diff --git a/bind/v4k.lua b/bind/v4k.lua
index 5f7377b..fec03ff 100644
--- a/bind/v4k.lua
+++ b/bind/v4k.lua
@@ -3015,7 +3015,7 @@ unsigned shader;
 } lightmap_t;
  lightmap_t lightmap(int hmsize , float near, float far, vec3 color , int passes , float threshold , float distmod );
  void       lightmap_setup(lightmap_t *lm, int w, int h);
- void          lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void *userdata);
+ void          lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void (*progressupdate)(float progress), void *userdata);
  void       lightmap_destroy(lightmap_t *lm);
 typedef struct skybox_t {
 handle program;
diff --git a/demos/99-lmap.c b/demos/99-lmap.c
index 9b5c05c..ae8f534 100644
--- a/demos/99-lmap.c
+++ b/demos/99-lmap.c
@@ -10,9 +10,20 @@ void bakedrawmodel(lightmap_t *lm, model_t *m, float *view, float *proj, void *u
     model_render(litm, proj, view, litm.pivot, lm->shader);
 }
 
+void progressupdate(float progress) {
+    static double lastUpdateTime = 0.0;
+    double time = time_ss();
+    if (time - lastUpdateTime > 1.0) {
+        lastUpdateTime = time;
+        PRINTF("progress: %.02f%%", progress*100);
+    }
+    // window_swap();
+}
+
 int main()
 {
-    window_create(0.5, WINDOW_VSYNC_DISABLED);
+    window_create(0.5, 0);
+    window_title(__FILE__);
     camera_t cam = camera();
     sky = skybox(0, 0); skybox_mie_calc_sh(&sky, 2.0f);
     model_t mdl = model("gazebo.obj", 0);
@@ -31,11 +42,12 @@ int main()
     texture_t emission = texture_create(1,1,4,emissive,TEXTURE_LINEAR);
     model_set_texture(litm, emission);
 
-    lightmap_t baker = lightmap(64, 0.001, 1000, vec3(0,0,0), 2, 0.01, 0.0);
-    lightmap_setup(&baker, 640, 640);
+    lightmap_t baker = lightmap(64, 0.01, 100, vec3(0,0,0), 2, 0.01, 0.0);
+    lightmap_setup(&baker, 512, 512);
     array_push(baker.models, &mdl);
 
-    window_title("AO Baking demo");
+    bool do_bake=0;
+    int b=1;
 
     while (window_swap() && !input(KEY_ESC)) {
         bool active = ui_active() || ui_hover() || gizmo_active() ? false : input(MOUSE_L) || input(MOUSE_M) || input(MOUSE_R);
@@ -53,11 +65,16 @@ int main()
         if( ui_panel("Lightmapper", PANEL_OPEN) ) {
             ui_label2("Freecam", "Mouse + W/A/S/D/E/Q keys");
             ui_label("Warning " ICON_MD_WARNING "@This will take a few seconds and bake a lightmap illuminated by: The mesh itself (initially black) + A white sky (1.0f, 1.0f, 1.0f)");
-            int b=2;
+            ui_int("Bounces", &b);
             if( ui_button(va("Bake %d light bounce", b)) ) {
-                lightmap_bake(&baker, b, bakedrawmodel, 0);
+                do_bake=1;
             }
             ui_panel_end();
         }
+
+        if (do_bake) {
+            do_bake=0;
+            lightmap_bake(&baker, b, bakedrawmodel, progressupdate, 0);
+        }
     }
 }
diff --git a/engine/joint/v4k.h b/engine/joint/v4k.h
index d1381e6..239fff3 100644
--- a/engine/joint/v4k.h
+++ b/engine/joint/v4k.h
@@ -17589,7 +17589,7 @@ typedef struct lightmap_t {
 
 API lightmap_t lightmap(int hmsize /*64*/, float near, float far, vec3 color /*1,1,1 for AO*/, int passes /*2*/, float threshold /*0.01f*/, float distmod /*0.0f*/);
 API void       lightmap_setup(lightmap_t *lm, int w, int h);
-API void          lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void *userdata);
+API void          lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void (*progressupdate)(float progress), void *userdata);
 API void       lightmap_destroy(lightmap_t *lm);
 
