如何使用FFmpeg C API剪辑视频

Question

如何使用FFmpeg C API剪辑视频？例如，从00:10:00到00:20:00。

我需要使用哪些函数？

我使用以下函数转换视频：

int convert(char *file) {
    AVFrame *frame;
    AVPacket inPacket, outPacket;

    if(avio_open(&outFormatContext->pb, file, AVIO_FLAG_WRITE) < 0) {
        fprintf(stderr, "convert(): cannot open out file\n");
        return 0;
    }

    avformat_write_header(outFormatContext, NULL);
    frame = avcodec_alloc_frame();
    av_init_packet(&inPacket);

    while(av_read_frame(inFormatContext, &inPacket) >= 0) {
        if(inPacket.stream_index == inVideoStreamIndex) {
            avcodec_decode_video2(inCodecContext, frame, &frameFinished, &inPacket);
            if(frameFinished) {
                av_init_packet(&outPacket);
                avcodec_encode_video2(outCodecContext, &outPacket, frame, &outputed);
                if(outputed) {
                    if (av_write_frame(outFormatContext, &outPacket) != 0) {
                        fprintf(stderr, "convert(): error while writing video frame\n");
                        return 0;
                    }
                }
                av_free_packet(&outPacket);
            }
        }
    }

    av_write_trailer(outFormatContext);
    av_free_packet(&inPacket);
    return 1;
}

- Ilya Zharkov

---

请说明您已经尝试过什么。您在 Google 上搜索了这个主题吗？哪些方法对您没有起作用？ - Bas van Dijk

我没有找到C API的任何示例。也许我需要使用avformat_seek_file函数来寻找起始位置。但是我是FFmpeg的新手，不知道接下来该怎么做。 - Ilya Zharkov

Answer 1

正如Wagner Patriota所说，“如果只想剪辑视频，您不需要重新编码视频”。以下是基于ffmpeg remuxing.c 示例的代码，您无需重新编码视频。

#include <libavutil/timestamp.h>
#include <libavformat/avformat.h>

static void log_packet(const AVFormatContext *fmt_ctx, const AVPacket *pkt, const char *tag)
{
    AVRational *time_base = &fmt_ctx->streams[pkt->stream_index]->time_base;

    printf("%s: pts:%s pts_time:%s dts:%s dts_time:%s duration:%s duration_time:%s stream_index:%d\n",
           tag,
           av_ts2str(pkt->pts), av_ts2timestr(pkt->pts, time_base),
           av_ts2str(pkt->dts), av_ts2timestr(pkt->dts, time_base),
           av_ts2str(pkt->duration), av_ts2timestr(pkt->duration, time_base),
           pkt->stream_index);
}

int cut_video(double from_seconds, double end_seconds, const char* in_filename, const char* out_filename) {
    AVOutputFormat *ofmt = NULL;
    AVFormatContext *ifmt_ctx = NULL, *ofmt_ctx = NULL;
    AVPacket pkt;
    int ret, i;

    av_register_all();

    if ((ret = avformat_open_input(&ifmt_ctx, in_filename, 0, 0)) < 0) {
        fprintf(stderr, "Could not open input file '%s'", in_filename);
        goto end;
    }

    if ((ret = avformat_find_stream_info(ifmt_ctx, 0)) < 0) {
        fprintf(stderr, "Failed to retrieve input stream information");
        goto end;
    }

    av_dump_format(ifmt_ctx, 0, in_filename, 0);

    avformat_alloc_output_context2(&ofmt_ctx, NULL, NULL, out_filename);
    if (!ofmt_ctx) {
        fprintf(stderr, "Could not create output context\n");
        ret = AVERROR_UNKNOWN;
        goto end;
    }

    ofmt = ofmt_ctx->oformat;

    for (i = 0; i < ifmt_ctx->nb_streams; i++) {
        AVStream *in_stream = ifmt_ctx->streams[i];
        AVStream *out_stream = avformat_new_stream(ofmt_ctx, in_stream->codec->codec);
        if (!out_stream) {
            fprintf(stderr, "Failed allocating output stream\n");
            ret = AVERROR_UNKNOWN;
            goto end;
        }

        ret = avcodec_copy_context(out_stream->codec, in_stream->codec);
        if (ret < 0) {
            fprintf(stderr, "Failed to copy context from input to output stream codec context\n");
            goto end;
        }
        out_stream->codec->codec_tag = 0;
        if (ofmt_ctx->oformat->flags & AVFMT_GLOBALHEADER)
            out_stream->codec->flags |= AV_CODEC_FLAG_GLOBAL_HEADER;
    }
    av_dump_format(ofmt_ctx, 0, out_filename, 1);

    if (!(ofmt->flags & AVFMT_NOFILE)) {
        ret = avio_open(&ofmt_ctx->pb, out_filename, AVIO_FLAG_WRITE);
        if (ret < 0) {
            fprintf(stderr, "Could not open output file '%s'", out_filename);
            goto end;
        }
    }

    ret = avformat_write_header(ofmt_ctx, NULL);
    if (ret < 0) {
        fprintf(stderr, "Error occurred when opening output file\n");
        goto end;
    }

