Mirelo SFX 2.0 Video-to-Video: give a silent clip a soundtrack
Mirelo SFX 2.0 Video-to-Video returns your clip with a new synchronized sound effects track already in it. When to pick it over audio-only, with prompts.
Myth: sound is the last ten percent of a video
Anyone who has watched a silent AI clip next to the same clip with a few footsteps and room tone knows it is more like half. Mirelo SFX 2.0 Video-to-Video is built for that moment. You upload a video, it listens with its eyes, and you get the same video back with a new synchronized sound effects track inside. One file in, one file out, ready to post.
Video-to-Video or Video-to-Audio?
Both models read the same clip and take nearly the same settings. The question is what you want in your hands afterwards.
- Video-to-Video (this one): a finished clip with the sound already mixed in. Choose it for social posts, previews and anything you will not edit further.
- [Video-to-Audio](/blog/mirelo-sfx-2-video-to-audio-guide): only the sound, as WAV, MP3, AAC or FLAC, for you to place on your own timeline. Choose it when you edit, mix with music or need to fix one hit by hand.
A practical difference: the audio version has output format and bitrate settings and this one does not, because the sound goes inside the video.
What goes in, what you can steer
The video is the only required field and it has to be at least one second long. Everything else is optional. The prompt (up to 2000 characters) describes the sounds you want; the negative prompt names what to keep out, such as music, voices or traffic; the duration slider runs from 1 to 600 seconds and starts at 10, capped by your video's length; the seed makes a take repeatable.
Seaside promenade in the evening: waves breaking softly in the distance, gulls, the rhythmic crunch of footsteps on sand, a light breeze in clothing. Open air, calm, no music.Busy night market: muffled crowd chatter without clear words, a sizzling pan, a metal ladle tapping a wok, a motorbike passing close and fading away.Keep the voices, drop the rest
Preserve speech is on by default, under advanced settings. If your clip has a person talking, their voice is kept and the effects are built around it, so you do not lose the dialogue under a new layer. Turn the switch off only if the speech in the clip is something you want replaced, and add voices to the negative prompt if you want none at all.
A good pairing with a video model
Silent video from a video model is where this shines. Make the picture with something like Kling V4.0 Text-to-Video, keep the best take and let this model dress it. If your clip is for a loop on a website, you may want to keep it silent after all; the video loops guide explains when.
Where it falls short
You get one mixed track, so you cannot lift the footsteps from the wind afterwards. If you need that control, take the audio-only sibling. It makes sound effects and atmosphere, not songs and not dialogue. Its output is only as long as your input, and an unclear picture, such as dark footage or a fast cut, gives it less to catch. More audio models are on the models page.
Questions
What does Mirelo SFX 2.0 Video-to-Video return?
Your video again, with a newly generated, synchronized sound effects track in it.
How is it different from Video-to-Audio?
Video-to-Audio returns the sound only, as a file you edit yourself in a choice of formats. Video-to-Video returns the finished clip with the sound already inside.
Does it need a prompt?
No. Only the video is required. A prompt and a negative prompt are optional and help when the picture alone does not say what the place sounds like.
