Skip to content

Open-source AI podcast clipper. Generate vertical clips with face tracking and burned-in captions. CLI, MCP server, web app.

Add `studio` command: cut a fragment + Remotion intro/outro bookends (#4)

RIKI0afe2246c8cb
diff --git a/backend/cli.py b/backend/cli.py
index aa25ffe..3b8cd90 100644
--- a/backend/cli.py
+++ b/backend/cli.py
@@ -302,6 +302,67 @@ def _should_enter_post_render_loop(config: dict, interrupted: bool, results: lis
return bool(config.get("post_render_review", False) or (interrupted and _has_successful_results(results)))
+def cmd_studio(args):
+ """Cut a fragment + wrap it with Remotion intro/outro bookends.
+
+ Thin wrapper around scripts/clip_studio.py (which orchestrates the
+ fragment render, bookend renders, and the concat). Runs it as a
+ subprocess with this same interpreter so the venv is reused.
+ """
+ if not getattr(args, "save_brand", False) and not args.paragraph and (args.start is None or args.end is None):
+ print("Error: provide either --start and --end, or --paragraph \"text\"", file=sys.stderr)
+ sys.exit(1)
+
+ project_root = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+ script = os.path.join(project_root, "scripts", "clip_studio.py")
+ if not os.path.exists(script):
+ print(f"Error: clip_studio.py not found at {script}", file=sys.stderr)
+ sys.exit(1)
+
+ # When just saving the brand, no video is needed; pass a placeholder the
+ # script ignores because --save-brand returns before touching the video.
+ video_arg = os.path.abspath(args.video) if args.video else "_brand_only_"
+ cmd = [sys.executable, script, video_arg]
+ if getattr(args, "save_brand", False):
+ cmd += ["--save-brand"]
+ if args.start is not None:
+ cmd += ["--start", str(args.start)]
+ if args.end is not None:
+ cmd += ["--end", str(args.end)]
+ if args.paragraph:
+ cmd += ["--paragraph", args.paragraph]
+ if args.language:
+ cmd += ["--language", args.language]
+ cmd += [
+ "--caption-style", args.caption_style,
+ "--crop", args.crop,
+ "--intro-seconds", str(args.intro_seconds),
+ "--outro-seconds", str(args.outro_seconds),
+ ]
+ if args.outro_title is not None:
+ cmd += ["--outro-title", args.outro_title]
+ if args.platforms is not None:
+ cmd += ["--platforms", args.platforms]
+ if args.accent is not None:
+ cmd += ["--accent", args.accent]
+ if args.bg is not None:
+ cmd += ["--bg", args.bg]
+ if args.intro_title:
+ cmd += ["--intro-title", args.intro_title]
+ if args.handle:
+ cmd += ["--handle", args.handle]
+ if args.output:
+ cmd += ["--output", os.path.abspath(args.output)]
+ if args.no_intro:
+ cmd += ["--no-intro"]
+ if args.no_outro:
+ cmd += ["--no-outro"]
+
+ import subprocess
+ rc = subprocess.run(cmd).returncode
+ sys.exit(rc)
+
+
def cmd_process(args):
"""Full auto pipeline: transcribe → suggest → export."""
from services.clip_generator import generate_clip
@@ -3181,6 +3242,29 @@ def main():
proc.add_argument("--review-each", action="store_true", help="Review each rendered clip interactively")
proc.add_argument("--post-review", action="store_true", help="Open the post-render review loop after export")
+ # ── studio ──
+ studio = sub.add_parser("studio", help="Cut a fragment + add Remotion intro/outro (follow-us) bookends")
+ studio.add_argument("video", nargs="?", default=None, help="Path to the source video (omit only with --save-brand)")
+ studio.add_argument("--start", type=float, help="Fragment start (seconds)")
+ studio.add_argument("--end", type=float, help="Fragment end (seconds)")
+ studio.add_argument("--paragraph", help="Find the fragment by matching this text in the transcript")
+ studio.add_argument("--language", help="Transcription language (e.g. es). Auto-detect if omitted.")