 // -----------------------------------------------------------------------------
@@ -348237,6 +348237,7 @@ struct lm_context
 		{
 			GLuint texture;
 			lm_ivec2 writePosition;
+			int	width, height;
 			lm_ivec2 *toLightmapLocation;
 		} storage;
 	} hemisphere;
@@ -348497,10 +348498,72 @@ static lm_bool lm_findNextConservativeTriangleRasterizerPosition(lm_context *ctx
 	return lm_findFirstConservativeTriangleRasterizerPosition(ctx);
 }
 
-static void lm_integrateHemisphereBatch(lm_context *ctx)
+static void lm_writeResultsToLightmap(lm_context* ctx)
+{
+	// do the GPU->CPU transfer of downsampled hemispheres
+	float* hemi = (float*)LM_CALLOC(ctx->hemisphere.storage.width * ctx->hemisphere.storage.height, 4 * sizeof(float));
+	glBindTexture(GL_TEXTURE_2D, ctx->hemisphere.storage.texture);
+	glGetTexImage(GL_TEXTURE_2D, 0, GL_RGBA, GL_FLOAT, hemi);
+
+	// write results to lightmap texture
+	for (int y = 0; y < ctx->hemisphere.storage.writePosition.y + (int)ctx->hemisphere.fbHemiCountY; y++)
+	{
+		for (int x = 0; x < ctx->hemisphere.storage.width; x++)
+		{
+			int index = y * ctx->hemisphere.storage.width + x;
+			lm_ivec2 lmUV = ctx->hemisphere.storage.toLightmapLocation[index];
+			if (lmUV.x >= 0)
+			{
+				float* c = hemi + index * 4;
+				float validity = c[3];
+				float* lm = ctx->lightmap.data + (lmUV.y * ctx->lightmap.width + lmUV.x) * ctx->lightmap.channels;
+				if (!lm[0] && validity > 0.9)
+				{
+					float scale = 1.0f / validity;
+					switch (ctx->lightmap.channels)
+					{
+					case 1:
+						lm[0] = lm_maxf((c[0] + c[1] + c[2]) * scale / 3.0f, FLT_MIN);
+						break;
+					case 2:
+						lm[0] = lm_maxf((c[0] + c[1] + c[2]) * scale / 3.0f, FLT_MIN);
+						lm[1] = 1.0f; // do we want to support this format?
+						break;
+					case 3:
+						lm[0] = lm_maxf(c[0] * scale, FLT_MIN);
+						lm[1] = lm_maxf(c[1] * scale, FLT_MIN);
+						lm[2] = lm_maxf(c[2] * scale, FLT_MIN);
+						break;
+					case 4:
+						lm[0] = lm_maxf(c[0] * scale, FLT_MIN);
+						lm[1] = lm_maxf(c[1] * scale, FLT_MIN);
+						lm[2] = lm_maxf(c[2] * scale, FLT_MIN);
+						lm[3] = 1.0f;
+						break;
+					default:
+						assert(LM_FALSE);
+						break;
+					}
+
+#ifdef LM_DEBUG_INTERPOLATION
+					// set sampled pixel to red in debug output
+					ctx->lightmap.debug[(lmUV.y * ctx->lightmap.width + lmUV.x) * 3 + 0] = 255;
+#endif
+				}
+			}
+			ctx->hemisphere.storage.toLightmapLocation[index].x = -1; // reset
+		}
+	}
+
+	LM_FREE(hemi);
+	ctx->hemisphere.storage.writePosition = lm_i2(0, 0);
+}
+
+
+static lm_bool lm_integrateHemisphereBatch(lm_context *ctx)
 {
 	if (!ctx->hemisphere.fbHemiIndex)
-		return; // nothing to do
+		return LM_FALSE; // nothing to do
 
 	glDisable(GL_DEPTH_TEST);
 	glBindVertexArray(ctx->hemisphere.vao);
@@ -348568,83 +348631,30 @@ static void lm_integrateHemisphereBatch(lm_context *ctx)
 		{
 			int sx = ctx->hemisphere.storage.writePosition.x + x;
 			unsigned int hemiIndex = y * ctx->hemisphere.fbHemiCountX + x;
-			if (hemiIndex >= ctx->hemisphere.fbHemiIndex)
-				ctx->hemisphere.storage.toLightmapLocation[sy * ctx->lightmap.width + sx] = lm_i2(-1, -1);
-			else
-				ctx->hemisphere.storage.toLightmapLocation[sy * ctx->lightmap.width + sx] = ctx->hemisphere.fbHemiToLightmapLocation[hemiIndex];
+			ctx->hemisphere.storage.toLightmapLocation[sy * ctx->hemisphere.storage.width + sx] =
+				(hemiIndex >= ctx->hemisphere.fbHemiIndex) ? 
+				lm_i2(-1, -1) :
+				ctx->hemisphere.fbHemiToLightmapLocation[hemiIndex];
 		}
 	}
 
