开源C++智能语音识别库whisper.cpp开发使用入门

这篇具有很好参考价值的文章主要介绍了开源C++智能语音识别库whisper.cpp开发使用入门。希望对大家有所帮助。如果存在错误或未考虑完全的地方,请大家不吝赐教,您也可以点击"举报违法"按钮提交疑问。

whisper.cpp是一个C++编写的轻量级开源智能语音识别库,是基于openai的开源python智能语音模型whisper的移植版本,依赖项少,内存占用低,性能更优,方便作为依赖库集成的到应用程序中提供语音识别功能。

以下基于whisper.cpp的源码利用C++ api来开发实例demo演示读取本地音频文件并转成文字。

项目结构

whispercpp_starter
    - whisper.cpp-v1.5.0
    - src
      |- main.cpp
    - CMakeLists.txt

CMakeLists.txt

cmake_minimum_required(VERSION 3.15)

# this only works for unix, xapian source code not support compile in windows yet

project(whispercpp_starter)

set(CMAKE_CXX_STANDARD 14)
set(CMAKE_CXX_STANDARD_REQUIRED ON)

add_subdirectory(whisper.cpp-v1.5.0)

include_directories(
    ${CMAKE_CURRENT_SOURCE_DIR}/whisper.cpp-v1.5.0
    ${CMAKE_CURRENT_SOURCE_DIR}/whisper.cpp-v1.5.0/examples
)

file(GLOB SRC
    src/*.h
    src/*.cpp
)

add_executable(${PROJECT_NAME} ${SRC})

target_link_libraries(${PROJECT_NAME}
    common
    whisper # remember to copy dll or so to bin folder
)

main.cpp

#include <cmath>
#include <fstream>
#include <cstdio>
#include <string>
#include <thread>
#include <vector>
#include <cstring>

#include "common.h"
#include "whisper.h"

#if defined(_MSC_VER)
#pragma warning(disable: 4244 4267) // possible loss of data
#endif

// Terminal color map. 10 colors grouped in ranges [0.0, 0.1, ..., 0.9]
// Lowest is red, middle is yellow, highest is green.
const std::vector<std::string> k_colors = {
    "\033[38;5;196m", "\033[38;5;202m", "\033[38;5;208m", "\033[38;5;214m", "\033[38;5;220m",
    "\033[38;5;226m", "\033[38;5;190m", "\033[38;5;154m", "\033[38;5;118m", "\033[38;5;82m",
};

//  500 -> 00:05.000
// 6000 -> 01:00.000
std::string to_timestamp(int64_t t, bool comma = false)
{
    int64_t msec = t * 10;
    int64_t hr = msec / (1000 * 60 * 60);
    msec = msec - hr * (1000 * 60 * 60);
    int64_t min = msec / (1000 * 60);
    msec = msec - min * (1000 * 60);
    int64_t sec = msec / 1000;
    msec = msec - sec * 1000;

    char buf[32];
    snprintf(buf, sizeof(buf), "%02d:%02d:%02d%s%03d", (int)hr, (int)min, (int)sec, comma ? "," : ".", (int)msec);

    return std::string(buf);
}

int timestamp_to_sample(int64_t t, int n_samples)
{
    return std::max(0, std::min((int)n_samples - 1, (int)((t * WHISPER_SAMPLE_RATE) / 100)));
}

// helper function to replace substrings
void replace_all(std::string& s, const std::string& search, const std::string& replace)
{
    for (size_t pos = 0; ; pos += replace.length())
    {
        pos = s.find(search, pos);
        if (pos == std::string::npos) break;
        s.erase(pos, search.length());
        s.insert(pos, replace);
    }
}