    //    int indexs[8] = {0};


    //    int64_t start_from = 8*AV_TIME_BASE;
    ret = av_seek_frame(ifmt_ctx, -1, from_seconds*AV_TIME_BASE, AVSEEK_FLAG_ANY);
    if (ret < 0) {
        fprintf(stderr, "Error seek\n");
        goto end;
    }

    int64_t *dts_start_from = malloc(sizeof(int64_t) * ifmt_ctx->nb_streams);
    memset(dts_start_from, 0, sizeof(int64_t) * ifmt_ctx->nb_streams);
    int64_t *pts_start_from = malloc(sizeof(int64_t) * ifmt_ctx->nb_streams);
    memset(pts_start_from, 0, sizeof(int64_t) * ifmt_ctx->nb_streams);

    while (1) {
        AVStream *in_stream, *out_stream;

        ret = av_read_frame(ifmt_ctx, &pkt);
        if (ret < 0)
            break;

        in_stream  = ifmt_ctx->streams[pkt.stream_index];
        out_stream = ofmt_ctx->streams[pkt.stream_index];

        log_packet(ifmt_ctx, &pkt, "in");

        if (av_q2d(in_stream->time_base) * pkt.pts > end_seconds) {
            av_free_packet(&pkt);
            break;
        }

        if (dts_start_from[pkt.stream_index] == 0) {
            dts_start_from[pkt.stream_index] = pkt.dts;
            printf("dts_start_from: %s\n", av_ts2str(dts_start_from[pkt.stream_index]));
        }
        if (pts_start_from[pkt.stream_index] == 0) {
            pts_start_from[pkt.stream_index] = pkt.pts;
            printf("pts_start_from: %s\n", av_ts2str(pts_start_from[pkt.stream_index]));
        }

        /* copy packet */
        pkt.pts = av_rescale_q_rnd(pkt.pts - pts_start_from[pkt.stream_index], in_stream->time_base, out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);
        pkt.dts = av_rescale_q_rnd(pkt.dts - dts_start_from[pkt.stream_index], in_stream->time_base, out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);
        if (pkt.pts < 0) {
            pkt.pts = 0;
        }
        if (pkt.dts < 0) {
            pkt.dts = 0;
        }
        pkt.duration = (int)av_rescale_q((int64_t)pkt.duration, in_stream->time_base, out_stream->time_base);
        pkt.pos = -1;
        log_packet(ofmt_ctx, &pkt, "out");
        printf("\n");

        ret = av_interleaved_write_frame(ofmt_ctx, &pkt);
        if (ret < 0) {
            fprintf(stderr, "Error muxing packet\n");
            break;
        }
        av_free_packet(&pkt);
    }
    free(dts_start_from);
    free(pts_start_from);

    av_write_trailer(ofmt_ctx);
end:

    avformat_close_input(&ifmt_ctx);

    /* close output */
    if (ofmt_ctx && !(ofmt->flags & AVFMT_NOFILE))
        avio_closep(&ofmt_ctx->pb);
    avformat_free_context(ofmt_ctx);

    if (ret < 0 && ret != AVERROR_EOF) {
        fprintf(stderr, "Error occurred: %s\n", av_err2str(ret));
        return 1;
    }

    return 0;
}

- justin

---

我遇到了一个复用数据包的错误。我的代码可以编译，你能帮忙吗？我可以与你分享输出结果。 - Dinesh Kumar

是的，我也遇到了“复用数据包错误”的问题，因为某些时候 pkt.dts > pkt.pts，所以这段代码肯定有问题。 - Gleb Ignatev

Answer 2

如果你只想剪辑视频，而不需要重新编码视频，那么你不需要重新编码视频。所以我猜你想要为某种原因剪辑和重新编码视频。所以，根据你的代码：

请注意，你必须要有访问视频AVStream*结构的权限... 我将其命名为inVideoStream。

int convert_and_cut(char *file, float starttime, float endtime) {
    AVFrame *frame;
    AVPacket inPacket, outPacket;

    if(avio_open(&outFormatContext->pb, file, AVIO_FLAG_WRITE) < 0) {
        fprintf(stderr, "convert(): cannot open out file\n");
        return 0;
    }