+	lm_bool needWrite = LM_TRUE;
 	// advance storage texture write position
 	ctx->hemisphere.storage.writePosition.x += ctx->hemisphere.fbHemiCountX;
-	if (ctx->hemisphere.storage.writePosition.x + (int)ctx->hemisphere.fbHemiCountX > ctx->lightmap.width)
+	if (ctx->hemisphere.storage.writePosition.x + (int)ctx->hemisphere.fbHemiCountX > ctx->hemisphere.storage.width)
 	{
 		ctx->hemisphere.storage.writePosition.x = 0;
-		ctx->hemisphere.storage.writePosition.y += ctx->hemisphere.fbHemiCountY;
-		assert(ctx->hemisphere.storage.writePosition.y + (int)ctx->hemisphere.fbHemiCountY < ctx->lightmap.height);
+		// storage is full
+		if (ctx->hemisphere.storage.writePosition.y + (int)ctx->hemisphere.fbHemiCountY >= ctx->hemisphere.storage.height) {
+			lm_writeResultsToLightmap(ctx); // read storage data from gpu memory and write it to the lightmap
+			needWrite = LM_FALSE;
+		} else {
+			ctx->hemisphere.storage.writePosition.y += ctx->hemisphere.fbHemiCountY;
+		}
 	}
 
 	ctx->hemisphere.fbHemiIndex = 0;
-}
-
-static void lm_writeResultsToLightmap(lm_context *ctx)
-{
-	// do the GPU->CPU transfer of downsampled hemispheres
-	float *hemi = (float*)LM_CALLOC(ctx->lightmap.width * ctx->lightmap.height, 4 * sizeof(float));
-	glBindTexture(GL_TEXTURE_2D, ctx->hemisphere.storage.texture);
-	glGetTexImage(GL_TEXTURE_2D, 0, GL_RGBA, GL_FLOAT, hemi);
-
-	// write results to lightmap texture
-	for (int y = 0; y < ctx->hemisphere.storage.writePosition.y + (int)ctx->hemisphere.fbHemiCountY; y++)
-	{
-		for (int x = 0; x < ctx->lightmap.width; x++)
-		{
-			lm_ivec2 lmUV = ctx->hemisphere.storage.toLightmapLocation[y * ctx->lightmap.width + x];
-			if (lmUV.x >= 0)
-			{
-				float *c = hemi + (y * ctx->lightmap.width + x) * 4;
-				float validity = c[3];
-				float *lm = ctx->lightmap.data + (lmUV.y * ctx->lightmap.width + lmUV.x) * ctx->lightmap.channels;
-				if (!lm[0] && validity > 0.9)
-				{
-					float scale = 1.0f / validity;
-					switch (ctx->lightmap.channels)
-					{
-					case 1:
-						lm[0] = lm_maxf((c[0] + c[1] + c[2]) * scale / 3.0f, FLT_MIN);
-						break;
-					case 2:
-						lm[0] = lm_maxf((c[0] + c[1] + c[2]) * scale / 3.0f, FLT_MIN);
-						lm[1] = 1.0f; // do we want to support this format?
-						break;
-					case 3:
-						lm[0] = lm_maxf(c[0] * scale, FLT_MIN);
-						lm[1] = lm_maxf(c[1] * scale, FLT_MIN);
-						lm[2] = lm_maxf(c[2] * scale, FLT_MIN);
-						break;
-					case 4:
-						lm[0] = lm_maxf(c[0] * scale, FLT_MIN);
-						lm[1] = lm_maxf(c[1] * scale, FLT_MIN);
-						lm[2] = lm_maxf(c[2] * scale, FLT_MIN);
-						lm[3] = 1.0f;
-						break;
-					default:
-						assert(LM_FALSE);
-						break;
-					}
-
-#ifdef LM_DEBUG_INTERPOLATION
-					// set sampled pixel to red in debug output
-					ctx->lightmap.debug[(lmUV.y * ctx->lightmap.width + lmUV.x) * 3 + 0] = 255;
-#endif
-				}
-			}
-			ctx->hemisphere.storage.toLightmapLocation[y * ctx->lightmap.width + x].x = -1; // reset
-		}
-	}
-
-	LM_FREE(hemi);
-	ctx->hemisphere.storage.writePosition = lm_i2(0, 0);
+	return needWrite;
 }
 
