--- title: "How to add a lyric video with Claude (tested prompt and script)" description: "The footage dims and each sung line rises into the middle of the frame in big white capitals, then drifts up and fades as the next line comes in. Install one skill and ask Claude Code, or run the Python script we tested yourself." url: https://www.recharm.com/effects/lyric-video --- # How to add a lyric video with Claude The footage dims and each sung line rises into the middle of the frame in big white capitals, then drifts up and fades as the next line comes in. Made by the script on this page, from our own test clip. ## Get this effect 1. ### Switch to Code in your [Claude Desktop App](https://claude.com/download) 2. ### Install the skill (once) ``` Install the Recharm video effects skill: download https://recharm-growth-staging.s3.us-east-2.amazonaws.com/website/downloads/recharm-video-effects.zip and unzip it into ~/.claude/skills ``` 3. ### Ask for the effect ``` Add a lyric video to input.mp4 Most words on one lyric line: 5 A pause this long (seconds) also starts a new line: 0.4 ``` ## Several videos, several effects, one ask Put your clips in the folder, mention them with @, and ask for everything at once. Claude runs one tested script after another on each clip, then joins them. Add a lyric video to **@clip1.mp4**, put a sale badge on **@clip2.mp4**, give **@clip3.mp4** a VHS look, then join all three with crossfades. ## Settings you can change Ask Claude to change any of these, or edit them at the top of the script. | Setting | Default | What it does | | ------------ | ------- | -------------------------------------------------------------------- | | `LINE_WORDS` | `5` | Most words on one lyric line. | | `GAP` | `0.4` | A pause this long (seconds) also starts a new line. | | `SIZE` | `0.12` | Text height as a share of the frame width. | | `Y` | `0.45` | Where the line's center sits, from 0 (top) to 1 (bottom). | | `RISE` | `0.05` | How far a line rises as it comes in, as a share of the frame height. | | `FADE` | `0.3` | Seconds a line takes to fade in or out. | | `HOLD` | `0.8` | Seconds the last line stays after its last word. | | `DIM` | `0.45` | How much the footage darkens while lyrics are up. | ## Without the skill Don’t want to install the skill? Paste this prompt into Claude Code and Claude writes the script itself. ``` Make a lyric video from input.mp4. Transcribe its vocals with faster-whisper (model base.en, word timestamps) and group the words into lines of at most 5 words, starting a new line after a pause longer than 0.4 seconds or a word ending in punctuation. Show one line at a time in white capitals (Arial Black or another heavy bold font, 12% of the frame width tall, wrapped to fit 86% of the width) with a soft black shadow, centered horizontally with its middle 45% of the way down. Each line is on screen from its first word's start until the next line's first word starts (the last line stays 0.8 seconds after its last word). A line fades in over 0.3 seconds while rising into place from 5% of the frame height lower, and fades out over its last 0.3 seconds while drifting up by the same amount. Darken the whole footage with a 45% black layer from the first line until the last line has gone, fading that in and out over 0.3 seconds. Draw the text with Python and Pillow (don't use ffmpeg's drawtext filter), put it on the video with ffmpeg, keep the audio, and save it as output.mp4. ``` ## Run it without Claude This is the exact script we tested. Save it as `lyric-video.py` and run: ``` python3 lyric-video.py input.mp4 output.mp4 ``` ``` #!/usr/bin/env python3 """Lyric video: the vocals are transcribed and each lyric line rises onto the middle of the screen in big type as it is sung. Usage: python3 lyric-video.py input.mp4 output.mp4 Needs ffmpeg, Pillow and faster-whisper (pip install pillow faster-whisper); the "base.en" model downloads on first run. For a song, separate the vocals first (for example with Demucs) and transcribe those for cleaner timing. """ import json import math import subprocess import sys from pathlib import Path from faster_whisper import WhisperModel from PIL import Image, ImageDraw, ImageFilter, ImageFont LINE_WORDS = 5 # most words on one lyric line GAP = 0.4 # a pause this long (seconds) also starts a new line SIZE = 0.12 # text height as a share of the frame width (shrinks so each row fits) Y = 0.45 # where the line's center sits (0 = top, 1 = bottom) RISE = 0.05 # how far a line rises as it comes in, as a share of the frame height FADE = 0.3 # seconds a line takes to fade