+ studio.add_argument("--caption-style", choices=["hormozi", "karaoke", "subtle", "branded"], default="hormozi")
+ studio.add_argument("--crop", choices=["center", "face", "speaker", "speaker-hardcut"], default="face")
+ studio.add_argument("-o", "--output", help="Final output path")
+ studio.add_argument("--intro-title", help="Intro headline (default: derived from first words)")
+ studio.add_argument("--outro-title", default=None)
+ studio.add_argument("--handle", help="Handle shown on cards, e.g. @yourbrand")
+ studio.add_argument("--platforms", default=None)
+ studio.add_argument("--intro-seconds", type=float, default=2.0)
+ studio.add_argument("--outro-seconds", type=float, default=3.0)
+ studio.add_argument("--accent", default=None)
+ studio.add_argument("--bg", default=None)
+ studio.add_argument("--no-intro", action="store_true")
+ studio.add_argument("--no-outro", action="store_true")
+ studio.add_argument("--save-brand", action="store_true",
+ help="Save handle/platforms/outro-title/accent/bg as the default brand and exit")
+
# ── presets ──
pre = sub.add_parser("presets", help="Manage presets")
pre_sub = pre.add_subparsers(dest="presets_action")
@@ -3368,6 +3452,8 @@ def main():
if not getattr(args, "no_banner", False):
print()
cmd_process(args)
+ elif args.command == "studio":
+ cmd_studio(args)
elif args.command == "thumbnails":
cmd_thumbnails(args)
elif args.command == "swap-thumbnail":
diff --git a/remotion/render-bookend.mjs b/remotion/render-bookend.mjs
new file mode 100644
index 0000000..ee6c7df
--- /dev/null
+++ b/remotion/render-bookend.mjs
@@ -0,0 +1,118 @@
+#!/usr/bin/env node
+/**
+ * Render an intro or outro "bookend" card via Remotion, with a silent audio
+ * track so it concatenates cleanly with the main clip.
+ *
+ * Usage:
+ * node remotion/render-bookend.mjs \
+ * --mode outro \
+ * --title "Follow for more" \
+ * --handle "@yourbrand" \
+ * --platforms tiktok,instagram,youtube,x \
+ * --seconds 3 \
+ * --output /path/to/outro.mp4 \
+ * [--bg "#0B0B0F"] [--accent "#FFE000"] [--fps 30] [--width 1080] [--height 1920]
+ */
+
+import { bundle } from "@remotion/bundler";
+import { renderMedia, selectComposition } from "@remotion/renderer";
+import path from "path";
+import fs from "fs";
+import os from "os";
+import crypto from "crypto";
+import { spawnSync } from "child_process";
+import { fileURLToPath } from "url";
+
+const __dirname = path.dirname(fileURLToPath(import.meta.url));
+const PROJECT_ROOT = path.resolve(__dirname, "..");
+const CACHE_DIR = path.join(PROJECT_ROOT, ".podcli", "cache", "remotion-bundle");
+const ENTRY_POINT = path.join(__dirname, "src", "index.ts");
+
+function parseArgs() {
+ const args = process.argv.slice(2);
+ const opts = {};
+ for (let i = 0; i < args.length; i++) {
+ if (args[i].startsWith("--") && i + 1 < args.length) {
+ opts[args[i].replace(/^--/, "")] = args[i + 1];
+ i++;
+ }
+ }
+ return opts;
+}
+
+async function getBundle() {
+ // Reuse podcli's cached bundle if present; otherwise build fresh.
+ if (fs.existsSync(path.join(CACHE_DIR, "index.html"))) return CACHE_DIR;
+ return await bundle({ entryPoint: ENTRY_POINT, outDir: CACHE_DIR });
+}
+
+async function main() {
+ const opts = parseArgs();
+ if (!opts.mode || !opts.output) {
+ console.error("Usage: render-bookend.mjs --mode intro|outro --title <t> --output <path> [--handle --platforms --seconds --bg --accent]");
+ process.exit(1);
+ }
+
+ const fps = parseInt(opts.fps || "30", 10);
+ const width = parseInt(opts.width || "1080", 10);
+ const height = parseInt(opts.height || "1920", 10);
+ const seconds = parseFloat(opts.seconds || "3");
+ const durationInFrames = Math.round(seconds * fps);
+ const platforms = (opts.platforms || "tiktok,instagram,youtube,x")
+ .split(",").map((s) => s.trim()).filter(Boolean);
+
+ const inputProps = {
+ bookendMode: opts.mode,
+ bookendTitle: opts.title || (opts.mode === "outro" ? "Follow for more" : ""),
+ bookendHandle: opts.handle || undefined,
+ bookendPlatforms: platforms,
+ bookendBg: opts.bg || "#0B0B0F",
+ bookendAccent: opts.accent || "#FFE000",
+ };
+
+ const bundleLocation = await getBundle();
+ const composition = await selectComposition({ serveUrl: bundleLocation, id: "Bookend", inputProps });
+
+ const seed = `${path.resolve(opts.output)}:${process.pid}`;
+ const id = crypto.createHash("md5").update(seed).digest("hex").slice(0, 12);
+ const silentVideo = path.join(os.tmpdir(), `bookend_${id}.mp4`);
+
+ console.log(`Bookend: ${opts.mode}, "${inputProps.bookendTitle}", ${platforms.join("+")}, ${durationInFrames}f`);
+
+ await renderMedia({
+ composition: { ...composition, durationInFrames, fps, width, height },
+ serveUrl: bundleLocation,
+ codec: "h264",
+ outputLocation: silentVideo,
+ inputProps,
+ crf: 16,
+ concurrency: Math.max(2, Math.min(os.cpus().length, 8)),
+ });
+
+ // Add a silent stereo audio track so concat/crossfade with the main clip
+ // (which has audio) doesn't fail on a missing audio stream.