// command-line parameters
struct whisper_params
{
    int32_t n_threads = std::min(4, (int32_t)std::thread::hardware_concurrency());
    int32_t n_processors = 1;
    int32_t offset_t_ms = 0;
    int32_t offset_n = 0;
    int32_t duration_ms = 0;
    int32_t progress_step = 5;
    int32_t max_context = -1;
    int32_t max_len = 0;
    int32_t best_of = whisper_full_default_params(WHISPER_SAMPLING_GREEDY).greedy.best_of;
    int32_t beam_size = whisper_full_default_params(WHISPER_SAMPLING_BEAM_SEARCH).beam_search.beam_size;

    float word_thold = 0.01f;
    float entropy_thold = 2.40f;
    float logprob_thold = -1.00f;

    bool speed_up = false;
    bool debug_mode = false;
    bool translate = false;
    bool detect_language = false;
    bool diarize = false;
    bool tinydiarize = false;
    bool split_on_word = false;
    bool no_fallback = false;
    bool output_txt = false;
    bool output_vtt = false;
    bool output_srt = false;
    bool output_wts = false;
    bool output_csv = false;
    bool output_jsn = false;
    bool output_jsn_full = false;
    bool output_lrc = false;
    bool print_special = false;
    bool print_colors = false;
    bool print_progress = false;
    bool no_timestamps = false;
    bool log_score = false;
    bool use_gpu = true;

    std::string language = "en";
    std::string prompt;
    std::string font_path = "/System/Library/Fonts/Supplemental/Courier New Bold.ttf";
    std::string model = "models/ggml-base.en.bin";

    // [TDRZ] speaker turn string
    std::string tdrz_speaker_turn = " [SPEAKER_TURN]"; // TODO: set from command line

    std::string openvino_encode_device = "CPU";

    std::vector<std::string> fname_inp = {};
    std::vector<std::string> fname_out = {};
};

struct whisper_print_user_data
{
    const whisper_params* params;

    const std::vector<std::vector<float>>* pcmf32s;
    int progress_prev;
};

std::string estimate_diarization_speaker(std::vector<std::vector<float>> pcmf32s, int64_t t0, int64_t t1, bool id_only = false)
{
    std::string speaker = "";
    const int64_t n_samples = pcmf32s[0].size();

    const int64_t is0 = timestamp_to_sample(t0, n_samples);
    const int64_t is1 = timestamp_to_sample(t1, n_samples);

    double energy0 = 0.0f;
    double energy1 = 0.0f;

    for (int64_t j = is0; j < is1; j++)
    {
        energy0 += fabs(pcmf32s[0][j]);
        energy1 += fabs(pcmf32s[1][j]);
    }

    if (energy0 > 1.1 * energy1)
    {
        speaker = "0";
    }
    else if (energy1 > 1.1 * energy0)
    {
        speaker = "1";
    }
    else
    {
        speaker = "?";
    }

    //printf("is0 = %lld, is1 = %lld, energy0 = %f, energy1 = %f, speaker = %s\n", is0, is1, energy0, energy1, speaker.c_str());

    if (!id_only)
    {
        speaker.insert(0, "(speaker ");
        speaker.append(")");
    }

    return speaker;
}
void whisper_print_progress_callback(struct whisper_context* /*ctx*/, struct whisper_state* /*state*/, int progress, void* user_data)
{
    int progress_step = ((whisper_print_user_data*)user_data)->params->progress_step;
    int* progress_prev = &(((whisper_print_user_data*)user_data)->progress_prev);
    if (progress >= *progress_prev + progress_step)
    {
        *progress_prev += progress_step;
        fprintf(stderr, "%s: progress = %3d%%\n", __func__, progress);
    }
}

void whisper_print_segment_callback(struct whisper_context* ctx, struct whisper_state* /*state*/, int n_new, void* user_data)
{
    const auto& params = *((whisper_print_user_data*)user_data)->params;
    const auto& pcmf32s = *((whisper_print_user_data*)user_data)->pcmf32s;

    const int n_segments = whisper_full_n_segments(ctx);

    std::string speaker = "";

    int64_t t0 = 0;
    int64_t t1 = 0;

