Spaces:
Sleeping
Sleeping
File size: 8,473 Bytes
3493993 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 | from __future__ import annotations
from dataclasses import dataclass
from pathlib import Path
from app.projects.editor_schemas import AudioClip, EditorDocument, MediaClip
from app.projects.errors import ProjectRenderInvalidError, ProjectRenderUnsupportedError
@dataclass(frozen=True, slots=True)
class RenderInput:
asset_id: str
path: Path
kind: str
source_start_ms: int
duration_ms: int
input_index: int
@dataclass(frozen=True, slots=True)
class RenderPlan:
args: tuple[str | Path, ...]
inputs: tuple[RenderInput, ...]
duration_ms: int
output_extension: str
def validate_renderable_document(document: EditorDocument) -> None:
if document.timeline.transitions:
raise ProjectRenderUnsupportedError(
"Timeline transitions are not supported by the render compiler yet."
)
for track in document.timeline.tracks:
if not track.visible:
continue
for clip in track.clips:
if clip.visible and clip.kind in {"caption", "effect"}:
raise ProjectRenderUnsupportedError(
f"The render compiler does not support {clip.kind} clips yet."
)
def compile_render(
document: EditorDocument,
*,
asset_paths: dict[str, tuple[Path, str]],
width: int,
height: int,
frame_rate: float,
output_format: str,
quality: str,
preset: str,
) -> RenderPlan:
"""Compile the currently supported editor primitives into FFmpeg args.
The compiler is deterministic: all inputs, offsets, trims, canvas settings,
and codecs are derived solely from the immutable editor snapshot/settings.
Captions, effects, transitions, and unsupported clip types fail closed.
"""
if output_format not in {"mp4", "webm"}:
raise ProjectRenderInvalidError("Unsupported render output format.")
if width % 2 or height % 2 or width < 2 or height < 2:
raise ProjectRenderInvalidError("Render resolution must be positive and even.")
validate_renderable_document(document)
inputs: list[RenderInput] = []
video_clips: list[tuple[MediaClip, RenderInput]] = []
audio_clips: list[tuple[AudioClip, RenderInput]] = []
for track in sorted(document.timeline.tracks, key=lambda item: item.order):
if not track.visible:
continue
for clip in track.clips:
if not clip.visible:
continue
if clip.kind in {"caption", "effect"}:
raise ProjectRenderUnsupportedError(
f"The render compiler does not support {clip.kind} clips yet."
)
if clip.kind == "audio" and track.muted:
continue
if clip.asset_id not in asset_paths:
raise ProjectRenderInvalidError("A referenced media asset is unavailable.")
path, mime_type = asset_paths[clip.asset_id]
if not path.is_file():
raise ProjectRenderInvalidError("A referenced media asset file is unavailable.")
input_index = len(inputs) + 1 # 0 is the generated black canvas.
item = RenderInput(
asset_id=clip.asset_id,
path=path,
kind=clip.kind,
source_start_ms=clip.source_start_ms,
duration_ms=clip.duration_ms,
input_index=input_index,
)
inputs.append(item)
if clip.kind == "media":
if clip.media_type == "image" and not mime_type.startswith("image/"):
raise ProjectRenderInvalidError("An image clip references a non-image asset.")
if clip.media_type == "video" and not mime_type.startswith("video/"):
raise ProjectRenderInvalidError("A video clip references a non-video asset.")
video_clips.append((clip, item))
elif clip.kind == "audio":
if not mime_type.startswith("audio/"):
raise ProjectRenderInvalidError("An audio clip references a non-audio asset.")
audio_clips.append((clip, item))
if not video_clips and not audio_clips:
raise ProjectRenderInvalidError("The editor timeline contains no renderable media.")