+ const mux = spawnSync(
+ "ffmpeg",
+ [
+ "-y",
+ "-i", silentVideo,
+ "-f", "lavfi",
+ "-i", "anullsrc=channel_layout=stereo:sample_rate=44100",
+ "-c:v", "copy",
+ "-c:a", "aac",
+ "-b:a", "192k",
+ "-ar", "44100",
+ "-shortest",
+ path.resolve(opts.output),
+ ],
+ { stdio: "ignore", timeout: 60000 }
+ );
+ if (mux.status !== 0) {
+ throw new Error(`ffmpeg failed to add silent audio track (exit ${mux.status})`);
+ }
+ try { fs.unlinkSync(silentVideo); } catch {}
+
+ console.log(`OK ${opts.output}`);
+}
+
+main().catch((e) => { console.error("Bookend render failed:", e?.message || e); process.exit(1); });
diff --git a/remotion/src/Bookend.tsx b/remotion/src/Bookend.tsx
new file mode 100644
index 0000000..4ae4702
--- /dev/null
+++ b/remotion/src/Bookend.tsx
@@ -0,0 +1,180 @@
+import React from "react";
+import { AbsoluteFill, interpolate, spring, useCurrentFrame, useVideoConfig } from "remotion";
+
+export interface BookendProps {
+ mode: "intro" | "outro";
+ title: string; // intro: the headline; outro: the call-to-action ("Follow for more")
+ handle?: string; // e.g. "@yourbrand"
+ platforms: string[]; // subset of: tiktok, instagram, youtube, x
+ bg?: string; // background color
+ accent?: string; // accent color (active word / underline / icons)
+}
+
+const FONT = "'DM Sans', sans-serif";
+
+// --- Recognizable social glyphs as inline SVG (monochrome, fill=currentColor) ---
+const Glyph: React.FC<{ name: string; size: number; color: string }> = ({ name, size, color }) => {
+ const common = { width: size, height: size, viewBox: "0 0 24 24", fill: color } as const;
+ switch (name) {
+ case "tiktok":
+ return (
+ <svg {...common}>
+ <path d="M16.6 5.82A4.28 4.28 0 0 1 15.54 3h-3.1v12.4a2.59 2.59 0 1 1-2.59-2.59c.27 0 .53.04.78.12V9.77a5.7 5.7 0 0 0-.78-.05A5.69 5.69 0 1 0 15.54 15.4V9.01a7.34 7.34 0 0 0 4.3 1.38V7.3a4.28 4.28 0 0 1-3.24-1.48z" />
+ </svg>
+ );
+ case "instagram":
+ return (
+ <svg {...common}>
+ <path d="M12 2.16c3.2 0 3.58.01 4.85.07 1.17.05 1.8.25 2.23.41.56.22.96.48 1.38.9.42.42.68.82.9 1.38.16.42.36 1.06.41 2.23.06 1.27.07 1.65.07 4.85s-.01 3.58-.07 4.85c-.05 1.17-.25 1.8-.41 2.23-.22.56-.48.96-.9 1.38-.42.42-.82.68-1.38.9-.42.16-1.06.36-2.23.41-1.27.06-1.65.07-4.85.07s-3.58-.01-4.85-.07c-1.17-.05-1.8-.25-2.23-.41a3.7 3.7 0 0 1-1.38-.9 3.7 3.7 0 0 1-.9-1.38c-.16-.42-.36-1.06-.41-2.23C2.17 15.58 2.16 15.2 2.16 12s.01-3.58.07-4.85c.05-1.17.25-1.8.41-2.23.22-.56.48-.96.9-1.38.42-.42.82-.68 1.38-.9.42-.16 1.06-.36 2.23-.41C8.42 2.17 8.8 2.16 12 2.16zm0 3.68A6.16 6.16 0 1 0 18.16 12 6.16 6.16 0 0 0 12 5.84zm0 10.16A4 4 0 1 1 16 12a4 4 0 0 1-4 4zm6.4-10.4a1.44 1.44 0 1 1-1.44-1.44 1.44 1.44 0 0 1 1.44 1.44z" />
+ </svg>
+ );
+ case "youtube":
+ return (
+ <svg {...common}>
+ <path d="M23.5 6.2a3 3 0 0 0-2.1-2.1C19.5 3.5 12 3.5 12 3.5s-7.5 0-9.4.6A3 3 0 0 0 .5 6.2 31 31 0 0 0 0 12a31 31 0 0 0 .5 5.8 3 3 0 0 0 2.1 2.1c1.9.6 9.4.6 9.4.6s7.5 0 9.4-.6a3 3 0 0 0 2.1-2.1A31 31 0 0 0 24 12a31 31 0 0 0-.5-5.8zM9.6 15.6V8.4l6.3 3.6z" />
