diff --git a/Cargo.toml b/Cargo.toml
index 9ecf5e2..7a5c1d9 100644
--- a/Cargo.toml
+++ b/Cargo.toml
@@ -4,6 +4,9 @@ version = "0.1.0"
 edition = "2024"
 
 [dependencies]
+image = "0.25.6"
+rand = "0.9.2"
+rand_chacha = "0.9.0"
 [lib]
 name = "worldgen"
 path = "src/lib.rs"
diff --git a/src/lib.rs b/src/lib.rs
index e1b5a28..2312b48 100644
--- a/src/lib.rs
+++ b/src/lib.rs
@@ -1,44 +1,119 @@
+use rand::{Rng, SeedableRng};
+use rand_chacha::ChaCha8Rng;
+
+#[derive(Clone, Copy, Debug, PartialEq)]
+
 pub struct Noise {
-    scale: f64,
-    amplitude: f64,
+    scale: f32,
+    amplitude: f32,
+    seed: u64,
 }
+#[derive(Clone, Copy, Debug, PartialEq)]
 pub struct Vector {
     x: f32,
     y: f32,
 }
 impl Noise {
-    pub fn new(scale: f64, amplitude: f64) -> Self {
-        Self { scale, amplitude }
+    pub fn new(scale: f32, amplitude: f32, seed: u64) -> Self {
+        assert!(scale != 0.0, "scale must be non-zero");
+        Self {
+            scale,
+            amplitude,
+            seed,
+        }
     }
-    pub fn get(&self, x: f64, z: f64) -> f64 {
+
+    pub fn get(&self, x: f32, z: f32) -> f32 {
         let xs = x / self.scale;
         let zs = z / self.scale;
         self.perlin(xs, zs) * self.amplitude
     }
-    fn perlin(&self, x: f64, z: f64) -> f64 {
-        // implement Perlin noise here (then simplex because it's harder)
-        todo!()
+
+    fn perlin(&self, x: f32, y: f32) -> f32 {
+        // 1) cellule et coords locales
+        let xi = x.floor() as i32;
+        let yi = y.floor() as i32;
+        let xf = x - xi as f32; // ∈ [0,1)
+        let yf = y - yi as f32; // ∈ [0,1)
+
+        // 2) gradients aux 4 coins (via tes helpers)
+        let g00 = gradient_at(xi, yi, self.seed);
+        let g10 = gradient_at(xi + 1, yi, self.seed);
+        let g01 = gradient_at(xi, yi + 1, self.seed);
+        let g11 = gradient_at(xi + 1, yi + 1, self.seed);
+
+        // 3) vecteurs locaux depuis chaque coin
+        let v00 = Vector { x: xf, y: yf };
+        let v10 = Vector { x: xf - 1.0, y: yf };
+        let v01 = Vector { x: xf, y: yf - 1.0 };
+        let v11 = Vector {
+            x: xf - 1.0,
+            y: yf - 1.0,
+        };
+
+        // 4) produits scalaires
+        let n00 = dot_product(&g00, &v00);
+        let n10 = dot_product(&g10, &v10);
+        let n01 = dot_product(&g01, &v01);
+        let n11 = dot_product(&g11, &v11);
+
+        // 5) interpolation quintique
+        let u = fade(xf);
+        let v = fade(yf);
+
+        let nx0 = linear_interpolation(n00, n10, u); // sur x, bas
+        let nx1 = linear_interpolation(n01, n11, u); // sur x, haut
+        linear_interpolation(nx0, nx1, v)
     }
 }
-fn dot_product(v1: Vector, v2: Vector) -> f32 {
-    // Calculate the dot product between v1 and v2 using their coordinates.  the result->f32.
+
+fn hash2(ix: i32, iy: i32, seed: u64) -> u64 {
+    let mut h = seed
+        ^ (ix as u64).wrapping_mul(0x9E3779B97F4A7C15)
+        ^ (iy as u64).wrapping_mul(0xC2B2AE3D27D4EB4F);
+    h ^= h >> 33;
+    h = h.wrapping_mul(0xFF51AFD7ED558CCD);
+    h ^= h >> 33;
+    h = h.wrapping_mul(0xC4CEB9FE1A85EC53);
+    h ^= h >> 33;
+    h
+}
+
+// Gradient unitaire au coin (réutilise `normalize`)
+fn gradient_at(ix: i32, iy: i32, seed: u64) -> Vector {
+    let h = hash2(ix, iy, seed);
+    let mut rng = ChaCha8Rng::seed_from_u64(h);
+    // angle uniforme → vecteur unitaire; pas besoin d’appeler normalize ensuite
+    let angle: f32 = rng.gen_range(0.0..std::f32::consts::TAU);
+    Vector {
+        x: angle.cos(),
+        y: angle.sin(),
+    }
+}
+#[allow(unused)]
+fn dot_product(v1: &Vector, v2: &Vector) -> f32 {
     v1.x * v2.x + v1.y * v2.y
 }
+#[allow(unused)]
 fn calculate_norm(v1: &Vector) -> f32 {
-    // Calculate the norm of a vector using it's coordinates.  the result -> f32.
