1: \begin{abstract}

2: We introduce a wide family of stochastic processes that are obtained as

3: sums of self-similar localized ``waveforms'' with multiplicative intensity 

4: in the spirit of the Richardson cascade picture of turbulence. We establish the convergence 

5: and the minimum regularity of our construction. We show that

6: its continuous wavelet transform is characterized by stochastic self-similarity and multifractal scaling properties. 

7: This model constitutes a stationary, "grid free", extension of $\cal W$-cascades introduced in the past 

8: by Arneodo, Bacry and Muzy using wavelet orthogonal basis.

9: Moreover our approach generically provides multifractal 

10: random functions that are not invariant by time reversal and therefore is able to account for skewed multifractal models and for the so-called ``leverage effect".

11: In that respect, it can be well suited to providing 

12: synthetic turbulence models or to reproducing the main observed 

13: features of asset price fluctuations in financial markets.

14: 

15: 

16: \end{abstract}

17: