通过FFmpeg DLL实现-ss与-to参数功能需调用哪些API方法？

To replicate this stream-copy segment command using FFmpeg's C API (exposed by its DLLs), you’ll follow a structured workflow that maps each command flag to specific API calls. Below’s a breakdown of the key functions you’ll need, organized by workflow stage:

1. Initialization #

First, set up FFmpeg’s core components to recognize formats and codecs:

• av_register_all(): Registers all supported file formats and codecs (required before any other FFmpeg operations).
• avformat_network_init(): Optional for local files, but useful if you might handle network streams later.

2. Input File Setup #

Open and analyze the input video to access its streams:

• avformat_open_input(): Opens input.mp4 and creates an AVFormatContext (the main structure for handling media files).
• avformat_find_stream_info(): Reads metadata and populates the context with details about the input’s video/audio streams.

3. Seeking to the Start Time (-ss 00:20:20) #

Efficiently jump to the 20:20 mark (critical for stream copy, which relies on keyframes):

• Convert the start time to microseconds: (20*3600 + 20) * 1000000 = 7320000000 μs.
• Use avformat_seek_file(): This flexible function lets you seek to the nearest keyframe before your target time (use AVSEEK_FLAG_BACKWARD flag). Pass -1 as the stream index to sync all streams to the seek position.

4. Output File Setup (-y and -c copy) #

Prepare the output file with overwrite support and stream-copy behavior:

• avformat_alloc_output_context2(): Creates an AVFormatContext for test.mp4; FFmpeg auto-detects the output format from the filename.
• For each input stream:
  • avformat_new_stream(): Creates a matching stream in the output context.
  • avcodec_parameters_copy(): Copies codec parameters from the input to output stream (this implements -c copy—no re-encoding needed, so skip codec opening calls).
• avio_open(): Opens the output file with flags AVIO_FLAG_WRITE | AVIO_FLAG_TRUNC (the TRUNC flag handles -y by overwriting existing files).
• avformat_write_header(): Writes the output file’s header (e.g., MP4 metadata).

5. Trimming and Writing the Segment (-to 00:30:00) #

Read packets from the input, filter by time, and write to the output:

• Convert the end time to microseconds: 30*3600*1000000 = 10800000000 μs.
• Loop with av_read_frame(): Reads packets from the input.
• For each packet:
  • Use av_rescale_q() to convert the packet’s timestamp (stored in the stream’s timebase) to microseconds.
  • Check if the timestamp falls between your start and end time.
  • Adjust the packet’s timestamps to start from 0 in the output (subtract the start time and convert to the output stream’s timebase with av_rescale_q()).
  • Write the packet to the output with av_interleaved_write_frame() (ensures proper audio/video interleaving for MP4).
  • Stop the loop when the packet’s timestamp exceeds the end time.

6. Cleanup #

Properly release resources to avoid memory leaks:

• av_write_trailer(): Writes the output file’s trailer (finalizes the MP4 container).
• avio_close(): Closes the output file handle.
• avformat_close_input(): Closes the input file and frees its context.
• avformat_free_context(): Frees the output context.

Key Notes #

• For stream copy (-c copy), you don’t need codec-related functions like avcodec_open2()—we’re just passing raw compressed packets between files.
• Always check return values of FFmpeg API calls (negative integers indicate errors).
• Timebases vary per stream, so use av_rescale_q() to convert between timebases and microseconds for accurate time checks.

内容的提问来源于stack exchange，提问作者DM1974