SHA256
Compare commits
5
Commits
| Author | SHA256 | Date | |
|---|---|---|---|
|
|
72d8d71928
|
||
|
|
413778e165
|
||
|
|
36552511f5
|
||
|
|
a1f9fa36ce
|
||
|
|
00971a6730
|
No files matched your search
@@ -48,3 +48,4 @@ desktop.ini
|
||||
|
||||
# Developer Utility Files
|
||||
/REVIEW_CHECKLIST.md
|
||||
/REVIEW_FFMPEG.md
|
||||
+1
-1
@@ -9,7 +9,7 @@ using MelonLoader.Utils;
|
||||
using UnityEngine;
|
||||
using Object = UnityEngine.Object;
|
||||
|
||||
[assembly: MelonInfo(typeof(RenderMod), "ADOFAI Renderer", "0.2.6-INDEV", "ADOFAI Editor Forum")]
|
||||
[assembly: MelonInfo(typeof(RenderMod), "ADOFAI Renderer", "0.2.7-INDEV", "ADOFAI Editor Forum")]
|
||||
[assembly: MelonGame("7th Beat Games", "A Dance of Fire and Ice")]
|
||||
|
||||
namespace ADOFAIRenderer;
|
||||
|
||||
+134
-111
@@ -191,6 +191,13 @@ internal sealed unsafe class FFmpegNativeDecoder : IDisposable
|
||||
private bool _isOpened;
|
||||
private bool _disposed;
|
||||
private bool _isDraining;
|
||||
|
||||
/// <summary>Number of video packets skipped because the decoder rejected them as invalid data.</summary>
|
||||
private int _skippedPacketCount;
|
||||
|
||||
/// <summary>Number of skipped packets for which a warning is logged.</summary>
|
||||
private const int MaxLoggedSkippedPackets = 5;
|
||||
|
||||
private readonly object _lock = new();
|
||||
|
||||
// Fields to detect SwsContext updates
|
||||
@@ -371,6 +378,16 @@ internal sealed unsafe class FFmpegNativeDecoder : IDisposable
|
||||
if (_videoStreamIndex == -1)
|
||||
throw new Exception("Failed to find video stream.");
|
||||
|
||||
// Only the video stream is used. The demuxer does not need to return the packets of the other streams
|
||||
// (audio, subtitles, ...) once they are marked as discarded.
|
||||
// Reference: AVStream.discard - "Selects which packets can be discarded at will and do not need to be demuxed."
|
||||
// https://ffmpeg.org/doxygen/trunk/structAVStream.html
|
||||
for (int i = 0; i < _pFormatContext->nb_streams; i++)
|
||||
{
|
||||
if (i != _videoStreamIndex)
|
||||
_pFormatContext->streams[i]->discard = AVDiscard.AVDISCARD_ALL;
|
||||
}
|
||||
|
||||
var stream = _pFormatContext->streams[_videoStreamIndex];
|
||||
_timeBase = stream->time_base;
|
||||
|
||||
@@ -441,9 +458,13 @@ internal sealed unsafe class FFmpegNativeDecoder : IDisposable
|
||||
TotalFrames = 0;
|
||||
}
|
||||
|
||||
Duration = _pFormatContext->duration > 0
|
||||
? _pFormatContext->duration / (double)ffmpeg.AV_TIME_BASE
|
||||
: (stream->duration > 0 ? stream->duration * ffmpeg.av_q2d(stream->time_base) : 0);
|
||||
// Duration of the video stream (AVStream.duration, in the stream time base).
|
||||
// When the stream does not carry a duration, the container duration (AVFormatContext.duration,
|
||||
// in AV_TIME_BASE units, the longest of all streams) is used.