# Hidden tracks/clips are non-rendering state and must not extend the
# generated canvas with a long black tail.
duration_ms = max(
(clip.start_ms + clip.duration_ms for clip, _ in [*video_clips, *audio_clips]),
default=0,
)
args: list[str | Path] = [
"-f",
"lavfi",
"-t",
f"{duration_ms / 1000:.6f}",
"-i",
f"color=c=black:s={width}x{height}:r={frame_rate:.6g}",
]
for item in inputs:
if item.kind == "media" and any(
clip.media_type == "image" and ref is item for clip, ref in video_clips
):
args.extend(["-loop", "1"])
if item.source_start_ms:
args.extend(["-ss", f"{item.source_start_ms / 1000:.6f}"])
args.extend(["-t", f"{item.duration_ms / 1000:.6f}", "-i", item.path])
filters: list[str] = []
current_video = "base0"
filters.append(f"[0:v]setpts=PTS-STARTPTS[{current_video}]")
for number, (clip, item) in enumerate(video_clips, start=1):
label = f"clipv{number}"
scale_x = f"{clip.transform.scale_x:.6g}"
scale_y = f"{clip.transform.scale_y:.6g}"
chain = f"setpts=PTS-STARTPTS,scale=trunc(iw*{scale_x}/2)*2:trunc(ih*{scale_y}/2)*2"
if abs(clip.transform.rotation) > 1e-9:
chain += f",rotate={clip.transform.rotation:.6g}*PI/180:c=none:ow=rotw(iw):oh=roth(ih)"
if clip.opacity < 1:
chain += f",format=rgba,colorchannelmixer=aa={clip.opacity:.6g}"
chain += f",setpts=PTS+{clip.start_ms / 1000:.6f}/TB"
filters.append(f"[{item.input_index}:v]{chain}[{label}]")
next_video = f"mixv{number}"
x = f"(W-w)/2+{clip.transform.x:.6g}"
y = f"(H-h)/2+{clip.transform.y:.6g}"
filters.append(
f"[{current_video}][{label}]overlay=x={x}:y={y}:eof_action=pass:shortest=0[{next_video}]"
)
current_video = next_video
audio_labels: list[str] = []
for number, (clip, item) in enumerate(audio_clips, start=1):
label = f"clipa{number}"
chain = f"atrim=duration={clip.duration_ms / 1000:.6f},asetpts=PTS-STARTPTS,volume={clip.volume:.6g}"
if clip.fade_in_ms:
chain += f",afade=t=in:st=0:d={clip.fade_in_ms / 1000:.6f}"
if clip.fade_out_ms:
chain += f",afade=t=out:st={(clip.duration_ms - clip.fade_out_ms) / 1000:.6f}:d={clip.fade_out_ms / 1000:.6f}"
chain += f",adelay={clip.start_ms}|{clip.start_ms}"
filters.append(f"[{item.input_index}:a]{chain}[{label}]")
audio_labels.append(label)
if audio_labels:
filters.append(
"".join(f"[{label}]" for label in audio_labels)
+ f"amix=inputs={len(audio_labels)}:duration=longest:dropout_transition=0[aout]"
)
args.extend(["-filter_complex", ";".join(filters), "-map", f"[{current_video}]"])
if audio_labels:
args.extend(["-map", "[aout]"])
else:
args.append("-an")
if output_format == "webm":
args.extend(
[
"-c:v",
"libvpx-vp9",
"-crf",
{"draft": "30", "standard": "24", "high": "18"}[quality],
"-b:v",
"0",
"-deadline",
"good",
"-cpu-used",
{"fast": "5", "balanced": "2", "quality": "0"}[preset],
"-c:a",
"libopus",
]
)
else:
crf = {"draft": "28", "standard": "23", "high": "18"}[quality]
args.extend(
[
"-c:v",
"libx264",
"-pix_fmt",
"yuv420p",
"-preset",
{"fast": "veryfast", "balanced": "medium", "quality": "slow"}[preset],
"-crf",
crf,
"-c:a",
"aac",
"-b:a",
"128k",
"-movflags",
"+faststart",
]
)
args.extend(["-r", f"{frame_rate:.6g}", "-t", f"{duration_ms / 1000:.6f}"])
return RenderPlan(tuple(args), tuple(inputs), duration_ms, output_format)
|