-    (v1.x.powi(2) + v1.y.powi(2)).sqrt()
+    v1.x.hypot(v1.y)
 }
+#[allow(unused)]
 fn normalize(v1: &Vector) -> Vector {
-    // This function aim that every vector created randomly has the same norm (1).
+    let n = calculate_norm(v1);
+    if n == 0.0 {
+        Vector { x: 0.0, y: 0.0 };
+    }
     Vector {
-        x: v1.x / calculate_norm(v1),
-        y: v1.y / calculate_norm(v1),
+        x: v1.x / n,
+        y: v1.y / n,
     }
 }
+#[allow(unused)]
 fn fade(x: f32) -> f32 {
-    // This is a mathematic equation, it take a value and return the result of the value by that
-    // equation. That equation if derivate should be easy when close to 0 and 1 and hard in the
-    // middle. Take a f32 value return f32.Both value are between 0 and 1
+    // Quintic fade function used in Perlin noise.
+    // Maps x ∈ [0,1] to a smooth curve with zero first derivative at 0 and 1.
+    // Formula: 6x⁵ − 15x⁴ + 10x³. Produces smooth interpolation weights.
     6_f32 * x.powi(5) - 15_f32 * x.powi(4) + 10_f32 * x.powi(3)
 }
 /// Linearly interpolates between two scalar values.
@@ -53,45 +128,291 @@ fn fade(x: f32) -> f32 {
 /// - if `t = 0.0`, yields `a`  
 /// - if `t = 1.0`, yields `b`  
 /// - otherwise `a + t * (b - a)`
+#[allow(unused)]
 fn linear_interpolation(a: f32, b: f32, t: f32) -> f32 {
     a + t * (b - a)
 }
+
+fn identify_cell(x: f32, y: f32) -> (i32, i32) {
+    // Use this function to identify the cell of the point that we're trying to calculate
+    (x.floor() as i32, y.floor() as i32)
+}
+
+fn get_random_float_number(seed: i64) -> f32 {
+    let mut rng = ChaCha8Rng::seed_from_u64(seed as u64);
+    {
+        let this = &mut rng;
+        let range = -1.0..1.0;
+        this.random_range(range)
+    }
+}
+fn get_gradient(seed: i64) -> Vector {
+    normalize(&Vector {
+        x: get_random_float_number(seed),
+        y: get_random_float_number(seed + 1), // éviter x == y
+    })
+}
 #[cfg(test)]
+
 mod tests {
-    use crate::{Vector, calculate_norm, dot_product, linear_interpolation, normalize};
+    fn deriv(f: fn(f32) -> f32, x: f32) -> f32 {
+        let h = 1e-3;
+        (f(x + h) - f(x - h)) / (2.0 * h)
+    }
+
+    use super::*;
+
+    fn approx_eq(a: f32, b: f32, eps: f32) -> bool {
+        (a - b).abs() <= eps
+    }
+
     #[test]
     fn test_dot_product() {
-        assert_eq!(
-            dot_product(Vector { x: 1.0, y: 0.0 }, Vector { x: 0.0, y: 1.0 }),
-            0.0
-        );
-        assert_eq!(
-            dot_product(Vector { x: 1.0, y: 0.5 }, Vector { x: 0.2, y: 1.0 }),
+        assert!(approx_eq(
+            dot_product(&Vector { x: 1.0, y: 0.0 }, &Vector { x: 0.0, y: 1.0 }),
+            0.0,
+            1e-6
+        ));
+        assert!(approx_eq(
+            dot_product(&Vector { x: 1.0, y: 0.5 }, &Vector { x: 0.2, y: 1.0 }),
             0.7,
-        );
-        assert_eq!(
-            dot_product(Vector { x: 1.0, y: 0.5 }, Vector { x: -0.2, y: -1.0 }),
+            1e-6
+        ));
+        assert!(approx_eq(
+            dot_product(&Vector { x: 1.0, y: 0.5 }, &Vector { x: -0.2, y: -1.0 }),
             -0.7,
-        );
+            1e-6
+        ));
     }
+
     #[test]
     fn test_calculate_norm() {
-        assert_eq!(calculate_norm(&Vector { x: 0.5, y: 0.5 }), 0.5_f32.sqrt());
-        assert_eq!(calculate_norm(&Vector { x: 0.7, y: 0.3 }), 0.58_f32.sqrt());
-        assert_eq!(
+        assert!(approx_eq(
+            calculate_norm(&Vector { x: 0.5, y: 0.5 }),
+            0.5_f32.sqrt(),
+            1e-6
+        ));
+        assert!(approx_eq(
+            calculate_norm(&Vector { x: 0.7, y: 0.3 }),