|
||||
// Reference: https://ffmpeg.org/doxygen/trunk/structAVStream.html, https://ffmpeg.org/doxygen/trunk/structAVFormatContext.html
|
||||
Duration = stream->duration > 0
|
||||
? stream->duration * ffmpeg.av_q2d(stream->time_base)
|
||||
: _pFormatContext->duration > 0 ? _pFormatContext->duration / (double)ffmpeg.AV_TIME_BASE : 0;
|
||||
|
||||
// 9. Allocate frame and packet buffers
|
||||
// Reference: https://ffmpeg.org/doxygen/trunk/group__lavu__frame.html
|
||||
@@ -655,6 +676,8 @@ internal sealed unsafe class FFmpegNativeDecoder : IDisposable
|
||||
/// <summary>
|
||||
/// Reads packets from the container until a packet belonging to the video stream is found,
|
||||
/// then sends it to the decoder via <c>avcodec_send_packet</c>.
|
||||
/// A packet that the decoder rejects as invalid data (<c>AVERROR_INVALIDDATA</c>, e.g. a damaged packet)
|
||||
/// is skipped and the next packet is read. Any other decoder error is thrown.
|
||||
/// Reference: https://ffmpeg.org/doxygen/trunk/group__lavc__encdec.html
|
||||
/// </summary>
|
||||
/// <returns>True if a video packet was sent; false on EOF.</returns>
|
||||
@@ -670,6 +693,18 @@ internal sealed unsafe class FFmpegNativeDecoder : IDisposable
|
||||
{
|
||||
ret = ffmpeg.avcodec_send_packet(_pCodecContext, _pPacket);
|
||||
ffmpeg.av_packet_unref(_pPacket);
|
||||
|
||||
if (ret == ffmpeg.AVERROR_INVALIDDATA)
|
||||
{
|
||||
// The warning is logged for the first few packets only (a damaged file can have many).
|
||||
_skippedPacketCount++;
|
||||
if (_skippedPacketCount <= MaxLoggedSkippedPackets)
|
||||
{
|
||||
Melon<RenderMod>.Logger.Warning($"[FFmpegNativeDecoder] Skipped an invalid video packet ({_skippedPacketCount}): {GetFFmpegErrorString(ret)}");
|
||||
}
|
||||
continue;
|
||||
}
|
||||
|
||||
if (ret < 0)
|
||||
{
|
||||
throw new Exception($"Error sending packet to decoder: {GetFFmpegErrorString(ret)}");
|
||||
@@ -1212,107 +1247,24 @@ internal sealed unsafe class FFmpegNativeDecoder : IDisposable
|
||||
// ──────────────────────────────────────────────────────────────
|
||||
|
||||
/// <summary>
|
||||
/// Decodes and discards a specified number of frames immediately without GPU-to-CPU transfer overhead.
|
||||
/// Maximum distance in frames over which <see cref="SeekToTime"/> moves forward by decoding.
|
||||
/// Decoding forward needs no keyframe seek and no decoder flush, which is cheaper than seeking for short distances.
|
||||
/// </summary>
|
||||
/// <param name="count">Number of frames to discard.</param>
|
||||
private void SkipFrames(int count)
|
||||
{
|
||||
lock (_lock)
|
||||
{
|
||||
if (!_isOpened || _disposed) return;
|
||||
if (count <= 0) return;
|
||||
|
||||
int skipped = 0;
|
||||
if (_hasCachedFrame)
|
||||
{
|
||||
ffmpeg.av_frame_unref(_pCachedFrame);
|
||||
_hasCachedFrame = false;
|
||||
skipped++;
|
||||
}
|
||||
|
||||
while (skipped < count)
|
||||
{
|
||||
int ret = ffmpeg.avcodec_receive_frame(_pCodecContext, _pDecodedFrame);
|
||||
if (ret == 0)
|
||||
{
|
||||
GetTimestampAndUpdateIndex(_pDecodedFrame);
|
||||
ffmpeg.av_frame_unref(_pDecodedFrame);
|
||||
skipped++;
|
||||
continue;
|
||||
}
|
||||
|
||||
if (ret == ffmpeg.AVERROR(ffmpeg.EAGAIN))
|
||||
{
|
||||
if (_isDraining) break;
|
||||
|
||||
if (!ReadAndSendPacket())
|
||||
{
|
||||
_isDraining = true;
|
||||
ffmpeg.avcodec_send_packet(_pCodecContext, null);
|
||||
}
|
||||
}
|
||||
else if (ret == ffmpeg.AVERROR_EOF)
|
||||
{
|
||||
break;
|
||||
}
|
||||
else
|
||||
{
|
||||
throw new Exception($"Error skipping frames: {GetFFmpegErrorString(ret)}");
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
private const int MaxForwardDecodeFrames = 60;
|
||||
|
||||
/// <summary>
|
||||
/// Seeks to a specific playback time in seconds using backward keyframe seek followed by forward decode.
