your pipeline already produced.

There’s also no prompt engineering involved. The agent doesn’t have to describe a video in words and hope the description was good. It passes the file.

Text-to-music

The plugin also generates music from a written prompt. That’s useful earlier in a pipeline, when an agent is working from a brief and the video doesn’t exist yet, or when a project calls for standalone tracks.

Video-to-sound-effects

From the same video upload, Sonilo can also generate royalty-free sound effects matched to what happens on screen: ambience under a long scene, an impact landing on a hard cut. The same video can be passed to video-to-music and video-to-sound-effects to generate the score and the effects layer as separate tracks.

Text-to-sound-effects

The fourth tool generates royalty-free sound effects from a written prompt: a door slam, rain on glass, a crowd swell. Useful when a specific moment needs a sound and there’s no video to work from.

Getting started

1. In Dify, open the Marketplace, search for Sonilo, and install sonilo-ai/sonilo.
2. Create an account at platform.sonilo.com and copy your API key from the dashboard. New accounts come with free credits.
3. Back in Dify, open the plugin’s authorization settings and add the key as SONILO_API_KEY.

Once authorized, the Sonilo actions appear alongside your other tools in the workflow editor.

The free trial covers a first pass on each tool: one video-to-music run, one video-to-sound-effects run, and two generations each for text-to-music and text-to-sound-effects. Commercial publishing rights come with the applicable paid plan (terms apply), so you can validate a workflow first and upgrade when it starts shipping real work.

02Sonilo plugin on the Dify Marketplace

Using it in a workflow

The fastest way to see it working is the template we published together: search the Template Marketplace for “Generate a licensed soundtrack for your video”, from the @sonilo org (look for the Partner badge).

The template is deliberately small. Four nodes:

• Start takes the URL of your rendered video.
• Sonilo sends the video off for scoring and returns the generated music.
• Code parses the response into clean outputs: the audio UR
• L, status, and duration.
• End returns them to the user.

03The template’s four-node workflow canvas

To run it:

1. Open the template from the Template Marketplace and add it to your workspace.
2. Click the Sonilo node to check its settings: the video input, an optional style prompt, and the output format.
3. Hit Test Run, give it a video URL, and start the run.

04Configuring the Video to Music node

Run it with any video and you get a soundtrack matched to that edit, licensed and safe for commercial use (terms apply).

05A successful run: the result panel returns the soundtrack URL

Listen: press play

▶ Open spiderwoman-original-track.m4a

0:00 / 0:19 ♪What the output sounds like: the track video-to-music composed for the cut below. Unedited output.

Here’s what that sounds like on a real cut, before and after:

▶ Open demo-before.mp4

Before: without music

▶ Open demo-before.mp4

After: with the composed track

♬The same cut without music, and with the track video-to-music composed for it. For this comparison, we merged the returned soundtrack with the video afterward.

If the first take isn’t right, run it again. Each generation is a fresh original track for the same cut, so auditioning options costs one more run.

The template is meant to be extended. Swap the Start node for the output of a video-generation step and the workflow scores each video as it’s produced; a merge step downstream (ffmpeg, or your editing tool) combines the returned audio with the video file. Wrap it in a loop and it scores a whole batch.

You can also add a second Sonilo node that calls video-to-sound-effects on the same video for a royalty-free effects track.

A second template, “Describe the vibe, get a licensed soundtrack”, turns a plain-language description into a precise style prompt and applies it when composing from your video.

Real-world use cases

Imagine building a marketing pipeline that scores its own videos: demo cuts and feature announcements run through the workflow, and every video leaves with an original track composed for it, licensed for commercial use (terms apply).

Or an agent that hands over finished work: chain a video-generation step into Sonilo, merge the returned track downstream, and the agent’s output arrives with a licensed soundtrack (terms apply) instead of waiting on a human to pick music. The same loop covers batch work, where each cut of a master gets a track composed for its own edit.

Try it

Install sonilo-ai/sonilo from the Dify Marketplace, get a key at platform.sonilo.com, and run the template on a video you’ve already made. Three steps, and you have a soundtrack composed for your video, ready to merge into the final cut.

About Sonilo

Sonilo is the world’s first professionally licensed video-to-music AI platform. It generates original soundtracks and sound effects from video, trained on a licensed catalog (including a Shutterstock partnership). Sonilo partners with ComfyUI, fal, Scenario, WaveSpeed, Shutterstock, Tapnow, and Pika, and is backed by B Capital. Learn more at sonilo.com.

About Dify.AI

Placeholder

Dify team to supply the standard About Dify.AI boilerplate from previous partner posts. Stand-in copy below.

Dify is an open-source platform for developing LLM applications. It brings agentic workflow building, RAG pipelines, model management, and observability together in one interface, and the Dify Marketplace connects those workflows to a growing catalog of model and tool plugins. Learn more at dify.ai.

Hear the result

/