+ </svg>
+ );
+ case "x":
+ return (
+ <svg {...common}>
+ <path d="M18.9 2H22l-7.1 8.1L23.3 22h-6.6l-5.2-6.8L5.6 22H2.5l7.6-8.7L1 2h6.8l4.7 6.2zm-1.1 18h1.8L7.3 3.9H5.4z" />
+ </svg>
+ );
+ default:
+ return null;
+ }
+};
+
+const PLATFORM_LABEL: Record<string, string> = {
+ tiktok: "TikTok",
+ instagram: "Instagram",
+ youtube: "YouTube",
+ x: "X",
+};
+
+export const Bookend: React.FC<BookendProps> = ({
+ mode,
+ title,
+ handle,
+ platforms,
+ bg = "#0B0B0F",
+ accent = "#FFE000",
+}) => {
+ const frame = useCurrentFrame();
+ const { fps, width, height } = useVideoConfig();
+
+ // Title slides up + fades in
+ const titleProgress = spring({ frame, fps, config: { damping: 200 }, durationInFrames: Math.round(fps * 0.6) });
+ const titleY = interpolate(titleProgress, [0, 1], [60, 0]);
+ const titleOpacity = interpolate(titleProgress, [0, 1], [0, 1]);
+
+ // Icons pop in staggered, after the title
+ const iconBaseFrame = Math.round(fps * 0.5);
+
+ const iconSize = Math.round(width * 0.085);
+
+ return (
+ <AbsoluteFill
+ style={{
+ backgroundColor: bg,
+ fontFamily: FONT,
+ justifyContent: "center",
+ alignItems: "center",
+ padding: width * 0.08,
+ }}
+ >
+ {/* Accent bar */}
+ <div
+ style={{
+ width: interpolate(titleProgress, [0, 1], [0, width * 0.18]),
+ height: 10,
+ backgroundColor: accent,
+ borderRadius: 5,
+ marginBottom: height * 0.04,
+ }}
+ />
+
+ {/* Title / CTA */}
+ <div
+ style={{
+ transform: `translateY(${titleY}px)`,
+ opacity: titleOpacity,
+ color: "#FFFFFF",
+ fontSize: mode === "outro" ? width * 0.12 : width * 0.095,
+ fontWeight: 700,
+ lineHeight: 1.05,
+ textAlign: "center",
+ letterSpacing: -1,
+ textTransform: mode === "outro" ? "none" : "none",
+ }}
+ >
+ {title}
+ </div>
+
+ {handle && (
+ <div
+ style={{
+ opacity: titleOpacity,
+ color: accent,
+ fontSize: width * 0.06,
+ fontWeight: 700,
+ marginTop: height * 0.02,
+ }}
+ >
+ {handle}
+ </div>
+ )}
+
+ {/* Social icons (outro emphasises them; intro shows them small if present) */}
+ {platforms.length > 0 && (
+ <div
+ style={{
+ display: "flex",
+ gap: width * 0.05,
+ marginTop: height * 0.05,
+ alignItems: "center",
+ justifyContent: "center",
+ flexWrap: "wrap",
+ }}
+ >
+ {platforms.map((p, i) => {
+ const f = frame - (iconBaseFrame + i * Math.round(fps * 0.12));
+ const pop = spring({ frame: Math.max(0, f), fps, config: { damping: 12, mass: 0.6 }, durationInFrames: Math.round(fps * 0.5) });
+ const scale = interpolate(pop, [0, 1], [0.2, 1]);
+ const op = interpolate(pop, [0, 1], [0, 1]);
+ return (
+ <div
+ key={p}
+ style={{
+ display: "flex",
+ flexDirection: "column",
+ alignItems: "center",
+ gap: 8,
+ transform: `scale(${scale})`,
+ opacity: op,
+ }}
+ >
+ <div
+ style={{
+ width: iconSize * 1.6,
+ height: iconSize * 1.6,
+ borderRadius: iconSize * 0.4,
+ backgroundColor: "rgba(255,255,255,0.08)",
+ display: "flex",
+ alignItems: "center",
+ justifyContent: "center",
+ }}
+ >
+ <Glyph name={p} size={iconSize} color="#FFFFFF" />