|
||||
/// Intermediate frames are discarded without GPU-to-host memory transfer to maximize seek speed;
|
||||
/// only the target frame is transferred and preserved in <see cref="_pCachedFrame"/>.
|
||||
/// Decodes forward from the current decoder position until the first frame whose presentation time is at least
|
||||
/// <paramref name="timeSeconds"/> minus half a frame, and keeps that frame in <see cref="_pCachedFrame"/>
|
||||
/// for the next <c>TryRead*</c> call.
|
||||
/// Frames before the target are discarded without GPU-to-host memory transfer; only the target frame is transferred.
|
||||
/// When the end of the stream is reached before the target, no frame is cached.
|
||||
/// </summary>
|
||||
/// <param name="timeSeconds">Target time in seconds to seek to.</param>
|
||||
public void SeekToTime(double timeSeconds)
|
||||
/// <param name="timeSeconds">Target time in seconds.</param>
|
||||
private void DecodeToTarget(double timeSeconds)
|
||||
{
|
||||
lock (_lock)
|
||||
{
|
||||
if (!_isOpened || _disposed) return;
|
||||
|
||||
if (timeSeconds < 0) timeSeconds = 0;
|
||||
double targetFrameIndex = timeSeconds * FPS;
|
||||
|
||||
// If already at target index or very close future, skip frames instead of seek to avoid I-frame seek latency
|
||||
bool useSkip = targetFrameIndex >= _currentFrameIndex && targetFrameIndex - _currentFrameIndex < 60;
|
||||
if (useSkip)
|
||||
{
|
||||
SkipFrames((int)(targetFrameIndex - _currentFrameIndex));
|
||||
return;
|
||||
}
|
||||
|
||||
if (_hasCachedFrame)
|
||||
{
|
||||
ffmpeg.av_frame_unref(_pCachedFrame);
|
||||
_hasCachedFrame = false;
|
||||
}
|
||||
|
||||
long targetPts = (long)Math.Round(timeSeconds / ffmpeg.av_q2d(_timeBase));
|
||||
int ret = ffmpeg.av_seek_frame(_pFormatContext, _videoStreamIndex, targetPts, ffmpeg.AVSEEK_FLAG_BACKWARD);
|
||||
if (ret < 0)
|
||||
{
|
||||
// Fallback to beginning of file on seek failure
|
||||
ret = ffmpeg.av_seek_frame(_pFormatContext, _videoStreamIndex, 0, ffmpeg.AVSEEK_FLAG_BACKWARD);
|
||||
if (ret < 0)
|
||||
{
|
||||
throw new Exception($"Seek failed: {GetFFmpegErrorString(ret)}");
|
||||
}
|
||||
}
|
||||
|
||||
// Flush decoder internal buffers
|
||||
// Reference: https://ffmpeg.org/doxygen/trunk/group__lavc__encdec.html
|
||||
ffmpeg.avcodec_flush_buffers(_pCodecContext);
|
||||
_isDraining = false;
|
||||
|
||||
// Decode frames until reaching the target presentation timestamp
|
||||
while (true)
|
||||
{
|
||||
ret = ffmpeg.avcodec_receive_frame(_pCodecContext, _pDecodedFrame);
|
||||
int ret = ffmpeg.avcodec_receive_frame(_pCodecContext, _pDecodedFrame);
|
||||
if (ret == 0)
|
||||
{
|
||||
double currentPts = GetFrameTimestamp(_pDecodedFrame);
|
||||
@@ -1343,28 +1295,18 @@ internal sealed unsafe class FFmpegNativeDecoder : IDisposable
|
||||
}
|
||||
|
||||
_hasCachedFrame = true;
|
||||
break;
|
||||
return;
|
||||
}
|
||||
|
||||
// Discard non-matching intermediate frame without transferring to CPU
|
||||
long ptsValue = _pDecodedFrame->best_effort_timestamp != ffmpeg.AV_NOPTS_VALUE
|
||||
? _pDecodedFrame->best_effort_timestamp