in or out HOLD = 0.8 # seconds the last line stays after its last word DIM = 0.45 # how much the footage darkens while lyrics are up (0 = not at all) FONTS = ["/System/Library/Fonts/Supplemental/Arial Black.ttf", # macOS "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", # Linux "C:/Windows/Fonts/ariblk.ttf"] # Windows FONT_FILE = next((f for f in FONTS if Path(f).exists()), None) or sys.exit("No bold font found: add one to FONTS") def lyric_lines(src): """Word timings from Whisper, grouped into lines of at most LINE_WORDS, breaking early at pauses and punctuation.""" model = WhisperModel("base.en", device="cpu", compute_type="int8") segments, _ = model.transcribe(src, word_timestamps=True) lines, line = [], [] for word in (w for s in segments for w in s.words): if line and (len(line) == LINE_WORDS or word.start - line[-1].end > GAP or line[-1].word.strip()[-1] in ".,!?"): lines.append(line) line = [] line.append(word) lines.append(line) timed = [] for i, line in enumerate(lines): end = lines[i + 1][0].start if i + 1 < len(lines) else line[-1].end + HOLD timed.append((line[0].start, end, " ".join(w.word.strip(" .,!?") for w in line).upper())) return timed def line_image(text): """The line in white, wrapped onto rows that fit 86% of the frame width, over a soft shadow.""" size = round(w * SIZE) while True: font = ImageFont.truetype(FONT_FILE, size) rows, row = [], [] for word in text.split(): if row and font.getlength(" ".join(row + [word])) > w * 0.86: rows.append(" ".join(row)) row = [] row.append(word) rows.append(" ".join(row)) if all(font.getlength(r) <= w * 0.86 for r in rows) or size <= 16: break size -= 2 leading = round(size * 1.15) img = Image.new("RGBA", (w, leading * len(rows) + size)) draw = ImageDraw.Draw(img) for i, r in enumerate(rows): draw.text((w / 2, size / 2 + leading * (i + 0.5)), r, font=font, anchor="mm", fill="white") shadow = Image.new("RGBA", img.size, (0, 0, 0, 0)) shadow.putalpha(img.getchannel("A").filter(ImageFilter.GaussianBlur(size / 6)).point(lambda a: a * 0.9)) return Image.alpha_composite(shadow, img) def overlay(t): img = Image.new("RGBA", (w, h)) if lines[0][0] <= t < lines[-1][1]: dim = DIM * min(1, (t - lines[0][0]) / FADE, (lines[-1][1] - t) / FADE) img.paste((0, 0, 0, round(255 * dim)), (0, 0, w, h)) for start, end, art in lines: if not start <= t < end: continue come = min(1, (t - start) / FADE) # rises in from below... go = min(1, (end - t) / FADE) # ...and drifts up as it fades before the next line rise = ((1 - come) ** 2 - (1 - go) ** 2) * RISE * h layer = art.copy() layer.putalpha(art.getchannel("A").point(lambda a: a * come * go)) img.alpha_composite(layer, (0, round(h * Y - art.height / 2 + rise))) return img def probe(path): info = json.loads(subprocess.check_output(["ffprobe", "-v", "error", "-select_streams", "v:0", "-show_entries", "stream=width,height,r_frame_rate:format=duration", "-of", "json", path])) s = info["streams"][0] num, den = s["r_frame_rate"].split("/") return s["width"], s["height"], s["r_frame_rate"], float(num) / float(den), float(info["format"]["duration"]) def render(src, out): """Draw overlay(t) on every frame of src and save to out, keeping the audio.""" ff = subprocess.Popen(["ffmpeg", "-y", "-v", "error", "-i", src, "-f", "rawvideo", "-pix_fmt", "rgba", "-s", f"{w}x{h}", "-r", rate, "-i", "-", "-filter_complex", "[0:v][1:v]overlay=shortest=1,format=yuv420p[v]", "-map", "[v]", "-map", "0:a?", "-c:v", "libx264", "-crf", "18", "-c:a", "copy", out], stdin=subprocess.PIPE) for i in range(math.ceil(duration * fps)): ff.stdin.write(overlay(i / fps).tobytes()) ff.stdin.close() if ff.wait(): sys.exit("ffmpeg failed") src, out = sys.argv[1:3] w, h, rate, fps, duration = probe(src) lines = [(start, end, line_image(text)) for start, end, text in lyric_lines(src)] render(src, out) ``` ## More music-synced visuals - [AI voice orb **AI voice orb**](https://www.recharm.com/effects/ai-voice-orb) - [Audio-reactive blob **Audio-reactive blob**](https://www.recharm.com/effects/audio-reactive-blob) - [Audio-reactive particles **Audio-reactive particles**](https://www.recharm.com/effects/audio-reactive-particles) - [Audio-reactive text **Audio-reactive text**](https://www.recharm.com/effects/audio-reactive-text) - [Audio waveform **Audio waveform**](https://www.recharm.com/effects/audio-waveform) ## Find 3x _winning ads_ Recharm cuts your ad footage into tagged, searchable clips: hooks, testimonials, product shots. Find the shots that win, add the effect with Claude, and ship 3x more ad edits. [Start 14-day free trial](https://www.recharm.com/get-started)