+ </div>
+ <div style={{ color: "rgba(255,255,255,0.7)", fontSize: width * 0.028, fontWeight: 700 }}>
+ {PLATFORM_LABEL[p] || p}
+ </div>
+ </div>
+ );
+ })}
+ </div>
+ )}
+ </AbsoluteFill>
+ );
+};
diff --git a/remotion/src/Root.tsx b/remotion/src/Root.tsx
index 5ed5c53..5a92408 100644
--- a/remotion/src/Root.tsx
+++ b/remotion/src/Root.tsx
@@ -3,6 +3,7 @@ import "@fontsource/dm-sans/700.css";
import React from "react";
import { Composition, getInputProps } from "remotion";
import { CaptionedClip } from "./CaptionedClip";
+import { Bookend } from "./Bookend";
import { STYLES } from "./types";
import type { Word } from "./types";
@@ -14,6 +15,13 @@ const inputProps = getInputProps() as {
faceY?: number | null;
durationInFrames?: number;
fps?: number;
+ // Bookend props
+ bookendMode?: "intro" | "outro";
+ bookendTitle?: string;
+ bookendHandle?: string;
+ bookendPlatforms?: string[];
+ bookendBg?: string;
+ bookendAccent?: string;
};
export const RemotionRoot: React.FC = () => {
@@ -37,6 +45,22 @@ export const RemotionRoot: React.FC = () => {
faceY: inputProps.faceY ?? null,
}}
/>
+ <Composition
+ id="Bookend"
+ component={Bookend}
+ durationInFrames={durationInFrames}
+ fps={fps}
+ width={1080}
+ height={1920}
+ defaultProps={{
+ mode: inputProps.bookendMode || "outro",
+ title: inputProps.bookendTitle || "Follow for more",
+ handle: inputProps.bookendHandle,
+ platforms: inputProps.bookendPlatforms || ["tiktok", "instagram", "youtube", "x"],
+ bg: inputProps.bookendBg || "#0B0B0F",
+ accent: inputProps.bookendAccent || "#FFE000",
+ }}
+ />
</>
);
};
diff --git a/scripts/clip_studio.py b/scripts/clip_studio.py
new file mode 100644
index 0000000..cac3eb2
--- /dev/null
+++ b/scripts/clip_studio.py
@@ -0,0 +1,302 @@
+#!/usr/bin/env python3
+"""
+clip_studio — cut a precise fragment from a video and wrap it with a Remotion
+intro + "Follow for more" outro (with social icons).
+
+Cut a fragment by TIMESTAMP or by PARAGRAPH text (matched against the
+transcript). The fragment is rendered through podcli's existing face-crop +
+caption pipeline, then an intro card and an outro card are stitched on:
+
+ [ intro ] → [ captioned fragment ] → [ outro: Follow for more + icons ]
+
+Usage:
+ # by timestamp
+ python scripts/clip_studio.py VIDEO --start 12 --end 38 --caption-style hormozi
+
+ # by paragraph (finds the matching span in the transcript)
+ python scripts/clip_studio.py VIDEO --paragraph "the part where they talk about X"
+
+ # tune the bookends
+ python scripts/clip_studio.py VIDEO --start 0 --end 30 \
+ --intro-title "BIG IDEA" --outro-title "Follow for more" \
+ --handle "@yourbrand" --platforms tiktok,instagram,youtube,x \
+ --no-intro # skip the intro
+"""
+import argparse
+import json
+import os
+import subprocess
+import sys
+
+ROOT = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+# Capture the caller's CWD before we chdir into backend/ so relative video
+# paths on the command line still resolve correctly.