 static void lm_setView(
@@ -348921,6 +348931,7 @@ static void lm_setMeshPosition(lm_context *ctx, unsigned int indicesTriangleBase
 		// TODO: signed formats
 		case LM_FLOAT: {
 			n = *(const lm_vec3*)nPtr;
+			n = lm_normalize3(lm_transformNormal(ctx->mesh.normalMatrix, n));
 		} break;
 		case LM_NONE: {
 			n = flatNormal;
@@ -348929,7 +348940,7 @@ static void lm_setMeshPosition(lm_context *ctx, unsigned int indicesTriangleBase
 			assert(LM_FALSE);
 		} break;
 		}
-		ctx->meshPosition.triangle.n[i] = lm_normalize3(lm_transformNormal(ctx->mesh.normalMatrix, n));
+		ctx->meshPosition.triangle.n[i] = n;
 	}
 
 	// calculate area of interest (on lightmap) for conservative rasterization
@@ -349286,16 +349297,13 @@ void lmSetHemisphereWeights(lm_context *ctx, lm_weight_func f, void *userdata)
 	LM_FREE(weights);
 }
 
-void lmSetTargetLightmap(lm_context *ctx, float *outLightmap, int w, int h, int c)
+static void lm_initStorage(lm_context *ctx, int w, int h)
 {
-	ctx->lightmap.data = outLightmap;
-	ctx->lightmap.width = w;
-	ctx->lightmap.height = h;
-	ctx->lightmap.channels = c;
-
 	// allocate storage texture
 	if (!ctx->hemisphere.storage.texture)
 		glGenTextures(1, &ctx->hemisphere.storage.texture);
+	ctx->hemisphere.storage.width = w;
+	ctx->hemisphere.storage.height = h;
 	glBindTexture(GL_TEXTURE_2D, ctx->hemisphere.storage.texture);
 	glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
 	glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
@@ -349310,6 +349318,18 @@ void lmSetTargetLightmap(lm_context *ctx, float *outLightmap, int w, int h, int
 	// invalidate all positions
 	for (int i = 0; i < w * h; i++)
 		ctx->hemisphere.storage.toLightmapLocation[i].x = -1;
+}
+
+void lmSetTargetLightmap(lm_context *ctx, float *outLightmap, int w, int h, int c)
+{
+	ctx->lightmap.data = outLightmap;
+	ctx->lightmap.width = w;
+	ctx->lightmap.height = h;
+	ctx->lightmap.channels = c;
+
+	unsigned int sw = w > ctx->hemisphere.fbHemiCountX ? w : ctx->hemisphere.fbHemiCountX;
+	unsigned int sh = h > ctx->hemisphere.fbHemiCountY ? h : ctx->hemisphere.fbHemiCountY;
+	lm_initStorage(ctx, sw, sh);
 
 #ifdef LM_DEBUG_INTERPOLATION
 	if (ctx->lightmap.debug)
@@ -349363,8 +349383,8 @@ lm_bool lmBegin(lm_context *ctx, int* outViewport4, float* outView4x4, float* ou
 			}
 			else
 			{ // ...and there are no triangles left: finish
-				lm_integrateHemisphereBatch(ctx); // integrate and store last batch
-				lm_writeResultsToLightmap(ctx); // read storage data from gpu memory and write it to the lightmap
+				if (lm_integrateHemisphereBatch(ctx)) // integrate and store last batch
+					lm_writeResultsToLightmap(ctx); // read storage data from gpu memory and write it to the lightmap
 
 				if (++ctx->meshPosition.pass == ctx->meshPosition.passCount)
 				{
@@ -349639,8 +349659,7 @@ lm_bool lmImageSaveTGAf(const char *filename, const float *image, int w, int h,
 	return success;
 }
 
-#endif // LIGHTMAPPER_IMPLEMENTATION
-#line 0
+#endif // LIGHTMAPPER_IMPLEMENTATION#line 0
 
 #endif // V4K_3RD
 /* game framework.
@@ -370328,7 +370347,7 @@ void lightmap_setup(lightmap_t *lm, int w, int h) {
     lm->h = h;
 }
 
-void lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void *userdata) {
+void lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void (*progressupdate)(float progress), void *userdata) {
     ASSERT(lm->ready);
     // @fixme: use xatlas to UV pack all models, update their UV1 and upload them to GPU.
 