|
||||
: _pDecodedFrame->pts;
|
||||
if (ptsValue == ffmpeg.AV_NOPTS_VALUE)
|
||||
{
|
||||
_currentFrameIndex++;
|
||||
}
|
||||
else
|
||||
{
|
||||
_currentFrameIndex = (long)Math.Round(currentPts * FPS);
|
||||
}
|
||||
GetTimestampAndUpdateIndex(_pDecodedFrame);
|
||||
ffmpeg.av_frame_unref(_pDecodedFrame);
|
||||
continue;
|
||||
}
|
||||
|
||||
if (ret == ffmpeg.AVERROR(ffmpeg.EAGAIN))
|
||||
{
|
||||
if (_isDraining) break;
|
||||
if (_isDraining) return;
|
||||
|
||||
if (!ReadAndSendPacket())
|
||||
{
|
||||
@@ -1374,7 +1316,7 @@ internal sealed unsafe class FFmpegNativeDecoder : IDisposable
|
||||
}
|
||||
else if (ret == ffmpeg.AVERROR_EOF)
|
||||
{
|
||||
break;
|
||||
return;
|
||||
}
|
||||
else
|
||||
{
|
||||
@@ -1382,6 +1324,87 @@ internal sealed unsafe class FFmpegNativeDecoder : IDisposable
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Moves the playback position to a time in seconds.
|
||||
/// The next <c>TryRead*</c> call returns the first frame whose presentation time is at least
|
||||
/// <paramref name="timeSeconds"/> minus half a frame (the frame nearest to the target time).
|
||||
/// Both paths below return that same frame:
|
||||
/// - Forward decode: the target is ahead of the current position by less than <see cref="MaxForwardDecodeFrames"/> frames.
|
||||
/// Frames are decoded and discarded up to the target.
|
||||
/// - Keyframe seek: all other cases (target before the current position, far ahead, or the target frame already read).
|
||||
/// A backward keyframe seek is followed by a forward decode up to the target.
|
||||
/// The target frame is transferred and preserved in <see cref="_pCachedFrame"/>; when it is already there, nothing is decoded.
|
||||
/// </summary>
|
||||
/// <param name="timeSeconds">Target time in seconds to seek to.</param>
|
||||
public void SeekToTime(double timeSeconds)
|
||||
{
|
||||
lock (_lock)
|
||||
{
|
||||
if (!_isOpened || _disposed) return;
|
||||
|
||||
if (timeSeconds < 0) timeSeconds = 0;
|
||||
|
||||
// Earliest presentation time that satisfies the target (target minus half a frame).
|
||||
double threshold = timeSeconds - (0.5 / FPS);
|
||||
|
||||
// Index of the first frame that satisfies the target. The epsilon keeps a frame that lies exactly on the
|
||||
// margin from being excluded by floating point rounding.
|
||||
double targetFrameIndex = Math.Ceiling(timeSeconds * FPS - 0.5 - 1e-6);
|
||||
|
||||
if (_hasCachedFrame)
|
||||
{
|
||||
// A frame is already positioned for the next read (e.g. by a previous SeekToTime call).
|
||||
double cachedPts = GetFrameTimestamp(_pCachedFrame);
|
||||
if (cachedPts >= threshold)
|
||||
{
|
||||
// The cached frame is the first frame that satisfies the target.
|
||||
// When the target lies before it, the frames in between are no longer available and a keyframe seek is required.
|
||||
if (cachedPts - threshold < 1.0 / FPS) return;
|
||||
}
|
||||
else if (targetFrameIndex - _currentFrameIndex < MaxForwardDecodeFrames)
|
||||
{
|
||||
// The cached frame lies before the target: drop it and continue decoding forward.
