Files
songs2vid/lib/ffmpeg/encode.ts
T
Atakan Doğan Özban f9b2a997a2 Add n8n community node, /api/v1/render alias, and job webhooks for OSS automation.
Payment-free self-hosted builds keep full API access with optional webhookUrl callbacks and the published n8n-nodes-songs2vid package source under integrations/n8n.
2026-08-08 16:32:28 +02:00

405 lines
12 KiB
TypeScript

import { spawn } from "child_process";
import fs from "fs/promises";
import path from "path";
import ffmpegStatic from "ffmpeg-static";
import { getVideoAttributionText } from "../branding";
import { getResolution } from "../constants";
import {
isCuratedFontKey,
sanitizeFontfileForFilter,
} from "../fonts";
import { resolveCuratedFontPath, resolveSystemFontPath, assertFontFile } from "../fonts-server";
import {
buildArtTrackFilterComplex,
buildClassicBlurFillFilterComplex,
isLowerCornerTemplate,
normalizeLayoutSettings,
type LayoutSettings,
} from "../layout";
import {
WATERMARK_WIDTH_FRACTION,
watermarkFontSizeForWidth,
} from "../preview-typography";
import { getWatermarkPath } from "../storage";
import {
buildDrawtextFilter,
normalizeWatermarkSettings,
overlayXy,
sanitizeDrawtext,
type WatermarkSettings,
} from "../watermark";
import { buildBrandWatermarkOverlayFilters } from "./brand-watermark";
function getFfmpegPath(): string {
if (process.env.FFMPEG_PATH) return process.env.FFMPEG_PATH;
if (ffmpegStatic) return ffmpegStatic;
return "ffmpeg";
}
export { getFfmpegPath };
async function fileExists(p: string): Promise<boolean> {
try {
await fs.access(p);
return true;
} catch {
return false;
}
}
/** Verify PNG magic bytes (prevents MIME spoof → FFmpeg surprises). */
export async function assertPngFile(filePath: string): Promise<void> {
const fh = await fs.open(filePath, "r");
try {
const buf = Buffer.alloc(8);
await fh.read(buf, 0, 8, 0);
const sig = Buffer.from([0x89, 0x50, 0x4e, 0x47, 0x0d, 0x0a, 0x1a, 0x0a]);
if (!buf.equals(sig)) {
throw new Error("Logo must be a valid PNG file");
}
} finally {
await fh.close();
}
}
async function resolveFontfileEscaped(
settings: WatermarkSettings,
weight: "regular" | "bold" = "regular",
): Promise<string | null> {
const key = settings.fontKey ?? "system";
if (key === "system") {
const preferred = resolveSystemFontPath(weight);
const fallback = weight === "bold" ? resolveSystemFontPath("regular") : preferred;
for (const fontPath of weight === "bold" ? [preferred, fallback] : [preferred]) {
if (!(await fileExists(fontPath))) continue;
try {
await assertFontFile(fontPath);
return sanitizeFontfileForFilter(fontPath);
} catch (err) {
console.warn(
`[ffmpeg] system font invalid at ${fontPath}: ${err instanceof Error ? err.message : err}`,
);
}
}
console.warn(`[ffmpeg] system font missing; using FFmpeg default`);
return null;
}
if (key === "custom") {
if (!settings.fontPath) return null;
if (!(await fileExists(settings.fontPath))) {
throw new Error("Custom watermark font file not found");
}
return sanitizeFontfileForFilter(path.resolve(settings.fontPath));
}
if (isCuratedFontKey(key)) {
const preferred = resolveCuratedFontPath(key, weight);
const fallback = weight === "bold" ? resolveCuratedFontPath(key, "regular") : preferred;
for (const fontPath of weight === "bold" && preferred !== fallback ? [preferred, fallback] : [preferred]) {
if (!(await fileExists(fontPath))) continue;
try {
await assertFontFile(fontPath);
return sanitizeFontfileForFilter(fontPath);
} catch {
/* try next */
}
}
console.warn(`[ffmpeg] curated font missing: ${key}; using system font`);
return null;
}
return null;
}
function classicScaleFilter(width: number, height: number): string {
return `scale=${width}:${height}:force_original_aspect_ratio=decrease,pad=${width}:${height}:(ow-iw)/2:(oh-ih)/2:black`;
}
/**
* Plan-aware audio for MP4.
* OSS / self-hosted defaults to Pro-quality 320k AAC.