    // seek to the start time you wish.
    // BEGIN
    AVRational default_timebase;
    default_timebase.num = 1;
    default_timebase.den = AV_TIME_BASE;

    // suppose you have access to the "inVideoStream" of course
    int64_t starttime_int64 = av_rescale_q((int64_t)( starttime * AV_TIME_BASE ), default_timebase, inVideoStream->time_base);
    int64_t endtime_int64 = av_rescale_q((int64_t)( endtime * AV_TIME_BASE ), default_timebase, inVideoStream->time_base);

    if(avformat_seek_file(inFormatContext, inVideoStreamIndex, INT64_MIN, starttime_int64, INT64_MAX, 0) < 0) {
        // error... do something...
        return 0; // usually 0 is used for success in C, but I am following your code.
    }

    avcodec_flush_buffers( inVideoStream->codec );
    // END

    avformat_write_header(outFormatContext, NULL);
    frame = avcodec_alloc_frame();
    av_init_packet(&inPacket);

    // you used avformat_seek_file() to seek CLOSE to the point you want... in order to give precision to your seek,
    // just go on reading the packets and checking the packets PTS (presentation timestamp) 
    while(av_read_frame(inFormatContext, &inPacket) >= 0) {
        if(inPacket.stream_index == inVideoStreamIndex) {
            avcodec_decode_video2(inCodecContext, frame, &frameFinished, &inPacket);
            // this line guarantees you are getting what you really want.
            if(frameFinished && frame->pkt_pts >= starttime_int64 && frame->pkt_pts <= endtime_int64) {
                av_init_packet(&outPacket);
                avcodec_encode_video2(outCodecContext, &outPacket, frame, &outputed);
                if(outputed) {
                    if (av_write_frame(outFormatContext, &outPacket) != 0) {
                        fprintf(stderr, "convert(): error while writing video frame\n");
                        return 0;
                    }
                }
                av_free_packet(&outPacket);
            }

            // exit the loop if you got the frames you want.
            if(frame->pkt_pts > endtime_int64) {
                break;
            }
        }
    }

    av_write_trailer(outFormatContext);
    av_free_packet(&inPacket);
    return 1;
}

- Wagner Patriota

---

请问您能解释一下 frameFinished 和 outputed 变量的含义吗？ - AnujAroshA

当你解码一个单独的数据包时，你仍然没有足够的信息来构成一帧（根据编解码器的类型，有些情况下是可以的），当你解码一组代表一帧的数据包时，你就有了一个完整的画面！这就是为什么frameFinished会告诉你已经解码到足够的程度以形成一帧。 - Wagner Patriota

outputed 发生了类似的逻辑。根据编解码器，每帧都可能有一个数据包，但其中有些没有...因此假设 avcodec_encode_video2 在累积信息直到准备好输出数据包时，你会保存这个数据包。 - Wagner Patriota

Answer 3

此外，正如ustin所说，该代码基于ffmpeg remuxing.c示例。

1. 在out_stream->codec->codec_tag = 0;之前添加out_stream->time_base = in_stream->time_base;

2. 在int stream_mapping_size = 0;之后添加int64_t *start_from = NULL;

3. 在stream_mapping = av_mallocz_array(stream_mapping_size, sizeof(*stream_mapping));之后添加start_from = av_mallocz_array(stream_mapping_size, sizeof(*start_from));

4. 在stream_mapping[i] = stream_index++;之后添加start_from[stream_mapping[i]] = INT16_MIN;

5. 添加

if (start_from[pkt.stream_index] == INT16_MIN) {
    int64_t dts = pkt.dts;
    int64_t pts = pkt.pts;
    int64_t min_ts = dts > pts ? pts : dts;
    start_from[pkt.stream_index] = min_ts < 0 ? 0 : min_ts;
}

之后

if (pkt.stream_index >= stream_mapping_size ||
   stream_mapping[pkt.stream_index] < 0) {
        av_packet_unref(&pkt);
        continue;
}

更改Justin的代码，关于pts/dst设置值为以下内容：

pkt.pts = av_rescale_q_rnd(pkt.pts - start_from[pkt.stream_index], in_stream->time_base, out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);

pkt.dts = av_rescale_q_rnd(pkt.dts - start_from[pkt.stream_index], in_stream->time_base, out_stream->time_base, AV_ROUND_NEAR_INF|AV_ROUND_PASS_MINMAX);

6. 在end:之前添加free(start_from);

提示：开始/结束时间戳就像Justin所说的那样

- Zhusong