|
||||
GetTimestampAndUpdateIndex(_pCachedFrame);
|
||||
ffmpeg.av_frame_unref(_pCachedFrame);
|
||||
_hasCachedFrame = false;
|
||||
DecodeToTarget(timeSeconds);
|
||||
return;
|
||||
}
|
||||
}
|
||||
else if (targetFrameIndex > _currentFrameIndex && targetFrameIndex - _currentFrameIndex < MaxForwardDecodeFrames)
|
||||
{
|
||||
// The target frame has not been read yet and is close: decode forward instead of seeking.
|
||||
DecodeToTarget(timeSeconds);
|
||||
return;
|
||||
}
|
||||
|
||||
if (_hasCachedFrame)
|
||||
{
|
||||
ffmpeg.av_frame_unref(_pCachedFrame);
|
||||
_hasCachedFrame = false;
|
||||
}
|
||||
|
||||
long targetPts = (long)Math.Round(timeSeconds / ffmpeg.av_q2d(_timeBase));
|
||||
int ret = ffmpeg.av_seek_frame(_pFormatContext, _videoStreamIndex, targetPts, ffmpeg.AVSEEK_FLAG_BACKWARD);
|
||||
if (ret < 0)
|
||||
{
|
||||
// Fallback to beginning of file on seek failure
|
||||
ret = ffmpeg.av_seek_frame(_pFormatContext, _videoStreamIndex, 0, ffmpeg.AVSEEK_FLAG_BACKWARD);
|
||||
if (ret < 0)
|
||||
{
|
||||
throw new Exception($"Seek failed: {GetFFmpegErrorString(ret)}");
|
||||
}
|
||||
}
|
||||
|
||||
// Flush decoder internal buffers
|
||||
// Reference: https://ffmpeg.org/doxygen/trunk/group__lavc__encdec.html
|
||||
ffmpeg.avcodec_flush_buffers(_pCodecContext);
|
||||
_isDraining = false;
|
||||
|
||||
// Decode frames until reaching the target presentation timestamp
|
||||
DecodeToTarget(timeSeconds);
|
||||
}
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
|
||||
@@ -83,7 +83,7 @@ namespace ADOFAIRenderer.Util.FFmpeg
|
||||
/// <summary>Frames per second (e.g., 60.0, 29.97, etc.).</summary>
|
||||
public double FPS;
|
||||
|
||||
/// <summary>Video codec type to use (H264, H265, AV1, VP9).</summary>
|
||||
/// <summary>Video codec type to use (H264, H265, AV1, VP9, VVC).</summary>
|
||||
public Codec Codec;
|
||||
|
||||
/// <summary>Bitrate control mode (CQP, CBR, VBR).</summary>
|
||||
@@ -723,6 +723,16 @@ namespace ADOFAIRenderer.Util.FFmpeg
|
||||
if (ffmpeg.avcodec_parameters_from_context(_pVideoStream->codecpar, _pCodecContext) < 0)
|
||||
throw new Exception("Failed to copy codec parameters to stream.");
|
||||
|
||||
// Stream time base: unit of the timestamps stored in the container.
|
||||
// Set to the codec time base (1 / framerate) so that every frame interval is a whole number of ticks.
|
||||
// When it is left unset, the muxer uses 1/90000, in which the interval of rates such as
|
||||
// 59.94 fps (1501.5 ticks), 23.976 fps (3753.75 ticks) or 165 fps (545.45 ticks) is not an integer,
|
||||
// and the frame durations alternate between two values.
|
||||
// The muxer may still adjust the value in avformat_write_header (e.g. mp4 scales it to at least 10000);
|
||||
// packet timestamps are rescaled to the final value in SendFrameAndWritePackets.
|
||||
// Reference: https://github.com/FFmpeg/FFmpeg/blob/master/libavformat/mux.c (init_muxer: default 1/90000)
|
||||
_pVideoStream->time_base = _pCodecContext->time_base;
|
||||
|
||||
// ── Step 4.5 (Optional): Initialize audio stream ──
|
||||
// If EncoderSettings.Audio is set, add audio stream.
