MMAudio - AI Model API and Pricing

Foley for silent clips

MMAudio is an open-source model that generates sound against video, paying attention to when things happen. Picture a door closing, a glass set on a table or waves hitting rocks. Each gets a sound placed where the motion is. That is the gap it fills. Plenty of video engines output picture only. A clip without room tone or impact sounds feels unfinished on social feeds. On Ugesi this model page sends only your text prompt. To have the sound follow an uploaded clip, use the Sound Effects tool. It runs the same model with your video.

Describe the sounds you expect rather than the visuals. "Footsteps on gravel, light wind, a distant dog bark" gives the model a clear brief. For a product shot: "soft click of a cap, liquid pour, quiet studio room tone".

Keep the clips short

Outputs are short, so MMAudio works best on single shots rather than edited sequences. Split a longer edit into its separate shots, run each one, and assemble the audio in your editor. This also lets you redo one bad sound without paying again for the rest.

What to listen for

Check sync first: does the impact land on the frame where contact happens? Then check that the sound fits the material, since wood, metal and glass should not sound alike. If a sound arrives early or late, nudge it in your editor instead of rerunning. If the material is wrong, name it explicitly in the prompt and run again.

Leave music out of the request. Foley and score are separate layers, and mixing them later gives you control over levels.

Cost

Each clip is 5 credits, about $0.05 at the largest pack. Ten separate clips cost 50 credits, about $0.53. MMAudio is one of 2 models for this job, and both charge 5 credits. Price does not decide between them. It is not linked to a specific tool at the moment.

MMAudio API

curl -X POST https://ugesi.com/api/generate \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "mmaudio",
    "prompt": "a product shot on white marble"
  }'

<?php
$ch = curl_init('https://ugesi.com/api/generate');
curl_setopt_array($ch, [
    CURLOPT_POST => true,
    CURLOPT_RETURNTRANSFER => true,
    CURLOPT_HTTPHEADER => [
        'Authorization: Bearer YOUR_API_KEY',
        'Content-Type: application/json',
    ],
    CURLOPT_POSTFIELDS => json_encode([
        'model' => 'mmaudio',
        'prompt' => 'a product shot on white marble',
    ]),
]);
$response = json_decode(curl_exec($ch), true);
curl_close($ch);
print_r($response);

type GenerateRequest = {
  model: string;
  prompt: string;
};
type GenerateResponse = {
  job_id: number;
  job_token: string;
  credits_held: number;
  status: "queued" | "succeeded";
  output_text?: string;
};

const body: GenerateRequest = {
  model: "mmaudio",
  prompt: "a product shot on white marble",
};

const resp = await fetch("https://ugesi.com/api/generate", {
  method: "POST",
  headers: {
    "Authorization": `Bearer ${process.env.UGESI_API_KEY}`,
    "Content-Type": "application/json",
  },
  body: JSON.stringify(body),
});
if (!resp.ok) throw new Error(`HTTP ${resp.status}`);
const job = (await resp.json()) as GenerateResponse;
console.log(job.job_id, job.status);

import requests

resp = requests.post(
    "https://ugesi.com/api/generate",
    headers={"Authorization": "Bearer YOUR_API_KEY"},
    json={
        "model": "mmaudio",
        "prompt": "a product shot on white marble",
    },
    timeout=60,
)
resp.raise_for_status()
print(resp.json())

{
  "model": "mmaudio",
  "prompt": "a product shot on white marble"
}

The API comes with every credit pack, from Lite to Enterprise, and uses the same balance as the website. API documentation and access →

License and terms of use

The license terms for MMAudio restrict commercial use (CC BY-NC 4.0). Use its results for personal and non-commercial projects. The full rules are in our Terms of service.

People also look for