    // print the last n_new segments
    const int s0 = n_segments - n_new;

    if (s0 == 0)
    {
        printf("\n");
    }

    for (int i = s0; i < n_segments; i++)
    {
        if (!params.no_timestamps || params.diarize)
        {
            t0 = whisper_full_get_segment_t0(ctx, i);
            t1 = whisper_full_get_segment_t1(ctx, i);
        }

        if (!params.no_timestamps)
        {
            printf("[%s --> %s]  ", to_timestamp(t0).c_str(), to_timestamp(t1).c_str());
        }

        if (params.diarize && pcmf32s.size() == 2)
        {
            speaker = estimate_diarization_speaker(pcmf32s, t0, t1);
        }

        if (params.print_colors)
        {
            for (int j = 0; j < whisper_full_n_tokens(ctx, i); ++j)
            {
                if (params.print_special == false)
                {
                    const whisper_token id = whisper_full_get_token_id(ctx, i, j);
                    if (id >= whisper_token_eot(ctx))
                    {
                        continue;
                    }
                }

                const char* text = whisper_full_get_token_text(ctx, i, j);
                const float  p = whisper_full_get_token_p(ctx, i, j);

                const int col = std::max(0, std::min((int)k_colors.size() - 1, (int)(std::pow(p, 3) * float(k_colors.size()))));

                printf("%s%s%s%s", speaker.c_str(), k_colors[col].c_str(), text, "\033[0m");
            }
        }
        else
        {
            const char* text = whisper_full_get_segment_text(ctx, i);

            printf("%s%s", speaker.c_str(), text);
        }

        if (params.tinydiarize)
        {
            if (whisper_full_get_segment_speaker_turn_next(ctx, i))
            {
                printf("%s", params.tdrz_speaker_turn.c_str());
            }
        }

        // with timestamps or speakers: each segment on new line
        if (!params.no_timestamps || params.diarize)
        {
            printf("\n");
        }

        fflush(stdout);
    }
}

bool output_txt(struct whisper_context* ctx, const char* fname, const whisper_params& params, std::vector<std::vector<float>> pcmf32s)
{
    std::ofstream fout(fname);
    if (!fout.is_open())
    {
        fprintf(stderr, "%s: failed to open '%s' for writing\n", __func__, fname);
        return false;
    }

    fprintf(stderr, "%s: saving output to '%s'\n", __func__, fname);

    const int n_segments = whisper_full_n_segments(ctx);
    for (int i = 0; i < n_segments; ++i)
    {
        const char* text = whisper_full_get_segment_text(ctx, i);
        std::string speaker = "";

        if (params.diarize && pcmf32s.size() == 2)
        {
            const int64_t t0 = whisper_full_get_segment_t0(ctx, i);
            const int64_t t1 = whisper_full_get_segment_t1(ctx, i);
            speaker = estimate_diarization_speaker(pcmf32s, t0, t1);
        }

        fout << speaker << text << "\n";
    }

    return true;
}

int main(int argc, char** argv)
{
    const std::string model_file_path = "./ggml-base.en.bin";
    const std::string audio_file_path = "sample.wav"; // should be wav 16bit format

    // set whisper params
    whisper_params params;
    params.model = model_file_path;
    params.fname_inp.emplace_back(audio_file_path);

    // whisper init
    struct whisper_context_params cparams;
    cparams.use_gpu = params.use_gpu;

    struct whisper_context* ctx = whisper_init_from_file_with_params(params.model.c_str(), cparams);

    if (ctx == nullptr)
    {
        fprintf(stderr, "error: failed to initialize whisper context\n");
        return 3;
    }

