FFmpegRecorder

FFmpegRecorder is a C++ class designed to handle real-time video and audio recording using the FFmpeg library. It supports encoding video (e.g., H.264) and audio (e.g., AAC) and writing them to a file or streaming endpoint.

This recorder handles frame timing, resampling, color format conversion, and interleaving streams into a muxed output file.

✅ Features

• Video encoding to H.264 (YUV420P)
• Audio encoding to AAC (FLT to target format)
• Automatic resampling and pixel format conversion
• Output to files (.mp4, .mkv, etc.) or network streams (e.g., rtmp://)
• Time synchronization and timestamping
• FIFO buffering for smoother audio input
• Adjustable parameters (resolution, FPS, sample rate, etc.)

🔧 Initialization

bool initialize(const char\* outputFile, int width, int height, int fps, int sampleRate, int channels);

This method configures the output context, selects appropriate encoders based on file/stream destination, and sets up:

• Video and audio codec contexts
• Resampling (SwrContext) and pixel format conversion (SwsContext)
• FIFO buffer for audio
• File or stream I/O
• FFmpeg muxing headers

Arguments:

• outputFile: Target file path or stream URL
• width, height: Video resolution
• fps: Video frame rate
• sampleRate: Audio sample rate (e.g., 44100)
• channels: Number of audio channels

🟢 Recording

bool sendVideoFrame(unsigned char\* rgbaData, int64_t timestamp);

Converts incoming RGBA video data to YUV420P and sends it for encoding.

• Timestamps are converted to stream time_base using FFmpeg scaling.
• Allocates and prepares an AVFrame for encoding.
• Encoded frames are written to the output stream.

⏱ Timestamping

int64_t getRecordingTimestamp();

Returns the number of milliseconds since recording started using a high-resolution clock. Used to maintain sync between audio and video.

🎵 Audio Flow

The functions are designed to:

• Accept interleaved AV_SAMPLE_FMT_FLT samples
• Use SwrContext to resample to codec-compatible format
• Accumulate samples in a FIFO until a full frame (e.g., 1024 samples) can be encoded

Key Members (Audio):

• m_aacFrameSize: Frame size for AAC
• m_audioFifo: Sample buffer to align input to required frame size
• m_audioSamplesProcessed, m_totalAudioSamplesQueued: Stats
• m_audioTimingInitialized: Ensures sync logic starts from a stable point

🧠 Internal Design Notes

Stream Format Decision:

• Chooses container and codec formats based on outputFile (e.g., flv for RTMP, guessed for .mp4)

• Encoder Hints: Preset: "medium"

• CRF: "23" for good quality and compression

Robust Logging:

• Uses av_strerror to provide clear FFmpeg error messages

🛑 Dependencies

You need FFmpeg development libraries installed:

• libavformat
• libavcodec
• libswscale
• libswresample
• libavutil