|
||||
if (_settings.Audio.HasValue)
|
||||
@@ -825,7 +835,7 @@ namespace ADOFAIRenderer.Util.FFmpeg
|
||||
/// Reference - CQP options per encoder:
|
||||
/// NVENC: "qp" (single value applied to I/P/B frames)
|
||||
/// AMF: "qp_i" + "qp_p" (I-frame and P-frame configured separately)
|
||||
/// QSV: "global_quality" (Intel unique option)
|
||||
/// QSV: AVCodecContext.global_quality (Intel ICQ mode)
|
||||
/// Software: "crf" (Constant Rate Factor, smarter quality control)
|
||||
///
|
||||
/// Reference: https://ffmpeg.org/doxygen/trunk/structAVCodecContext.html
|
||||
@@ -847,11 +857,21 @@ namespace ADOFAIRenderer.Util.FFmpeg
|
||||
_pCodecContext->rc_max_rate = _pCodecContext->rc_min_rate = _settings.TargetBitrate * 1000L;
|
||||
_pCodecContext->rc_buffer_size = _settings.TargetBitrate * 1000;
|
||||
|
||||
// nal-hrd=cbr: x264/x265 dedicated option. Generates CBR bitstream matching
|
||||
// NAL HRD (Hypothetical Reference Decoder) specification. May be required for streaming compatibility.
|
||||
if (encoder == VideoEncoder.Libx264 || encoder == VideoEncoder.Libx265)
|
||||
// HRD (Hypothetical Reference Decoder) signaling for CBR output, used for streaming compatibility.
|
||||
// libx264: "nal-hrd=cbr" writes the NAL HRD information and pads filler data to keep the bitrate constant.
|
||||
// FFmpeg's option description: "Signal HRD information (requires vbv-bufsize; cbr not allowed in .mp4)".
|
||||
// Reference: https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/libx264.c ("nal-hrd": none/vbr/cbr)
|
||||
// libx265: "x265-params" = "hrd=1" (--hrd: "Enable the signaling of HRD parameters to the decoder")
|
||||
// + "strict-cbr=1" (--strict-cbr: "Enables stricter conditions to control bitrate deviance
|
||||
// from the target bitrate in ABR mode").
|
||||
// Reference: https://x265.readthedocs.io/en/master/cli.html
|
||||
if (encoder == VideoEncoder.Libx264)
|
||||
{
|
||||
ffmpeg.av_opt_set(_pCodecContext->priv_data, "nal-hrd", "cbr", 0);
|
||||
TrySetOption("nal-hrd", "cbr");
|
||||
}
|
||||
else if (encoder == VideoEncoder.Libx265)
|
||||
{
|
||||
TrySetOption("x265-params", "hrd=1:strict-cbr=1");
|
||||
}
|
||||
}
|
||||
break;
|
||||
@@ -875,38 +895,77 @@ namespace ADOFAIRenderer.Util.FFmpeg
|
||||
break;
|
||||
|
||||
case RateControlMode.CQP:
|
||||
// Valid quality range per encoder. The range of the private options is read from the
|
||||
// encoder's option definition (AVOption min/max); a declared minimum of -1 means "auto".