    // initialize openvino encoder. this has no effect on whisper.cpp builds that don't have OpenVINO configured
    whisper_ctx_init_openvino_encoder(ctx, nullptr, params.openvino_encode_device.c_str(), nullptr);

    for (int f = 0; f < (int)params.fname_inp.size(); ++f)
    {
        const auto fname_inp = params.fname_inp[f];
        const auto fname_out = f < (int)params.fname_out.size() && !params.fname_out[f].empty() ? params.fname_out[f] : params.fname_inp[f];

        std::vector<float> pcmf32;               // mono-channel F32 PCM
        std::vector<std::vector<float>> pcmf32s; // stereo-channel F32 PCM

        if (!read_wav(fname_inp, pcmf32, pcmf32s, params.diarize))
        {
            fprintf(stderr, "error: failed to read WAV file '%s'\n", fname_inp.c_str());
            continue;
        }

        // print system information
        {
            fprintf(stderr, "\n");
            fprintf(stderr, "system_info: n_threads = %d / %d | %s\n",
                params.n_threads * params.n_processors, std::thread::hardware_concurrency(), whisper_print_system_info());
        }

        // print some info about the processing
        {
            fprintf(stderr, "\n");
            if (!whisper_is_multilingual(ctx))
            {
                if (params.language != "en" || params.translate)
                {
                    params.language = "en";
                    params.translate = false;
                    fprintf(stderr, "%s: WARNING: model is not multilingual, ignoring language and translation options\n", __func__);
                }
            }
            if (params.detect_language)
            {
                params.language = "auto";
            }
            fprintf(stderr, "%s: processing '%s' (%d samples, %.1f sec), %d threads, %d processors, %d beams + best of %d, lang = %s, task = %s, %stimestamps = %d ...\n",
                __func__, fname_inp.c_str(), int(pcmf32.size()), float(pcmf32.size()) / WHISPER_SAMPLE_RATE,
                params.n_threads, params.n_processors, params.beam_size, params.best_of,
                params.language.c_str(),
                params.translate ? "translate" : "transcribe",
                params.tinydiarize ? "tdrz = 1, " : "",
                params.no_timestamps ? 0 : 1);

            fprintf(stderr, "\n");
        }

        // run the inference
        {
            whisper_full_params wparams = whisper_full_default_params(WHISPER_SAMPLING_GREEDY);

            wparams.strategy = params.beam_size > 1 ? WHISPER_SAMPLING_BEAM_SEARCH : WHISPER_SAMPLING_GREEDY;

            wparams.print_realtime = false;
            wparams.print_progress = params.print_progress;
            wparams.print_timestamps = !params.no_timestamps;
            wparams.print_special = params.print_special;
            wparams.translate = params.translate;
            wparams.language = params.language.c_str();
            wparams.detect_language = params.detect_language;
            wparams.n_threads = params.n_threads;
            wparams.n_max_text_ctx = params.max_context >= 0 ? params.max_context : wparams.n_max_text_ctx;
            wparams.offset_ms = params.offset_t_ms;
            wparams.duration_ms = params.duration_ms;

            wparams.token_timestamps = params.output_wts || params.output_jsn_full || params.max_len > 0;
            wparams.thold_pt = params.word_thold;
            wparams.max_len = params.output_wts && params.max_len == 0 ? 60 : params.max_len;
            wparams.split_on_word = params.split_on_word;

            wparams.speed_up = params.speed_up;
            wparams.debug_mode = params.debug_mode;

            wparams.tdrz_enable = params.tinydiarize; // [TDRZ]

            wparams.initial_prompt = params.prompt.c_str();

            wparams.greedy.best_of = params.best_of;
            wparams.beam_search.beam_size = params.beam_size;

            wparams.temperature_inc = params.no_fallback ? 0.0f : wparams.temperature_inc;
            wparams.entropy_thold = params.entropy_thold;
            wparams.logprob_thold = params.logprob_thold;

            whisper_print_user_data user_data = { &params, &pcmf32s, 0 };