+_INVOKE_CWD = os.getcwd()
+sys.path.insert(0, os.path.join(ROOT, "backend"))
+os.chdir(os.path.join(ROOT, "backend"))
+
+FFMPEG = os.environ.get("PODCLI_FFMPEG", "ffmpeg")
+NODE = os.environ.get("PODCLI_NODE", "node")
+
+# Brand config: remembered handle / platforms / colors / outro title so they
+# don't have to be retyped each run. CLI flags override these; these override
+# the built-in defaults below.
+# Note: .env may set PODCLI_HOME to an empty string, so treat empty as unset.
+_PODCLI_HOME = os.environ.get("PODCLI_HOME") or os.path.join(ROOT, ".podcli")
+BRAND_PATH = os.path.join(_PODCLI_HOME, "brand.json")
+BRAND_DEFAULTS = {
+ "handle": None,
+ "platforms": "tiktok,instagram,youtube,x",
+ "outro_title": "Follow for more",
+ "accent": "#FFE000",
+ "bg": "#0B0B0F",
+}
+
+
+def _load_brand() -> dict:
+ try:
+ with open(BRAND_PATH) as f:
+ return json.load(f)
+ except (FileNotFoundError, json.JSONDecodeError):
+ return {}
+
+
+def _save_brand(data: dict):
+ os.makedirs(os.path.dirname(BRAND_PATH), exist_ok=True)
+ with open(BRAND_PATH, "w") as f:
+ json.dump(data, f, indent=2)
+
+
+def _probe_duration(path: str) -> float:
+ out = subprocess.run(
+ ["ffprobe", "-v", "error", "-show_entries", "format=duration",
+ "-of", "default=nw=1:nk=1", path],
+ capture_output=True, text=True,
+ )
+ try:
+ return float(out.stdout.strip())
+ except ValueError:
+ return 0.0
+
+
+def _transcribe(video: str, language: str | None):
+ """Transcribe (cached) and return the word list."""
+ from services.transcription import transcribe_file
+ print(" [transcribe] running Whisper...", flush=True)
+ res = transcribe_file(
+ file_path=video, model_size="base", language=language,
+ enable_diarization=False,
+ progress_callback=lambda p, m: print(f" {p}% {m}", flush=True),
+ )
+ return res.get("words", [])
+
+
+def _find_paragraph(words: list, phrase: str) -> tuple[float, float]:
+ """Find the time span of a paragraph by fuzzy-matching its text against
+ the transcript word stream. Returns (start, end) seconds."""
+ needle = "".join(c.lower() for c in phrase if c.isalnum() or c == " ").split()
+ if not needle:
+ raise SystemExit("Empty --paragraph text")
+ hay = [("".join(c.lower() for c in w["word"] if c.isalnum())) for w in words]
+
+ best_i, best_score = -1, 0
+ win = len(needle)
+ for i in range(0, max(1, len(hay) - win + 1)):
+ window = hay[i:i + win]
+ score = sum(1 for a, b in zip(window, needle) if a and (a == b or a.startswith(b) or b.startswith(a)))
+ if score > best_score:
+ best_score, best_i = score, i
+ if best_i < 0 or best_score < max(1, win // 3):
+ raise SystemExit(f"Could not confidently locate paragraph in transcript (best match {best_score}/{win} words).")
+ start = float(words[best_i]["start"])
+ end = float(words[min(best_i + win - 1, len(words) - 1)]["end"])
+ print(f" [paragraph] matched {best_score}/{win} words → {start:.1f}s–{end:.1f}s", flush=True)
+ return start, end
+
+
+def _render_fragment(video, start, end, words, style, crop, title, out_dir):
+ """Render the fragment with face-crop + captions via the existing engine."""
