--- title: "How to add build-up captions with Claude (tested prompt and script)" description: "The caption starts with the first spoken word and each new word joins it in place, wrapping onto new lines, until the whole sentence is on screen and then clears. Install one skill and ask Claude Code, or run the Python script we tested yourself." url: https://www.recharm.com/effects/build-up-captions --- # How to add build-up captions with Claude The caption starts with the first spoken word and each new word joins it in place, wrapping onto new lines, until the whole sentence is on screen and then clears. 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 build-up captions to input.mp4 Text height as a share of the frame width: 0.085 Where the caption block's center sits, from 0 (top) to 1 (bottom): 0.70 ``` ## 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 build-up captions 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 | | --------- | ------- | -------------------------------------------------------------------------------- | | `SIZE` | `0.085` | Text height as a share of the frame width. | | `Y` | `0.70` | Where the caption block's center sits, from 0 (top) to 1 (bottom). | | `WIDTH` | `0.86` | Widest a line can get before the next word wraps, as a share of the frame width. | | `LEADING` | `1.25` | Line spacing as a multiple of the text height. | | `HOLD` | `0.6` | Seconds the finished sentence stays on screen after its last word. | | `STROKE` | `0.014` | Outline thickness. 0 removes the outline. | ## Without the skill Don’t want to install the skill? Paste this prompt into Claude Code and Claude writes the script itself. ``` Add build-up captions to input.mp4 from its voiceover. Transcribe it with faster-whisper (model "base.en", device cpu, compute_type int8, word_timestamps=True) and split the words into sentences at words ending in . ? or !. Lay out each whole sentence first, in capitals with punctuation removed, as white bold text (Arial Bold) 8.5% of the frame width tall with a black outline 1.4% of the frame width thick, words separated by a space plus the outline thickness, wrapped into lines no wider than 86% of the frame width, each line centered horizontally, line spacing 1.25 times the text height, and the block of lines centered 70% of the way down the frame. Then show each word at its own place from the moment it starts being spoken, so words never move as later ones arrive, and keep every word of the sentence on screen until 0.6 seconds after its last word ends, then clear it. Nothing appears before the first word. Draw the text with Python and Pillow (not ffmpeg's drawtext filter) and 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 `build-up-captions.py` and run: ``` python3 build-up-captions.py input.mp4 output.mp4 ``` ``` #!/usr/bin/env python3 """Build-up captions: the voiceover is transcribed and each spoken word joins the caption, which stays until the sentence ends. Usage: python3 build-up-captions.py input.mp4 output.mp4 Needs ffmpeg, Pillow and faster-whisper (pip install pillow faster-whisper). The first run downloads Whisper's "base.en" model (about 150 MB); after that it runs offline. """ import json import math import subprocess import sys from pathlib import Path from faster_whisper import WhisperModel from PIL import Image, ImageDraw, ImageFont SIZE = 0.085 # text height as a share of the frame width Y = 0.70 # where the caption block's center sits (0 = top, 1 = bottom) WIDTH = 0.86 # widest a line can get, as a share of the frame width LEADING = 1.25 # line spacing as a multiple of the text height HOLD = 0.6 # seconds the full sentence stays after its last word STROKE = 0.014 # black outline thickness as a share of the frame width FONTS = ["/System/Library/Fonts/Supplemental/Arial Bold.ttf", # macOS "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf", # Linux "C:/Windows/Fonts/arialbd.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 transcribe(path): """Whisper's words, split into sentences at . ? or !""" model = WhisperModel("base.en", device="cpu", compute_type="int8") segments, _ = model.transcribe(path, word_timestamps=True) sentences, current = [], [] for segment in segments: for word in segment.words: current.append(word) if word.word.strip().endswith((".", "?", "!")): sentences.append(current) current = [] return sentences + [current] if current else sentences def layout(sentence): """Where each word of the finished sentence goes, wrapped into centered lines, so words never move as later ones arrive: [(text, x, y, start)].""" space = font.getlength(" ") + stroke # the outline would otherwise eat the gap between words lines, line = [], [] for word in sentence: text = word.word.strip(" .,!?").upper() width = sum(font.getlength(t) + space for t, _ in line) + font.getlength(text) if line and width > w * WIDTH: lines.append(line) line = [] line.append((text, word.start)) lines.append(line) step = w * SIZE * LEADING top = h * Y - step * (len(lines) - 1) / 2 placed = [] for row, line in enumerate(lines): x = w / 2 - (sum(font.getlength(t) for t, _ in line) + space * (len(line) - 1)) / 2 for text, start in line: placed.append((text, x, top + row * step, start)) x += font.getlength(text) + space return placed def overlay(t): img = Image.new("RGBA", (w, h)) draw = ImageDraw.Draw(img) for sentence, placed in zip(sentences, layouts): if sentence[0].start <= t < sentence[-1].end + HOLD: for text, x, y, start in placed: if start <= t: draw.text((x, y), text, font=font, anchor="lm", fill="white", stroke_width=stroke, stroke_fill="black") 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) font = ImageFont.truetype(FONT_FILE, round(w * SIZE)) stroke = round(w * STROKE) sentences = transcribe(src) layouts = [layout(sentence) for sentence in sentences] render(src, out) ``` ## More captions and text - [AI chat reply animation **AI chat reply animation**](https://www.recharm.com/effects/ai-chat-typing) - [Animated bar chart **Animated bar chart**](https://www.recharm.com/effects/animated-bar-chart) - [Animated checklist **Animated checklist**](https://www.recharm.com/effects/animated-checklist) - [Animated line graph **Animated line graph**](https://www.recharm.com/effects/animated-line-graph) - [Animated pie chart **Animated pie chart**](https://www.recharm.com/effects/animated-pie-chart) ## 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)