            // this callback is called on each new segment
            if (!wparams.print_realtime)
            {
                wparams.new_segment_callback = whisper_print_segment_callback;
                wparams.new_segment_callback_user_data = &user_data;
            }

            if (wparams.print_progress)
            {
                wparams.progress_callback = whisper_print_progress_callback;
                wparams.progress_callback_user_data = &user_data;
            }

            // examples for abort mechanism
            // in examples below, we do not abort the processing, but we could if the flag is set to true

            // the callback is called before every encoder run - if it returns false, the processing is aborted
            {
                static bool is_aborted = false; // NOTE: this should be atomic to avoid data race

                wparams.encoder_begin_callback = [](struct whisper_context* /*ctx*/, struct whisper_state* /*state*/, void* user_data) {
                    bool is_aborted = *(bool*)user_data;
                    return !is_aborted;
                    };
                wparams.encoder_begin_callback_user_data = &is_aborted;
            }

            // the callback is called before every computation - if it returns true, the computation is aborted
            {
                static bool is_aborted = false; // NOTE: this should be atomic to avoid data race

                wparams.abort_callback = [](void* user_data) {
                    bool is_aborted = *(bool*)user_data;
                    return is_aborted;
                    };
                wparams.abort_callback_user_data = &is_aborted;
            }

            if (whisper_full_parallel(ctx, wparams, pcmf32.data(), pcmf32.size(), params.n_processors) != 0)
            {
                fprintf(stderr, "%s: failed to process audio\n", argv[0]);
                return 10;
            }
        }

        // output stuff
        {
            printf("\n");

            // output to text file
            if (params.output_txt)
            {
                const auto fname_txt = fname_out + ".txt";
                output_txt(ctx, fname_txt.c_str(), params, pcmf32s);
            }
        }
    }

    // whisper release
    whisper_print_timings(ctx);
    whisper_free(ctx);

    return 0;
}

注:

  • whisper支持的模型文件需要自己去下载
  • whisper.cpp编译可以配置多种类型的增强选项,比如支持CPU/GPU加速,数据计算加速库
  • whisper.cpp的编译cmake文件做了少量改动,方便集成到项目,具体可参看demo

源码

whispercpp_starter

本文由博客一文多发平台 OpenWrite 发布!文章来源地址https://www.toymoban.com/news/detail-828523.html

到了这里,关于开源C++智能语音识别库whisper.cpp开发使用入门的文章就介绍完了。如果您还想了解更多内容,请在右上角搜索TOY模板网以前的文章或继续浏览下面的相关文章,希望大家以后多多支持TOY模板网!

本文来自互联网用户投稿,该文观点仅代表作者本人,不代表本站立场。本站仅提供信息存储空间服务,不拥有所有权,不承担相关法律责任。如若转载,请注明出处: 如若内容造成侵权/违法违规/事实不符,请点击违法举报进行投诉反馈,一经查实,立即删除!

领支付宝红包 赞助服务器费用

相关文章

  • OpenAI 开源语音识别 Whisper

            Whisper是一个通用语音识别模型。它是在各种音频的大型数据集上训练的,也是一个多任务模型,可以执行多语言语音识别以及语音翻译和语言识别。                人工智能公司 OpenAI 拥有 GTP-3 语言模型,并为 GitHub Copilot 提供技术支持的 ,宣布开源了

    2024年02月09日
    浏览(71)
  • Whisper OpenAI开源语音识别模型

    Whisper 是一个自动语音识别(ASR,Automatic Speech Recognition)系统,OpenAI 通过从网络上收集了 68 万小时的多语言(98 种语言)和多任务(multitask)监督数据对 Whisper 进行了训练。OpenAI 认为使用这样一个庞大而多样的数据集,可以提高对口音、背景噪音和技术术语的识别能力。除

    2024年02月16日
    浏览(55)
  • 可以白嫖的语音识别开源项目whisper的搭建详细过程 | 如何在Linux中搭建OpenAI开源的语音识别项目Whisper

