diff --git a/const_fmt/format.h b/const_fmt/format.h
index f84572c..cc8b0c3 100644
--- a/const_fmt/format.h
+++ b/const_fmt/format.h
@@ -47,7 +47,7 @@ constexpr inline void format_arg(char* dest, arg_t arg) {
 
 template <fmt_data_t fmt_data, std::floating_point arg_t>
 constexpr inline void format_arg(char* dest, arg_t arg){
-//    const_fmt_detail::format_float<arg_t, fmt_data>(dest, arg);
+    const_fmt_detail::format_float<arg_t, fmt_data>(dest, arg);
 };
 
 // TODO: Error handling
diff --git a/const_fmt/format_impl.h b/const_fmt/format_impl.h
index 3cb9af1..3fe0359 100644
--- a/const_fmt/format_impl.h
+++ b/const_fmt/format_impl.h
@@ -112,6 +112,25 @@ constexpr inline void format_decimal(char* out, uint_t value, int n_digits,
     copy2(out, digits2(static_cast<size_t>(value)));
 }
 
+// returns {abs_value, was_negative}
+template <std::signed_integral int_t>
+constexpr std::pair<typename std::make_unsigned<int_t>::type, bool>
+get_abs_value(int_t value) {
+    using uint_t = typename std::make_unsigned<int_t>::type;
+
+    uint_t abs_value = static_cast<uint_t>(value);
+
+    const bool negative = value < 0;
+    if (negative) abs_value = 0 - abs_value;
+
+    return {abs_value, negative};
+}
+
+template <std::unsigned_integral int_t>
+constexpr std::pair<int_t, bool> get_abs_value(int_t value) {
+    return {value, false};
+}
+
 
 /*
  *
@@ -127,9 +146,7 @@ constexpr inline void format_int(char* out, uint_t value) {
 
 template <std::signed_integral int_t, fmt_data_t t_fmt_node>
 constexpr inline void format_int(char* out, int_t value) {
-    uint64_t   abs_value = static_cast<uint64_t>(value);
-    const bool negative  = value < 0;
-    if (negative) abs_value = 0 - abs_value;
+    const auto [abs_value, negative] = get_abs_value(value);
 
     const int n_digits = count_digits(abs_value);
 
@@ -153,9 +170,34 @@ constexpr inline void format_int(char* out, int_t value) {
 
 template <std::floating_point float_t, fmt_data_t t_fmt_node>
 constexpr inline void format_float(char* out, float_t value) {
+    // clang-format off
+    constexpr fmt_data_t fmt_node_integral = {
+        t_fmt_node.has_zero_padding,                  // has_zero_padding
+        t_fmt_node.length - t_fmt_node.precision - 1, // length
+        t_fmt_node.precision,                         // ignored
+        t_fmt_node.type,                              // ignored
+        t_fmt_node.position                           // ignored
+    };
+    constexpr fmt_data_t fmt_node_fractional = {
+        true,                                         // has_zero_padding
+        t_fmt_node.precision,                         // length
+        t_fmt_node.precision,                         // ignored
+        t_fmt_node.type,                              // ignored
+        t_fmt_node.position                           // ignored
+    };
+    // clang-format on
 
-    *(out)                                                = 'f';
     *(out + t_fmt_node.length - t_fmt_node.precision - 1) = '.';
+
+    const int integral = static_cast<int>(value);
+
+    constexpr std::size_t factor = const_pow(10, t_fmt_node.precision);
+    const int fractional = static_cast<int>((value - integral) * factor);
+
+    const auto [fractional_abs, fractional_negative] = get_abs_value(fractional);
+
+    format_int<int, fmt_node_integral>(out, integral);
+    format_int<uint16_t, fmt_node_fractional>(out + t_fmt_node.length - t_fmt_node.precision, fractional_abs);
 }
 
 
diff --git a/examples/src/examples.cpp b/examples/src/examples.cpp
index 4769151..82ab7ea 100644
--- a/examples/src/examples.cpp
+++ b/examples/src/examples.cpp
@@ -5,7 +5,7 @@
 using namespace const_fmt;
 
 int main() {
-    constexpr auto s = "abcdef {:04}"_const_fmt(123);
+    constexpr auto s = "abcdef {:9.3}"_const_fmt(-876.3);
 
     // Convert s (with a type of 'std::array<char, N>') into something
     // writable to std::cout