+ from services.clip_generator import generate_clip
+ print(f" [fragment] rendering {start:.1f}s–{end:.1f}s ({style}, crop={crop})", flush=True)
+ res = generate_clip(
+ video_path=video, start_second=start, end_second=end,
+ caption_style=style, crop_strategy=crop,
+ transcript_words=words, title=title, output_dir=out_dir,
+ clean_fillers=True, allow_ass_fallback=True,
+ progress_callback=lambda p, m: print(f" {p}% {m}", flush=True),
+ )
+ return res["output_path"]
+
+
+def _render_bookend(mode, title, handle, platforms, seconds, out_path, accent, bg):
+ cmd = [
+ NODE, os.path.join(ROOT, "remotion", "render-bookend.mjs"),
+ "--mode", mode, "--title", title,
+ "--platforms", ",".join(platforms),
+ "--seconds", str(seconds), "--output", out_path,
+ "--accent", accent, "--bg", bg,
+ ]
+ if handle:
+ cmd += ["--handle", handle]
+ print(f" [{mode}] rendering bookend...", flush=True)
+ r = subprocess.run(cmd, capture_output=True, text=True)
+ if r.returncode != 0 or not os.path.exists(out_path):
+ raise SystemExit(f"Bookend ({mode}) render failed:\n{r.stderr[-800:]}")
+ return out_path
+
+
+def _concat(parts: list[str], out_path: str, fps: int = 30):
+ """Concatenate parts (intro + clip + outro) in a single ffmpeg pass.
+
+ Every part is normalized to 1080x1920 @ fps with stereo 44.1k audio inside
+ the filtergraph, then joined with the concat filter. This gives an exact
+ summed duration (no crossfade-offset drift) and a single clean re-encode.
+ """
+ n = len(parts)
+ inputs = []
+ for p in parts:
+ inputs += ["-i", p]
+
+ fc = []
+ labels = []
+ for i in range(n):
+ # Normalize video: scale to fit 1080x1920, pad, set fps + SAR + format.
+ fc.append(
+ f"[{i}:v]scale=1080:1920:force_original_aspect_ratio=decrease,"
+ f"pad=1080:1920:(ow-iw)/2:(oh-ih)/2:black,setsar=1,fps={fps},format=yuv420p[v{i}];"
+ )
+ # Normalize audio to a common format so concat doesn't choke.
+ fc.append(
+ f"[{i}:a]aformat=sample_fmts=fltp:sample_rates=44100:channel_layouts=stereo[a{i}];"
+ )
+ labels.append(f"[v{i}][a{i}]")
+ fc.append("".join(labels) + f"concat=n={n}:v=1:a=1[v][a]")
+ filtergraph = "".join(fc)
+
+ cmd = [
+ FFMPEG, "-y", *inputs,
+ "-filter_complex", filtergraph,
+ "-map", "[v]", "-map", "[a]",
+ "-c:v", "libx264", "-crf", os.environ.get("PODCLI_CONCAT_CRF", "18"),
+ "-preset", "medium", "-pix_fmt", "yuv420p",
+ "-c:a", "aac", "-b:a", "192k", "-ar", "44100",
+ "-movflags", "+faststart",
+ out_path,
+ ]
+ r = subprocess.run(cmd, capture_output=True, text=True)
+ if r.returncode != 0 or not os.path.exists(out_path):
+ raise SystemExit(f"Concat failed:\n{r.stderr[-1000:]}")
+ return out_path
+
+
+def main():
+ ap = argparse.ArgumentParser(description="Cut a fragment + add intro/outro bookends.")
+ ap.add_argument("video")
+ ap.add_argument("--start", type=float, help="Fragment start (seconds)")
+ ap.add_argument("--end", type=float, help="Fragment end (seconds)")
+ ap.add_argument("--paragraph", help="Find fragment by matching this text in the transcript")
+ ap.add_argument("--language", default=None, help="Transcription language (e.g. es). Auto-detect if omitted.")