    原文来自我个人的博客。 服务器为GPU服务器。点击这里跳转到我使用的GPU服务器。我搭建 whisper 选用的是 NVIDIA A 100显卡,4GB显存。 Python版本要在3.8~3.11之间。 输入下面命令查看使用的Python版本。 为啥要安装Anaconda? 为了减少不同项目使用的库的版本冲突,我们可以使用An

    2024年02月09日
    浏览(53)
  • 语音识别开源框架 openAI-whisper

    Whisper 是一种通用的语音识别模型。 它是OpenAI于2022年9月份开源的在各种音频的大型数据集上训练的语音识别模型,也是一个可以执行多语言语音识别、语音翻译和语言识别的多任务模型。 GitHub - yeyupiaoling/Whisper-Finetune: 微调Whisper语音识别模型和加速推理,支持Web部署和Andr

    2024年02月17日
    浏览(64)
  • 开源语音识别faster-whisper部署教程

    源码地址 模型下载地址: 下载 cuBLAS and cuDNN 在 conda 环境中创建 python 运行环境 激活虚拟环境 安装 faster-whisper 依赖 执行完以上步骤后,我们可以写代码了 说明: 更多内容欢迎访问博客 对应视频内容欢迎访问视频

    2024年02月04日
    浏览(54)
  • OpenAI开源!!Whisper语音识别实战!!【环境配置+代码实现】

    目录 环境配置 代码实现 ******  实现 .mp4转换为 .wav文件,识别后进行匹配并输出出现的次数 ******  完整代码实现请私信 安装 ffmpeg 打开网址   https://github.com/BtbN/FFmpeg-Builds/releases 下载如下图所示的文件 下载后解压  我的路径是G:ffmpeg-master-latest-win64-gpl-shared

    2024年02月13日
    浏览(54)
  • chatGPT的耳朵!OpenAI的开源语音识别AI:Whisper !

    语音识别是通用人工智能的重要一环!可以说是AI的耳朵! 它可以让机器理解人类的语音,并将其转换为文本或其他形式的输出。 语音识别的应用场景非常广泛,比如智能助理、语音搜索、语音翻译、语音输入等等。 然而,语音识别也面临着很多挑战,比如不同的语言、口

    2024年03月14日
    浏览(51)
  • OpenAI开源语音识别模型Whisper在Windows系统的安装详细过程

    Python的安装很简单,点击这里进行下载。 安装完成之后,输入python -V可以看到版本信息,说明已经安装成功了。 如果输入python -V命令没有看到上面的这样的信息,要么是安装失败,要么是安装好之后没有自动配置环境变量,如何配置环境变量可以从网上搜索。 Python的具体安

    2024年02月08日
    浏览(60)
  • OpenAI开源全新解码器和语音识别模型Whisper-v3

    在11月7日OpenAI的首届开发者大会上,除了推出一系列重磅产品之外,还开源了两款产品,全新解码器Consistency Decoder(一致性解码器)和最新语音识别模型Whisper v3。 据悉,Consistency Decoder可以替代Stable Diffusion VAE解码器。该解码器可以改善所有与Stable Diffusion 1.0+ VAE兼容的图像,

    2024年02月05日
    浏览(59)
  • whisper实践--基于whisper+pyqt5开发的语音识别翻译生成字幕工具

    大家新年快乐,事业生活蒸蒸日上,解封的第一个年,想必大家都回家过年,好好陪陪家人了吧,这篇文章也是我在老家码的,还记得上篇我带大家基本了解了whisper,相信大家对whisper是什么,怎么安装whisper,以及使用都有了一个认识,这次作为新年第一篇文章,我将介绍一

    2024年02月01日
    浏览(44)

觉得文章有用就打赏一下文章作者

支付宝扫一扫打赏

博客赞助

微信扫一扫打赏

请作者喝杯咖啡吧~博客赞助

支付宝扫一扫领取红包,优惠每天领

二维码1

领取红包

二维码2

领红包