|
||||
// NVENC: "qp" H.264/HEVC 0~51, AV1 0~255
|
||||
// (libavcodec/nvenc_h264.c, nvenc_hevc.c: "qp" -1~51 / libavcodec/nvenc_av1.c: "qp" -1~255)
|
||||
// AMF: "qp_i" H.264/HEVC 0~51, AV1 0~255
|
||||
// (libavcodec/amfenc_h264.c, amfenc_hevc.c: "qp_i" -1~51 / libavcodec/amfenc_av1.c: "qp_i" -1~255)
|
||||
// QSV: AVCodecContext.global_quality 1~51
|
||||
// VVC: "qp" 0~63 (libavcodec/libvvenc.c: "qp" -1~63)
|
||||
// Software: "crf" x264/x265 0~51 (https://x265.readthedocs.io/en/master/cli.html: --crf 0..51.0),
|
||||
// VP9/AV1 0~63 (libavcodec/libvpxenc.c, libaomenc.c: "crf" -1~63 / libavcodec/libsvtav1.c: "crf" 0~63)
|
||||
// Reference: https://github.com/FFmpeg/FFmpeg/tree/master/libavcodec
|
||||
int qpMin, qpMax;
|
||||
if (encoder.Kind == EncoderKind.QSV)
|
||||
{
|
||||
// QSV runs in ICQ (Intelligent Constant Quality) mode when global_quality > 0 and rc_max_rate is 0;
|
||||
// global_quality 0 selects the encoder's default CQP. The ICQ quality is clamped to 1~51 by FFmpeg:
|
||||
// libavcodec/qsvenc.c: q->param.mfx.ICQQuality = av_clip(avctx->global_quality, 1, 51);
|
||||
// Reference: https://github.com/FFmpeg/FFmpeg/blob/master/libavcodec/qsvenc.c
|
||||
qpMin = 1;
|
||||
qpMax = 51;
|
||||
}
|
||||
else if (encoder.Kind == EncoderKind.NVENC) (qpMin, qpMax) = GetQualityRange("qp");
|
||||
else if (encoder.Kind == EncoderKind.AMF) (qpMin, qpMax) = GetQualityRange("qp_i");
|
||||
else if (encoder == VideoEncoder.LibVvenc) (qpMin, qpMax) = GetQualityRange("qp");
|
||||
else (qpMin, qpMax) = GetQualityRange("crf"); // Software encoder (libx264, libx265, libvpx-vp9, etc.)
|
||||
|
||||
// Default value 18: Quality close to "visually lossless" in H.264.
|
||||
// Provides similar level of quality in other codecs as well.
|
||||
var qpValue = _settings.QualityValue is >= 0 and <= 51 ? _settings.QualityValue : 18;
|
||||
const int defaultQuality = 18;
|
||||
var requestedQp = _settings.QualityValue;
|
||||
int qpValue;
|
||||
if (requestedQp < 0)
|
||||
{
|
||||
qpValue = Math.Clamp(defaultQuality, qpMin, qpMax);
|
||||
Melon<RenderMod>.Logger.Warning($"{encoder.Name}: Invalid quality value {requestedQp}. Using {qpValue}.");
|
||||
}
|
||||
else
|
||||
{
|
||||
qpValue = Math.Clamp(requestedQp, qpMin, qpMax);
|
||||
if (qpValue != requestedQp)
|
||||
Melon<RenderMod>.Logger.Warning($"{encoder.Name}: Quality value {requestedQp} is outside the supported range {qpMin}~{qpMax}. Using {qpValue}.");
|
||||
}
|
||||
|
||||
if (encoder.Kind == EncoderKind.NVENC)
|
||||
{
|
||||
// NVENC: Applies same QP to I/P/B frames with single "qp" option
|
||||
ffmpeg.av_opt_set(_pCodecContext->priv_data, "qp", qpValue.ToString(), 0);
|
||||
TrySetOption("qp", qpValue.ToString());
|
||||
}
|
||||
else if (encoder.Kind == EncoderKind.AMF)
|
||||
{
|
||||
// AMF: QP for I-frame (qp_i) and P-frame (qp_p) must be set separately.
|
||||
// Since B-frames are disabled, qp_b is not set.
|
||||
ffmpeg.av_opt_set(_pCodecContext->priv_data, "qp_i", qpValue.ToString(), 0);
|
||||
ffmpeg.av_opt_set(_pCodecContext->priv_data, "qp_p", qpValue.ToString(), 0);
|
||||
TrySetOption("qp_i", qpValue.ToString());
|
||||
TrySetOption("qp_p", qpValue.ToString());
|
||||
}
|
||||
else if (encoder.Kind == EncoderKind.QSV)
|
||||
{
|
||||
// QSV: Uses ICQ (Intelligent Constant Quality) mode with "global_quality" option
|
||||
ffmpeg.av_opt_set(_pCodecContext->priv_data, "global_quality", qpValue.ToString(), 0);
|
||||
// QSV: ICQ (Intelligent Constant Quality) mode, configured with AVCodecContext.global_quality.