@@ -370371,6 +370390,7 @@ void lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm
                 // render to lightmapper framebuffer
                 glViewport(vp[0], vp[1], vp[2], vp[3]);
                 drawscene(lm, m, view, projection, userdata);
+                if (progressupdate) progressupdate(lmProgress(lm->ctx));
                 lmEnd(lm->ctx);
             }
         }
diff --git a/engine/split/v4k_render.c b/engine/split/v4k_render.c
index c27d397..99278d1 100644
--- a/engine/split/v4k_render.c
+++ b/engine/split/v4k_render.c
@@ -3767,7 +3767,7 @@ void lightmap_setup(lightmap_t *lm, int w, int h) {
     lm->h = h;
 }
 
-void lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void *userdata) {
+void lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void (*progressupdate)(float progress), void *userdata) {
     ASSERT(lm->ready);
     // @fixme: use xatlas to UV pack all models, update their UV1 and upload them to GPU.
 
@@ -3810,6 +3810,7 @@ void lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm
                 // render to lightmapper framebuffer
                 glViewport(vp[0], vp[1], vp[2], vp[3]);
                 drawscene(lm, m, view, projection, userdata);
+                if (progressupdate) progressupdate(lmProgress(lm->ctx));
                 lmEnd(lm->ctx);
             }
         }
diff --git a/engine/split/v4k_render.h b/engine/split/v4k_render.h
index a4b6518..b2c7462 100644
--- a/engine/split/v4k_render.h
+++ b/engine/split/v4k_render.h
@@ -598,7 +598,7 @@ typedef struct lightmap_t {
 
 API lightmap_t lightmap(int hmsize /*64*/, float near, float far, vec3 color /*1,1,1 for AO*/, int passes /*2*/, float threshold /*0.01f*/, float distmod /*0.0f*/);
 API void       lightmap_setup(lightmap_t *lm, int w, int h);
-API void          lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void *userdata);
+API void          lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void (*progressupdate)(float progress), void *userdata);
 API void       lightmap_destroy(lightmap_t *lm);
 
 // -----------------------------------------------------------------------------
diff --git a/engine/v4k.c b/engine/v4k.c
index a34e597..3291064 100644
--- a/engine/v4k.c
+++ b/engine/v4k.c
@@ -20683,7 +20683,7 @@ void lightmap_setup(lightmap_t *lm, int w, int h) {
     lm->h = h;
 }
 
-void lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void *userdata) {
+void lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void (*progressupdate)(float progress), void *userdata) {
     ASSERT(lm->ready);
     // @fixme: use xatlas to UV pack all models, update their UV1 and upload them to GPU.
 
@@ -20726,6 +20726,7 @@ void lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm
                 // render to lightmapper framebuffer
                 glViewport(vp[0], vp[1], vp[2], vp[3]);
                 drawscene(lm, m, view, projection, userdata);
+                if (progressupdate) progressupdate(lmProgress(lm->ctx));
                 lmEnd(lm->ctx);
             }
         }
diff --git a/engine/v4k.h b/engine/v4k.h
index b463ec7..7d4e3e2 100644
--- a/engine/v4k.h
+++ b/engine/v4k.h
@@ -3656,7 +3656,7 @@ typedef struct lightmap_t {
 
 API lightmap_t lightmap(int hmsize /*64*/, float near, float far, vec3 color /*1,1,1 for AO*/, int passes /*2*/, float threshold /*0.01f*/, float distmod /*0.0f*/);
 API void       lightmap_setup(lightmap_t *lm, int w, int h);
-API void          lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void *userdata);
+API void          lightmap_bake(lightmap_t *lm, int bounces, void (*drawscene)(lightmap_t *lm, model_t *m, float *view, float *proj, void *userdata), void (*progressupdate)(float progress), void *userdata);
 API void       lightmap_destroy(lightmap_t *lm);
 
 // -----------------------------------------------------------------------------