*/
export type AudioEncodeOptions = {
bitrateKbps: 192 | 320;
copyWhenSafe: boolean;
audioPath: string;
};
function isCopySafeAudioPath(audioPath: string): boolean {
const ext = path.extname(audioPath).toLowerCase();
return ext === ".aac" || ext === ".m4a" || ext === ".mp4";
}
function pushAudioEncodeArgs(args: string[], audio: AudioEncodeOptions): void {
if (audio.copyWhenSafe && isCopySafeAudioPath(audio.audioPath)) {
args.push("-c:a", "copy");
return;
}
args.push(
"-c:a",
"aac",
"-b:a",
`${audio.bitrateKbps}k`,
"-ar",
"44100",
"-ac",
"2",
);
}
/**
* Art-track filter that ends at `outLabel` instead of hardcoded [laid].
*/
function artTrackFilterEndingAt(
opts: Parameters<typeof buildArtTrackFilterComplex>[0],
outLabel: string,
): string {
return buildArtTrackFilterComplex(opts).replace(/\[laid\]$/, `[${outLabel}]`);
}
function classicBlurFillEndingAt(
opts: Parameters<typeof buildClassicBlurFillFilterComplex>[0],
outLabel: string,
): string {
return buildClassicBlurFillFilterComplex(opts).replace(/\[laid\]$/, `[${outLabel}]`);
}
export async function encodeVideo(options: {
imagePath: string;
audioPath: string;
outputPath: string;
resolution: string;
includeWatermark: boolean;
watermark?: Partial<WatermarkSettings> | null;
layout?: LayoutSettings | null;
/** On-video song title (art-track layouts). */
songTitle?: string;
artist?: string | null;
/**
* Optional full-frame background for lower-corner templates.
* When set with LOWER_LEFT / LOWER_RIGHT, blur fill uses this image; cover stays sharp.
*/
backgroundImagePath?: string | null;
/** Defaults to 320k (self-hosted / studio). */
audioBitrateKbps?: 192 | 320;
audioCopyWhenSafe?: boolean;
}): Promise<void> {
const res = getResolution(options.resolution);
if (!res) throw new Error(`Invalid resolution: ${options.resolution}`);
const audioEncode: AudioEncodeOptions = {
bitrateKbps: options.audioBitrateKbps ?? 320,
copyWhenSafe: options.audioCopyWhenSafe ?? true,
audioPath: options.audioPath,
};
await fs.mkdir(path.dirname(options.outputPath), { recursive: true });
const settings = normalizeWatermarkSettings(options.watermark, options.includeWatermark);
const layoutSettings = options.layout
? normalizeLayoutSettings(options.layout)
: null;
const artTrack = Boolean(layoutSettings?.template);
const classicBlurFill = Boolean(layoutSettings && !artTrack && layoutSettings.blurFill);
const layout = artTrack && layoutSettings ? layoutSettings : null;
const watermarkWidth = Math.max(1, Math.round(res.width * WATERMARK_WIDTH_FRACTION));
const fontSize = watermarkFontSizeForWidth(res.width);
const fontfile = await resolveFontfileEscaped(settings, "regular");
const titleFontfile =
layout?.titleBold ? await resolveFontfileEscaped(settings, "bold") : null;
// Only pass a distinct bold face when it differs from regular (avoids faux-bold when bold TTF exists).
const titleFontfileDistinct =
titleFontfile && titleFontfile !== fontfile ? titleFontfile : null;
const args = ["-y", "-loop", "1", "-r", "1", "-i", options.imagePath, "-i", options.audioPath];
let nextInput = 2;
let separateBackgroundInputIndex: number | null = null;
let logoInputIndex: number | null = null;
let defaultWmInputIndex: number | null = null;
const useSeparateBackground =
Boolean(layout && isLowerCornerTemplate(layout.template) && options.backgroundImagePath);
if (useSeparateBackground && options.backgroundImagePath) {
if (!(await fileExists(options.backgroundImagePath))) {
throw new Error("Background image file not found");
}
args.push("-loop", "1", "-r", "1", "-i", options.backgroundImagePath);
separateBackgroundInputIndex = nextInput++;
}
const needsLogo = settings.mode === "logo" && Boolean(settings.logoPath);
const defaultPath = getWatermarkPath();
const useDefaultPng =
settings.mode === "default" && (await fileExists(defaultPath));
if (needsLogo && settings.logoPath) {
if (!(await fileExists(settings.logoPath))) {
throw new Error("Watermark logo file not found");