|
||||
_pCodecContext->global_quality = qpValue;
|
||||
}
|
||||
else if (encoder == VideoEncoder.LibVvenc)
|
||||
{
|
||||
// libvvenc: Specifies Constant QP mode with "qp" option
|
||||
ffmpeg.av_opt_set(_pCodecContext->priv_data, "qp", qpValue.ToString(), 0);
|
||||
TrySetOption("qp", qpValue.ToString());
|
||||
}
|
||||
else // Software encoder (libx264, libx265, libvpx-vp9, etc.)
|
||||
else
|
||||
{
|
||||
// CRF (Constant Rate Factor): Smarter quality control than QP.
|
||||
// Automatically adjusts bitrate according to scene complexity to maintain consistent visual quality.
|
||||
// Range: 0 (lossless) ~ 51 (lowest quality). Typically between 18 and 28.
|
||||
ffmpeg.av_opt_set(_pCodecContext->priv_data, "crf", qpValue.ToString(), 0);
|
||||
// Range: 0 (lossless) ~ 51 (lowest quality) for x264/x265, 0~63 for VP9/AV1. Typically between 18 and 28.
|
||||
TrySetOption("crf", qpValue.ToString());
|
||||
}
|
||||
break;
|
||||
default:
|
||||
@@ -914,6 +973,49 @@ namespace ADOFAIRenderer.Util.FFmpeg
|
||||
}
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Sets a private option (priv_data) of the encoder.
|
||||
/// Returns true when FFmpeg accepts the option. Otherwise logs a warning with the FFmpeg error and returns false.
|
||||
/// </summary>
|
||||
private bool TrySetOption(string name, string value)
|
||||
{
|
||||
var ret = ffmpeg.av_opt_set(_pCodecContext->priv_data, name, value, 0);
|
||||
if (ret >= 0) return true;
|
||||
|
||||
Melon<RenderMod>.Logger.Warning($"[Encoder] Option '{name}={value}' was not applied: {GetFFmpegErrorString(ret)} (Code: {ret})");
|
||||
return false;
|
||||
}
|
||||
|
||||
/// <summary>
|
||||
/// Largest declared maximum that is treated as an actual quality range.
|
||||
/// A declared maximum above this value is treated as unbounded, and the fallback range 0~51 is used.
|
||||
/// The largest bounded quality range among the supported encoders is 0~255 (AV1 NVENC/AMF), while
|
||||
/// libx264/libx265 declare "crf" as -1~FLT_MAX (libavcodec/libx264.c, libx265.c).
|
||||
/// </summary>
|
||||
private const double MaxMeaningfulQualityLimit = 1000;
|
||||
|
||||
/// <summary>
|
||||
/// Returns the valid range of a private quality option of the encoder from its option definition (AVOption min/max).
|
||||
/// e.g. 0~51 for H.264/HEVC, 0~63 for VP9/AV1/VVC, 0~255 for AV1 NVENC/AMF.
|
||||
/// The lower bound is at least 0 (a declared minimum of -1 means "auto").
|
||||
/// When the option is missing or its declared maximum is unbounded (see <see cref="MaxMeaningfulQualityLimit"/>), returns 0~51,
|
||||
/// the QP scale of H.264/HEVC 8-bit (x264/x265 CRF 0~51).
|
||||
/// Reference: https://x265.readthedocs.io/en/master/cli.html (--crf 0..51.0)
|
||||
/// </summary>
|
||||
/// <param name="optionName">Name of the quality option in priv_data.</param>
|
||||
private (int Min, int Max) GetQualityRange(string optionName)
|
||||
{
|
||||
const int fallbackMax = 51;
|
||||
|
||||
var option = ffmpeg.av_opt_find(_pCodecContext->priv_data, optionName, null, 0, 0);
|
||||
if (option == null || !(option->max > 0 && option->max <= MaxMeaningfulQualityLimit))
|
||||
return (0, fallbackMax);
|
||||
|
||||
var min = Math.Max(0, (int)Math.Ceiling(option->min));
|
||||
var max = (int)Math.Floor(option->max);
|
||||
return min <= max ? (min, max) : (0, fallbackMax);
|
||||
}
|
||||
|
||||
// ================================================================
|
||||
// Write Frame
|
||||
// ================================================================
|
||||
|
||||
Reference in new issue
Block a user