+            0.58_f32.sqrt(),
+            1e-6
+        ));
+        assert!(approx_eq(
             calculate_norm(&Vector { x: -0.7, y: -0.3 }),
-            calculate_norm(&Vector { x: 0.7, y: 0.3 })
-        );
+            calculate_norm(&Vector { x: 0.7, y: 0.3 }),
+            1e-6
+        ));
     }
+
     #[test]
     fn test_normalize() {
         let v1 = Vector { x: 0.5, y: 0.5 };
-        assert_eq!(calculate_norm(&normalize(&v1)).round(), 1_f32);
+        assert!(approx_eq(calculate_norm(&normalize(&v1)), 1.0, 1e-6));
+        let z = Vector { x: 0.0, y: 0.0 };
+        assert_eq!(normalize(&z), z);
     }
-    //fn test_fade() {todo!()}
+
     #[test]
     fn test_linear_interpolation() {
-        assert_eq!(linear_interpolation(0.0, 10.0, 0.25), 2.5);
-        assert_eq!(linear_interpolation(5.0, 15.0, 0.75), 12.5);
+        assert!(approx_eq(linear_interpolation(0.0, 10.0, 0.25), 2.5, 1e-6));
+        assert!(approx_eq(linear_interpolation(5.0, 15.0, 0.75), 12.5, 1e-6));
+    }
+    #[test]
+    fn dot_product_properties() {
+        let a = Vector { x: 0.3, y: -0.7 };
+        let b = Vector { x: -0.2, y: 1.1 };
+        assert!((dot_product(&a, &b) - dot_product(&b, &a)).abs() < 1e-6);
+
+        let k = 2.5;
+        let kb = Vector {
+            x: k * b.x,
+            y: k * b.y,
+        };
+        assert!((dot_product(&a, &kb) - k * dot_product(&a, &b)).abs() < 1e-6);
+
+        let p = Vector { x: 2.0, y: 4.0 };
+        assert!(dot_product(&p, &p) >= 0.0);
+    }
+    #[test]
+    fn norm_properties() {
+        let a = Vector { x: -0.7, y: 0.3 };
+        let b = Vector { x: 0.4, y: -1.2 };
+        assert!(
+            (calculate_norm(&a)
+                - calculate_norm(&Vector {
+                    x: a.x.abs(),
+                    y: a.y.abs()
+                }))
+            .abs()
+                < 1e-6
+        );
+
+        let dab = dot_product(&a, &b).abs();
+        assert!(dab <= calculate_norm(&a) * calculate_norm(&b) + 1e-6);
+
+        let a_plus_b = Vector {
+            x: a.x + b.x,
+            y: a.y + b.y,
+        };
+        assert!(calculate_norm(&a_plus_b) <= calculate_norm(&a) + calculate_norm(&b) + 1e-6);
+    }
+    #[test]
+    fn normalize_properties() {
+        let v = Vector { x: -0.5, y: 0.25 };
+        let u = normalize(&v);
+        assert!((calculate_norm(&u) - 1.0).abs() < 1e-6);
+
+        let uu = normalize(&u);
+        assert!((uu.x - u.x).abs() < 1e-6 && (uu.y - u.y).abs() < 1e-6);
+
+        if calculate_norm(&v) > 0.0 {
+            let cross = v.x * u.y - v.y * u.x;
+            assert!(cross.abs() < 1e-6);
+        }
+
+        let z = Vector { x: 0.0, y: 0.0 };
+        assert_eq!(normalize(&z), z);
+    }
+    #[test]
+    fn lerp_properties() {
+        assert!((linear_interpolation(2.0, 8.0, 0.0) - 2.0).abs() < 1e-6);
+        assert!((linear_interpolation(2.0, 8.0, 1.0) - 8.0).abs() < 1e-6);
+        assert!((linear_interpolation(0.0, 10.0, -0.5) + 5.0).abs() < 1e-6);
+
+        let (a, b, t) = (3.0f32, 7.0f32, 0.3f32);
+        assert!((linear_interpolation(a, b, t) - (a + t * (b - a))).abs() < 1e-6);
+    }
+
+    #[test]
+    fn fade_constraints() {
+        assert!((fade(0.0) - 0.0).abs() < 1e-6);
+        assert!((fade(1.0) - 1.0).abs() < 1e-6);
+        assert!(deriv(fade, 0.0).abs() < 1e-2);
+        assert!(deriv(fade, 1.0).abs() < 1e-2);
+
+        let mut prev = fade(0.0);
+        for i in 1..=100 {
+            let x = i as f32 / 100.0;
+            let y = fade(x);
+            assert!(y >= prev - 1e-6);
+            prev = y;
+        }
+    }