}
await assertPngFile(settings.logoPath);
args.push("-i", settings.logoPath);
logoInputIndex = nextInput++;
} else if (useDefaultPng) {
args.push("-i", defaultPath);
defaultWmInputIndex = nextInput++;
}
const applyWm =
settings.mode === "logo" && logoInputIndex !== null
? ("logo" as const)
: settings.mode === "text" && settings.text?.trim()
? ("text" as const)
: settings.mode === "default" && defaultWmInputIndex !== null
? ("default-png" as const)
: settings.mode === "default"
? ("default-text" as const)
: ("none" as const);
// Fast path: classic letterbox (black bars), no watermark, no blur fill
if (!layout && !classicBlurFill && applyWm === "none") {
args.push("-vf", classicScaleFilter(res.width, res.height));
args.push(
"-c:v",
"libx264",
"-tune",
"stillimage",
);
pushAudioEncodeArgs(args, audioEncode);
args.push(
"-shortest",
"-pix_fmt",
"yuv420p",
options.outputPath,
);
await runFfmpeg(args);
return;
}
const outDirect = applyWm === "none";
const baseLabel = outDirect ? "vout" : "base";
const filterParts: string[] = [];
if (layout) {
const titleEscaped = sanitizeDrawtext(options.songTitle?.trim() || "Untitled");
const artistRaw = options.artist?.trim();
const artistEscaped = artistRaw ? sanitizeDrawtext(artistRaw) : null;
filterParts.push(
artTrackFilterEndingAt(
{
width: res.width,
height: res.height,
layout,
titleEscaped,
artistEscaped,
fontfileEscaped: fontfile,
titleFontfileEscaped: titleFontfileDistinct,
separateBackgroundInputIndex,
},
baseLabel,
),
);
} else if (classicBlurFill && layoutSettings) {
filterParts.push(
classicBlurFillEndingAt(
{
width: res.width,
height: res.height,
blurAmount: layoutSettings.blurAmount,
blurOpacity: layoutSettings.blurOpacity,
},
baseLabel,
),
);
} else {
filterParts.push(`[0:v]${classicScaleFilter(res.width, res.height)}[${baseLabel}]`);
}
if (applyWm === "logo" && logoInputIndex !== null) {
const { x, y } = overlayXy(settings.position, settings.offsetX, settings.offsetY);
filterParts.push(
`[${logoInputIndex}:v]scale=${watermarkWidth}:-1[wm]`,
`[${baseLabel}][wm]overlay=${x}:${y}[vout]`,
);
} else if (applyWm === "text") {
const draw = buildDrawtextFilter({
text: settings.text!.trim(),
fontSize,
fontColor: "white@0.9",
position: settings.position,
offsetX: settings.offsetX,
offsetY: settings.offsetY,
fontfileEscaped: fontfile,
});
filterParts.push(`[${baseLabel}]${draw}[vout]`);
} else if (applyWm === "default-png" && defaultWmInputIndex !== null) {
filterParts.push(
...buildBrandWatermarkOverlayFilters({
baseLabel,
watermarkInputIndex: defaultWmInputIndex,
watermarkWidthPx: watermarkWidth,
offsetX: settings.offsetX,
offsetY: settings.offsetY,
}),
);
} else if (applyWm === "default-text") {
const draw = buildDrawtextFilter({
text: getVideoAttributionText(),
fontSize,
fontColor: "white@0.85",
position: settings.position,
offsetX: settings.offsetX,
offsetY: settings.offsetY,
fontfileEscaped: fontfile,
});
filterParts.push(`[${baseLabel}]${draw}[vout]`);
}
args.push("-filter_complex", filterParts.join(";"), "-map", "[vout]", "-map", "1:a");
args.push(
"-c:v",
"libx264",
"-tune",
"stillimage",
);
pushAudioEncodeArgs(args, audioEncode);
args.push(
"-shortest",
"-pix_fmt",
"yuv420p",
options.outputPath,
);
await runFfmpeg(args);
}
function runFfmpeg(args: string[]): Promise<void> {
return new Promise((resolve, reject) => {
const proc = spawn(getFfmpegPath(), args, { stdio: ["ignore", "pipe", "pipe"] });
let stderr = "";
proc.stderr.on("data", (chunk) => {
stderr += chunk.toString();
if (stderr.length > 64_000) stderr = stderr.slice(-32_000);
});
proc.on("close", (code) => {
if (code === 0) resolve();
else reject(new Error(`FFmpeg failed (code ${code}): ${stderr.slice(-500)}`));
});
proc.on("error", (err) => {
reject(new Error(`FFmpeg not found or failed to start: ${err.message}`));
});
});
}
export async function cleanupFiles(paths: string[]) {
for (const p of paths) {
try {
await fs.unlink(p);
} catch {
// ignore missing files
}
}
}