# AI sound effect generator: turn text into sound effects

Describe a sound and Audo makes it in seconds with ElevenLabs' SFX v3: whooshes, impacts, footsteps, rain, and crowds, up to 30 seconds, as MP3 or 48 kHz WAV.

## How it works

1. Describe the sound: what happens, where, and how it feels. Or start from an idea, such as Whoosh, Rain, or Footsteps.
2. Choose how long it lasts, from half a second to 30 seconds, and turn on Loop for rain, wind, or a room.
3. Generate it, play your takes side by side, and download the best one as an MP3 or a 48 kHz WAV.

## Sound effects from text, made by ElevenLabs SFX v3

Audo's sound effect generator turns a written description into a sound effect in a few seconds. It runs ElevenLabs' SFX v3 model, makes sounds from half a second to 30 seconds long, and downloads them as an MP3 (44.1 kHz) or a WAV (48 kHz, 16-bit stereo), each named after what you asked for, so a folder of takes stays easy to read.

Use it for whooshes and impacts on video transitions, footsteps and doors in a scene, rain, wind, and crowds for ambience, notification chimes for an app, and game sounds. Every generation is a new take, so describe the sound once, generate it a few times, and keep the one that fits; Audo lists this visit's takes side by side to play and compare. Loop makes a sound that repeats seamlessly, and How closely to follow the words trades variety for a literal reading.

Sounds up to 10 seconds are free 3 times a day with no sign-up, and longer ones are paid per second, with no subscription. Open a sound in Audo's editor, or mix it under a voice in Mix audio. The same generator works from ChatGPT, Claude, and other assistants through Audo's MCP server, and from code through the API as generate_sound_effect.

Tool: Sound effect generator (`generate_sound_effect`). Price: 0.5 credits per generated second — $0.005 per generated second, pay-as-you-go with no monthly subscription — per job’s measured quantity, charged only when a job succeeds.

## Free API

```bash
curl -F 'text=Glass shattering on concrete' -F duration_seconds=3 https://audo.ai/api/try/sound-effect-generator -D response.headers -o response.body
```

Read response.headers and response.body before naming media. Handle 202, 200 JSON and errors with the complete response client at https://audo.ai/docs/api?utm_source=markdown&utm_medium=agent#responses.

Free for sounds up to 10 seconds long, 3 times a day per person, with no key or account. Descriptions up to 450 characters. Free results last one hour. The response is the result itself; a job that runs longer than about 30 seconds returns 202 with a link to check on it. With an API key (`-H "Authorization: Bearer $AUDO_API_KEY"`), omitting free uses credits if the balance covers the sound, otherwise makes it free when it's up to 10 seconds and quota remains. Set free=true to prohibit spending; for paid sounds use free=false with an approved positive max_credits. Nothing is cut short: a sound is made whole or not at all. Details: https://audo.ai/docs/api.md

## Options

Form fields with the same names as the MCP tool's inputs:

- `text`: What the sound is, up to 450 characters: the action, the place, and the feel, such as "Footsteps on gravel, approaching slowly, then stopping". Not speech to read aloud (text_to_speech).
- `duration_seconds`: The sound's length in seconds, 0.5 to 30, such as 2.5. Default 5. Billed per second.
- `loop`: Makes a sound that repeats seamlessly, with no audible start or end, for ambiences and beds. Default false.
- `prompt_influence`: How closely the sound follows the words, 0 to 1: higher is more literal and less varied. Default 0.3.
- `model`: The sound model and its price a generated second: eleven_sfx_v3 (Eleven SFX v3, 0.5 credits). Default eleven_sfx_v3.
- `output_format`: mp3 (44.1 kHz, 192 kbps) or wav (48 kHz, 16-bit, stereo). Default mp3.

## Languages

Describe sounds in English, the language ElevenLabs documents its sound effect descriptions in. The sounds themselves have no language, though a crowd or a voice in one may speak.

## From an assistant (MCP)

Add https://audo.ai/mcp and sign in (setup for each assistant: https://audo.ai/setup.md). The tool is `generate_sound_effect`:

> Makes a sound effect from a description, such as "glass shattering on concrete" or "rain on a window", as an MP3 or 48 kHz WAV, 0.5 to 30 seconds long (duration_seconds, default 5). loop makes it repeat seamlessly, for ambiences. Model: eleven_sfx_v3 (default eleven_sfx_v3). Not for speech (text_to_speech). It costs 0.5 credits a generated second, at least 1, charged only if the job succeeds; free up to 10 seconds, 3 times a day. Returns the sound with download links, or job_id for get_job if it takes longer.

Example: "Make a 3-second sound effect of glass shattering on concrete."

## Questions

**Is Audo's sound effect generator free?** Yes, for short sounds: Audo makes sound effects up to 10 seconds long free, 3 times a day, with no sign-up. Longer ones, up to 30 seconds, cost $0.005 a second, so $0.15 for the longest, shown before you generate and charged only if it works. There's no subscription or monthly plan.

**Which AI model makes the sounds?** Audo runs ElevenLabs' SFX v3, the newest ElevenLabs sound effects model, released in November 2026: the same model ElevenLabs' own Sound Effects offers. It follows descriptions with several parts more closely than earlier versions, sounds more natural, and handles musical and vocal sounds, such as a crowd chanting or a drum hit, better. On Audo you pay per sound, with no subscription.

**Can I use the sound effects in YouTube videos or commercial projects?** Yes. Audo's terms say your files and results stay yours, so a sound you make on Audo is yours to use in videos, podcasts, ads, and games, with no attribution. Don't describe sounds you don't have the rights to, such as a brand's trademarked jingle.

**How do I describe a sound so it comes out right?** Say what happens, where, and how it feels, rather than naming a category: "footsteps on gravel, approaching slowly, then stopping" works better than "footsteps". Name the material and the space ("a heavy metal door in an empty warehouse"), and put a sequence in order ("a door creaks open, then slams"). Exact counts, such as "four knocks", aren't always followed, so generate a few takes and keep the best.

**How long can a sound effect be, and how long does it take?** Each sound is half a second to 30 seconds long, in half-second steps. It's usually ready in a few seconds: in Audo's tests in November 2026, sounds of 1 to 12 seconds came back in 2 to 4 seconds.

**Can it make a sound that loops?** Yes. Turn on Loop and the sound has no audible start or end, so rain, wind, room tone, or a crowd can play on repeat under a video, a podcast, or a game scene.

**Should I download MP3 or WAV?** MP3 (44.1 kHz, 192 kbps) is small and plays anywhere. WAV is uncompressed 48 kHz, 16-bit stereo, which video editors such as Premiere Pro and DaVinci Resolve and game engines such as Unity take without converting.

**Can it make retro 8-bit game sounds?** Describe them, such as "a retro 8-bit coin pickup blip", and it makes them like any other sound. It's an AI model that makes a sound from words, not a chiptune synthesizer with sliders: for an exact 8-bit sound you tweak by hand, a synthesizer such as sfxr or Bfxr fits better.

**Can ChatGPT or Claude make sound effects?** Yes, with Audo added to it. Ask it for "a 3-second whoosh for a video transition", and it makes the sound with Audo's generator and hands you the MP3 to download. Claude works the same way. Add Audo to your assistant once, and it's there in every chat.

**What if I need speech, not a sound effect?** Use Text to speech to read a script aloud in the voice you choose. To put a sound effect under a recording, mix the two in Mix audio.

## More

- Every tool: https://audo.ai/llms.txt
- OpenAPI: https://audo.ai/api/openapi.json
- This page for people: https://audo.ai/sound-effect-generator?utm_source=markdown&utm_medium=agent