+ ap.add_argument("--caption-style", default="hormozi", choices=["hormozi", "karaoke", "subtle", "branded"])
+ ap.add_argument("--crop", default="face", choices=["center", "face", "speaker", "speaker-hardcut"])
+ ap.add_argument("--output", default=None, help="Final output path")
+ # bookends (defaults are None so we can tell what the user explicitly set;
+ # unset values fall back to the saved brand config, then BRAND_DEFAULTS)
+ ap.add_argument("--intro-title", default=None, help="Intro headline (default: derived from first words)")
+ ap.add_argument("--outro-title", default=None, help="Outro call-to-action (brand default: 'Follow for more')")
+ ap.add_argument("--handle", default=None, help="Handle shown on cards, e.g. @yourbrand")
+ ap.add_argument("--platforms", default=None, help="Comma-separated: tiktok,instagram,youtube,x")
+ ap.add_argument("--intro-seconds", type=float, default=2.0)
+ ap.add_argument("--outro-seconds", type=float, default=3.0)
+ ap.add_argument("--accent", default=None, help="Accent color hex (brand default: #FFE000)")
+ ap.add_argument("--bg", default=None, help="Background color hex (brand default: #0B0B0F)")
+ ap.add_argument("--no-intro", action="store_true")
+ ap.add_argument("--no-outro", action="store_true")
+ ap.add_argument("--save-brand", action="store_true",
+ help="Save the given handle/platforms/outro-title/accent/bg as the default brand and exit")
+ args = ap.parse_args()
+
+ # Resolve brand fields: CLI flag > saved brand.json > BRAND_DEFAULTS
+ brand = {**BRAND_DEFAULTS, **_load_brand()}
+ handle = args.handle if args.handle is not None else brand["handle"]
+ platforms_str = args.platforms if args.platforms is not None else brand["platforms"]
+ outro_title = args.outro_title if args.outro_title is not None else brand["outro_title"]
+ accent = args.accent if args.accent is not None else brand["accent"]
+ bg = args.bg if args.bg is not None else brand["bg"]
+
+ if args.save_brand:
+ saved = {"handle": handle, "platforms": platforms_str,
+ "outro_title": outro_title, "accent": accent, "bg": bg}
+ _save_brand(saved)
+ print(f" ✓ Brand saved to {BRAND_PATH}")
+ for k, v in saved.items():
+ print(f" {k}: {v}")
+ return
+
+ video = args.video if os.path.isabs(args.video) else os.path.join(_INVOKE_CWD, args.video)
+ video = os.path.abspath(video)
+ if not os.path.exists(video):
+ raise SystemExit(f"Video not found: {video}")
+ out_dir = os.path.join(ROOT, "data", "output")
+ os.makedirs(out_dir, exist_ok=True)
+
+ # Need a transcript if cutting by paragraph or if rendering captions.
+ words = _transcribe(video, args.language)
+
+ if args.paragraph:
+ start, end = _find_paragraph(words, args.paragraph)
+ elif args.start is not None and args.end is not None:
+ start, end = args.start, args.end
+ else:
+ raise SystemExit("Provide either --start/--end or --paragraph")
+
+ # 1. Fragment
+ fragment = _render_fragment(
+ video, start, end, words, args.caption_style, args.crop, "fragment", out_dir,
+ )
+
+ platforms = [p.strip() for p in platforms_str.split(",") if p.strip()]
+ parts = []
+
+ # 2. Intro (optional)
+ if not args.no_intro:
+ intro_title = args.intro_title
+ if not intro_title:
+ # derive a short headline from the first ~4 words of the fragment
+ frag_words = [w["word"] for w in words if start <= w["start"] < end][:4]
+ intro_title = " ".join(frag_words) or "Watch this"
+ intro = _render_bookend(
+ "intro", intro_title, handle, platforms,
+ args.intro_seconds, os.path.join(out_dir, "_intro.mp4"), accent, bg,
+ )
+ parts.append(intro)
+
+ parts.append(fragment)
+
+ # 3. Outro (optional)
+ if not args.no_outro:
+ outro = _render_bookend(
+ "outro", outro_title, handle, platforms,
+ args.outro_seconds, os.path.join(out_dir, "_outro.mp4"), accent, bg,
+ )
+ parts.append(outro)
+
+ # 4. Stitch
+ if args.output:
+ final = args.output if os.path.isabs(args.output) else os.path.join(_INVOKE_CWD, args.output)
+ else:
+ final = os.path.join(out_dir, "studio_final.mp4")
+ if len(parts) == 1:
+ # no bookends — fragment is the result
+ import shutil
+ shutil.copy(parts[0], final)
+ else:
+ _concat(parts, final)
+
+ dur = _probe_duration(final)
+ print(f"\n ✓ DONE {final}")
+ print(f" duration={dur:.1f}s (intro {'on' if not args.no_intro else 'off'}, "
+ f"outro {'on' if not args.no_outro else 'off'})")
+
+
+if __name__ == "__main__":
+ main()