Images
sfx-gen
Sound effects from words or from your clip.
<!-- Generated by apps/docs/scripts/gen-tools.mjs. Do not edit here: the words live in apps/web/src/lib/tools/sfx-gen.tool.json (agent, hint) and packages/shared/src/tools/setting-hints.ts. --> On the rail as **sfx-gen**. ## When to use it Making sound effects and ambience: describe a sound in words, give a silent clip a soundtrack that follows what happens on screen, or get the clip back with that sound mixed in and its dialogue kept. ## When to reach for something else Not for music or a voice reading lines, which have their own tools. A clip must run at least one second, and its price grows with its length, up to ten minutes. ## How to use it 1. Drop a clip onto the card's **Video** slot. 2. Write what you want in **Prompt**. 3. Pick a mode on the card: **Text to sound**, **Video to sound**, **Video with sound**. 4. Pick a model in the pill at the top of the card, or keep **ElevenLabs SFX v2**. 5. Press **Run**. ## On the card | Control | What it does | Choices | | - | - | - | | Video | The clip the tool works from. Drop one from the board or your computer. | One clip | | Prompt | What you want, in your own words. The card sends it with your inputs. | Your words | | Run | Starts the run. The number on it is what this run costs in credits, worked out before you press. | | | ×1 | Runs the card several times in a row with the same settings. | 1 to 99 | ## Modes ### Text to sound A sound effect from words. Models in this mode: **ElevenLabs SFX v2**, **Mirelo SFX 2.0**. ### Video to sound A clip's soundtrack, as a separate sound. Models in this mode: **Mirelo SFX 2.0**. ### Video with sound The clip back, with its new sound mixed in. Models in this mode: **MMAudio v2**. ## Models The pill at the top of the card picks the model. **ElevenLabs SFX v2** is the one a new card starts on. | Model | Good for | | - | - | | [ElevenLabs SFX v2](#elevenlabs-sfx-v2) | One short sound from a text description in the Sound Effect tool: a door latch, a phone notification, wind in a stairwell, 1 to 22 seconds as mp3, with a Loop toggle that returns it seamless for ambience beds and UI states. | | [Mirelo SFX 2.0](#mirelo-sfx-20) | One sound effect or a stretch of ambience from a description, up to 30 seconds as a wav, when it should run longer than a short effect or leave something specific out, such as music or voices. | | [Mirelo SFX 2.0](#mirelo-sfx-20-1) | Giving a clip its own soundtrack as a separate sound file: footsteps, doors, weather and room tone that follow what happens on screen, to mix or edit apart from the picture. | | [MMAudio v2](#mmaudio-v2) | Giving a silent clip a soundtrack synced to the action, whether effects, room tone or music, from a short description or no prompt at all, baked back into the same video. | ### ElevenLabs SFX v2 **Not the one for:** Not music and not a soundtrack for a clip: text in only, no image, video or audio input, and it stops at 22 seconds. A track is the Music tool's job, and sound that follows a video is a video-to-audio job. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Duration | Length of the sound in seconds (1-22). | 1 to 22 | | Loop | Make a sound whose end runs smoothly into its start. | On or off | ### Mirelo SFX 2.0 **Not the one for:** Not music and not a voice reading lines, and it has no loop switch, so a seamless bed is the other text model's job. It costs about four times as much a second as the default text model. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Duration | Length of the sound in seconds (1-30). | 1 to 30 | | Leave out | Sounds to keep out of the result, as a short list. | Your words | ### Mirelo SFX 2.0 **Not the one for:** It returns sound only, so a clip with the sound already on it is the Video with sound mode. Not for music or a voice reading lines. Behind the settings button: | Setting | What it does | Choices | | - | - | - | | Keep dialogue | Keep the speech in the clip and make the rest of the sound around it. | On or off | | Leave out | Sounds to keep out of the result, as a short list. | Your words | ### MMAudio v2 **Not the one for:** The source calls it experimental and says it misses often; the sound length is set separately (1 to 30 seconds) and must match the clip, or the audio stops early. This model has no settings behind the settings button.On the rail as sfx-gen.