+    #[test]
+    fn perlin_is_deterministic() {
+        let n = Noise::new(8.0, 1.0, 42);
+        let a = n.get(12.345, -6.789);
+        let b = n.get(12.345, -6.789);
+        assert!((a - b).abs() < 1e-6);
+    }
+    #[test]
+    fn perlin_bounds_reasonable() {
+        let n = Noise::new(4.0, 1.0, 7);
+        for i in -5..=5 {
+            for j in -5..=5 {
+                let v = n.get(i as f32 * 0.37, j as f32 * 0.41);
+                assert!(v.is_finite());
+                assert!(v >= -1.5 && v <= 1.5);
+            }
+        }
+    }
+    #[test]
+    fn perlin_continuity_c1_local() {
+        let n = Noise::new(5.0, 1.0, 1);
+        let x = 2.3f32;
+        let y = -1.7f32;
+        let h = 1e-3f32;
+        let c0 = n.get(x, y);
+        let cx = n.get(x + h, y);
+        let cy = n.get(x, y + h);
+        assert!((cx - c0).abs() < 0.1 && (cy - c0).abs() < 0.1);
+    }
+    #[test]
+    fn test_perlin() -> () {
+        let n = Noise::new(8.0, 1.0, 42);
+        let h = 1e-3;
+        let p = n.get(3.2, 4.7);
+        assert!((n.get(3.2 + h, 4.7) - p).abs() < 0.1);
+        assert!((n.get(3.2, 4.7 + h) - p).abs() < 0.1);
+        assert!((n.get(4.0 - h, 4.0) - n.get(4.0 + h, 4.0)).abs() < 0.2);
+    }
+}
+// The following code is GPT-5 generated and allows you to take a look at the noise
+#[cfg(test)]
+mod perlin_tests {
+    use super::*;
+
+    #[test]
+    fn perlin_deterministic() {
+        let n = Noise::new(32.0, 1.0, 123);
+        let a = n.get(12.34, -5.67);
+        let b = n.get(12.34, -5.67);
+        assert!((a - b).abs() < 1e-6);
+    }
+
+    #[test]
+    fn perlin_reasonable_bounds() {
+        let n = Noise::new(16.0, 1.0, 7);
+        for i in -4..=4 {
+            for j in -4..=4 {
+                let v = n.get(i as f32 * 0.37, j as f32 * 0.41);
+                assert!(v.is_finite());
+                assert!(v >= -1.5 && v <= 1.5);
+            }
+        }
+    }
+
+    #[test]
+    fn perlin_local_continuity() {
+        let n = Noise::new(24.0, 1.0, 42);
+        let (x, y) = (3.2f32, -4.7f32);
+        let h = 1e-3f32;
+        let c0 = n.get(x, y);
+        assert!((n.get(x + h, y) - c0).abs() < 0.1);
+        assert!((n.get(x, y + h) - c0).abs() < 0.1);
+    }
+}
+#[cfg(test)]
+mod viz {
+    use super::*;
+    use image::{GrayImage, Luma};
+
+    // try it with `cargo test -- --ignored`
+    #[test]
+    #[ignore]
+    fn dump_perlin_png() {
+        let noise = Noise::new(64.0, 10.0, 42);
+
+        let (w, h) = (512u32, 512u32);
+        let mut img: GrayImage = GrayImage::new(w, h);
+
+        // Simple Perlin. Pour un rendu plus riche, active la section "octaves" plus bas.
+        for y in 0..h {
+            for x in 0..w {
+                let fx = x as f32;
+                let fy = y as f32;
+
+                // --- Perlin simple ---
+                let mut v = noise.get(fx, fy);
+
+                let mut v = 0.0f32;
+                let mut amp = 1.0f32;
+                let mut freq = 1.0f32;
+                for _ in 0..5 {
+                    v += amp * noise.perlin(fx / noise.scale * freq, fy / noise.scale * freq);
+                    amp *= 0.5; // persistance
+                    freq *= 2.0; // lacunarité
+                }
+
+                // Normalisation grossière vers [0,1] puis 8 bits
+                let n01 = (v * 0.5 + 0.5).clamp(0.0, 1.0);
+                let p = (n01 * 255.0).round() as u8;
+
+                img.put_pixel(x, y, Luma([p]));
+            }
+        }
+
+        std::fs::create_dir_all("target").ok();
+        img.save("perlin.png").expect("save png");
     }
 }