Sound · declare-lab/tangoflux
TangoFlux — fast, accurate sound from a description
TangoFlux generates sound effects from text quickly and accurately: a sound usually takes a few seconds. In limko it costs 0.2 ₽ per generation, and the duration is set from 1 to 10 seconds. The model is handy for quick tests, accents in video edits and interface sounds.
from 0.2 ₽ per sound
What makes TangoFlux good
- Fast sound from a description: usually a few seconds, so it is easy to try wordings one after another.
- Accuracy: the model is trained to pick the sound that matches the description more closely.
- Price: 0.2 ₽ per sound.
- Duration from 1 to 10 seconds: short accents, transitions, background fragments.
- AudioGen right next to it: give the same description to the second model and compare.
How TangoFlux matches the description
TangoFlux is built on the flow matching technique and fine-tuned on rankings: it was taught to choose, out of several sound variants, the one that matches the text more closely.
That is why the description should be specific: what is making the sound, what it is made of, where it is and how far it is from the listener.
Speed and price make the model handy for rough work: try several wordings in a row and keep the best one.
TangoFlux and AudioGen
Both models work in the limko sound effects generator with the same settings — description and duration. They differ in price, speed and how each one reads your text.
| What we compare | TangoFlux | AudioGen |
|---|---|---|
| Price per sound | from 0.2 ₽ | from 1.5 ₽ |
| Generation time | usually a few seconds | usually about 20 seconds |
| Sound duration | 1–10 seconds | 1–10 seconds |
| When to pick it | quick tests, lots of short sounds | a second variant for the same description |
Where to start: description examples
Write in your own language — the service will translate the description for the model. One or two phrases are enough: the sound source, material, setting.
Video transition
A short, sharp whoosh of air, like a fast cut between shots, clean sound with no background.
Notification
A soft double notification chime, high-pitched and pleasant, no echo.
Ambience
A quiet forest at night: crickets chirping, wind rustling the leaves, an owl hooting in the distance.
Action
A glass tumbler falls onto a tiled floor and shatters, the shards ringing and scattering.
Frequently asked about TangoFlux
How is TangoFlux different from AudioGen?
TangoFlux is cheaper and faster: 0.2 ₽ versus 1.5 ₽ per sound, and usually a few seconds versus twenty for AudioGen. The easiest way to see which model renders a particular sound better is to run both with the same description.
How much does one sound cost?
One generation costs 0.2 ₽. You pay with credits, no subscription, and are charged only for a finished result.
How long will the sound be?
From 1 to 10 seconds — the duration is set before the run.
Can I write the description in my own language?
Yes, the service will translate it for the model on its own.
Is TangoFlux suitable for music?
For songs and instrumentals, limko has a separate Music screen with the MiniMax Music model. Use TangoFlux for short sounds, accents and ambience.