When to use it
Section titled “When to use it”Making sound effects and ambience: describe a sound in words, give a silent clip a soundtrack that follows what happens on screen, or get the clip back with that sound mixed in and its dialogue kept.
When to reach for something else
Section titled “When to reach for something else”Not for music or a voice reading lines, which have their own tools. A clip must run at least one second, and its price grows with its length, up to ten minutes.
How to use it
Section titled “How to use it”- Drop a clip onto the card’s Video slot.
- Write what you want in Prompt.
- Pick a mode on the card: Text to sound, Video to sound, Video with sound.
- Pick a model in the pill at the top of the card, or keep ElevenLabs SFX v2.
- Press Run.
On the card
Section titled “On the card”| Control | What it does | Choices |
|---|---|---|
| Video | The clip the tool works from. Drop one from the board or your computer. | One clip |
| Prompt | What you want, in your own words. The card sends it with your inputs. | Your words |
| Run | Starts the run. The number on it is what this run costs in credits, worked out before you press. | |
| ×1 | Runs the card several times in a row with the same settings. | 1 to 99 |
Text to sound
Section titled “Text to sound”A sound effect from words.
Models in this mode: ElevenLabs SFX v2, Mirelo SFX 2.0.
Video to sound
Section titled “Video to sound”A clip’s soundtrack, as a separate sound.
Models in this mode: Mirelo SFX 2.0.
Video with sound
Section titled “Video with sound”The clip back, with its new sound mixed in.
Models in this mode: MMAudio v2.
Models
Section titled “Models”The pill at the top of the card picks the model. ElevenLabs SFX v2 is the one a new card starts on.
| Model | Good for |
|---|---|
| ElevenLabs SFX v2 | One short sound from a text description in the Sound Effect tool: a door latch, a phone notification, wind in a stairwell, 1 to 22 seconds as mp3, with a Loop toggle that returns it seamless for ambience beds and UI states. |
| Mirelo SFX 2.0 | One sound effect or a stretch of ambience from a description, up to 30 seconds as a wav, when it should run longer than a short effect or leave something specific out, such as music or voices. |
| Mirelo SFX 2.0 | Giving a clip its own soundtrack as a separate sound file: footsteps, doors, weather and room tone that follow what happens on screen, to mix or edit apart from the picture. |
| MMAudio v2 | Giving a silent clip a soundtrack synced to the action, whether effects, room tone or music, from a short description or no prompt at all, baked back into the same video. |
ElevenLabs SFX v2
Section titled “ElevenLabs SFX v2”Not the one for: Not music and not a soundtrack for a clip: text in only, no image, video or audio input, and it stops at 22 seconds. A track is the Music tool’s job, and sound that follows a video is a video-to-audio job.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Duration | Length of the sound in seconds (1-22). | 1 to 22 |
| Loop | Make a sound whose end runs smoothly into its start. | On or off |
Mirelo SFX 2.0
Section titled “Mirelo SFX 2.0”Not the one for: Not music and not a voice reading lines, and it has no loop switch, so a seamless bed is the other text model’s job. It costs about four times as much a second as the default text model.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Duration | Length of the sound in seconds (1-30). | 1 to 30 |
| Leave out | Sounds to keep out of the result, as a short list. | Your words |
Mirelo SFX 2.0
Section titled “Mirelo SFX 2.0”Not the one for: It returns sound only, so a clip with the sound already on it is the Video with sound mode. Not for music or a voice reading lines.
Behind the settings button:
| Setting | What it does | Choices |
|---|---|---|
| Keep dialogue | Keep the speech in the clip and make the rest of the sound around it. | On or off |
| Leave out | Sounds to keep out of the result, as a short list. | Your words |
MMAudio v2
Section titled “MMAudio v2”Not the one for: The source calls it experimental and says it misses often; the sound length is set separately (1 to 30 seconds) and must match the clip, or the audio stops early.
This model has no settings behind the settings button.