HY-2012

Upload ASR and TTS models

a37072b verified 4 months ago

2.57 kB

/**************************************************************************************************
 * ZipVoice AXERA C++ Port
 *
 * Simple WAV file writer for float32 mono audio.
 **************************************************************************************************/

#pragma once

#include <vector>
#include <string>
#include <fstream>
#include <cstdint>
#include <cstring>

class WavWriter {
public:
    /**
     * Write float32 mono audio samples to a WAV file.
     *
     * @param path        Output file path
     * @param samples     Float32 audio samples in range [-1.0, 1.0]
     * @param sample_rate Sample rate in Hz (e.g., 24000)
     * @return true on success
     */
    static bool Write(const std::string& path,
                      const std::vector<float>& samples,
                      int sample_rate) {
        std::ofstream file(path, std::ios::binary);
        if (!file.is_open()) return false;

        int num_samples = static_cast<int>(samples.size());
        int byte_rate = sample_rate * 2;  // 16-bit mono
        int data_size = num_samples * 2;
        int chunk_size = 36 + data_size;

        // RIFF header
        Write4CC(file, "RIFF");
        WriteU32LE(file, chunk_size);
        Write4CC(file, "WAVE");

        // fmt chunk
        Write4CC(file, "fmt ");
        WriteU32LE(file, 16);           // chunk size
        WriteU16LE(file, 1);            // PCM format
        WriteU16LE(file, 1);            // mono
        WriteU32LE(file, sample_rate);
        WriteU32LE(file, byte_rate);
        WriteU16LE(file, 2);            // block align
        WriteU16LE(file, 16);           // bits per sample

        // data chunk
        Write4CC(file, "data");
        WriteU32LE(file, data_size);

        // Convert float32 to int16 (with clipping)
        for (float sample : samples) {
            // Clamp to [-1, 1]
            if (sample > 1.0f) sample = 1.0f;
            if (sample < -1.0f) sample = -1.0f;
            int16_t val = static_cast<int16_t>(sample * 32767.0f);
            WriteU16LE(file, static_cast<uint16_t>(val));
        }

        file.close();
        return true;
    }

private:
    static void WriteU16LE(std::ofstream& f, uint16_t v) {
        f.put(v & 0xFF);
        f.put((v >> 8) & 0xFF);
    }

    static void WriteU32LE(std::ofstream& f, uint32_t v) {
        f.put(v & 0xFF);
        f.put((v >> 8) & 0xFF);
        f.put((v >> 16) & 0xFF);
        f.put((v >> 24) & 0xFF);
    }

    static void Write4CC(std::ofstream& f, const char* s) {
        f.write(s, 4);
    }
};

	/**************************************************************************************************
	* ZipVoice AXERA C++ Port
	*
	* Simple WAV file writer for float32 mono audio.
	**************************************************************************************************/

	#pragma once

	#include <vector>
	#include <string>
	#include <fstream>
	#include <cstdint>
	#include <cstring>

	class WavWriter {
	public:
	/**
	* Write float32 mono audio samples to a WAV file.
	*
	* @param path Output file path
	* @param samples Float32 audio samples in range [-1.0, 1.0]
	* @param sample_rate Sample rate in Hz (e.g., 24000)
	* @return true on success
	*/
	static bool Write(const std::string& path,
	const std::vector<float>& samples,
	int sample_rate) {
	std::ofstream file(path, std::ios::binary);
	if (!file.is_open()) return false;

	int num_samples = static_cast<int>(samples.size());
	int byte_rate = sample_rate * 2; // 16-bit mono
	int data_size = num_samples * 2;
	int chunk_size = 36 + data_size;

	// RIFF header
	Write4CC(file, "RIFF");
	WriteU32LE(file, chunk_size);
	Write4CC(file, "WAVE");

	// fmt chunk
	Write4CC(file, "fmt ");
	WriteU32LE(file, 16); // chunk size
	WriteU16LE(file, 1); // PCM format
	WriteU16LE(file, 1); // mono
	WriteU32LE(file, sample_rate);
	WriteU32LE(file, byte_rate);
	WriteU16LE(file, 2); // block align
	WriteU16LE(file, 16); // bits per sample

	// data chunk
	Write4CC(file, "data");
	WriteU32LE(file, data_size);

	// Convert float32 to int16 (with clipping)
	for (float sample : samples) {
	// Clamp to [-1, 1]
	if (sample > 1.0f) sample = 1.0f;
	if (sample < -1.0f) sample = -1.0f;
	int16_t val = static_cast<int16_t>(sample * 32767.0f);
	WriteU16LE(file, static_cast<uint16_t>(val));
	}

	file.close();
	return true;
	}

	private:
	static void WriteU16LE(std::ofstream& f, uint16_t v) {
	f.put(v & 0xFF);
	f.put((v >> 8) & 0xFF);
	}

	static void WriteU32LE(std::ofstream& f, uint32_t v) {
	f.put(v & 0xFF);
	f.put((v >> 8) & 0xFF);
	f.put((v >> 16) & 0xFF);
	f.put((v >> 24) & 0xFF);
	}

	static void Write4CC(std::ofstream& f, const char* s) {
	f.write(s, 4);
	}
	};