showtime's own launch video

The hero video for showtime's README, made with showtime. Every clip in it is a real render from this examples/ folder, every line in its terminal is real command output, and its narration was synthesized on the same machine that rendered it.

• Video: final.mp4, 1920x1080, 30 fps, 30.00 s, H.264 + AAC, 16.1 MB
• Square version: final.square.mp4, 1080x1080, 4.9 MB (from showtime deliver exports)
• Poster: poster.jpg: the end card at 29.3 s, also baked into frame 0.
• Captions sidecar: final.srt (17 cues from the narration's word times; not burned in)
• Share copy: share.txt
• Source: project/. Before re-rendering, copy the six example renders into project/media/ and rebuild the voice lines (project/media/README.md).

The request#

Make a 30-second launch video for showtime itself. Show real things: real renders from the other examples, real terminal commands, the studio board. Kokoro voice-over optional, music + SFX, 1920x1080 plus a 1080x1080 cut-down via showtime deliver exports. This is the hero video for the README. Quick mode.

Mode: quick. Nothing was asked. The opening line stated the assumptions: "Quick mode: 30 s 16:9 launch for showtime, confident default tone, Kokoro voice-over (it doubles as proof of the local voices), composed music with synced UI effects; real clips from examples 01-08, real command output, the studio board from example 09; a 1080x1080 version from the master."

What it demonstrates#

• Result first, then how. The hook types example 01's real request ("Make a 20-second launch video for Tidepool"), and the next scene plays that example's finished final.mp4. The terminal then shows the commands behind it, with their real output: showtime audio mix of example 01's mix (-14.01 LUFS -1.1 dBTP LRA 1.86, and its waveform drawn from the mixed file), a render progress line in showtime's own format (frames 600/600 100%), and showtime qa of example 01's final, verbatim (verdict: PASS (0 fail, 0 warn, 0 note)).
• A wall of real renders. Six <video> tiles play the finished examples (launch, explainer, vertical short, footage edit, data story, beat montage). All six are on screen from the cut, dimmed, and each one lights up as the narration names it (word times from timeline.json).
• The studio board. A showtime site capture --serve of example 09's exported board (studio-board.html), in a browser-frame whose page is live DOM (the capture as an <img> plus an overlay), so when the cursor clicks the recommended concept it visibly gets picked: an amber ring, a "picked" chip, the other two concepts dim. The camera pushes the whole browser, not the page inside it.
• A voice-led timeline. voice script --fit lands the narration on 25.5 s and retime --from-voice sets every scene length, places each line as its own track and ducks the music. The "local" scene draws the waveform of the sentence being spoken ("Even this voice was made locally.") with a playhead synced to it.
• One visual spine. The amber play mark from showtime's studio board, and a frame counter of this video itself (frame 0517 / 900 · 00:17.23) running under every scene.
• Honest copy. On-screen claims come from the README: "No cloud APIs. No API keys. No uploads." (short for the README's "no cloud AI services, no API keys, nothing you make is uploaded"; web features such as media search only fetch), "A local video studio for Claude Code", and the two /plugin install lines. There are no numbers other than real command output. (A "macOS · Windows · Linux" line was cut in the review round: only macOS x86_64 has actually been run so far.)

Storyboard#

Commands, in order#

Everything ran through skills/showtime/bin/showtime (below it is just showtime), from the work folder, so the job went to showtime-out/showtime-launch-<time>/ (written <job> below).

showtime doctor --quick
showtime job init showtime-launch --goal "Make a 30-second launch video for showtime itself ..." --mode quick --platform youtube \
  --assumed "30 s, 16:9 1920x1080, 30 fps, dom template; square 1080x1080 via deliver exports" --assumed "Kokoro voice-over, composed bed + synced SFX" ...

# material
showtime footage scenes examples/<each example>/final.mp4 --every 2 -o <job>/work/inv/<example>   # 8 contact sheets: picked clips and in-points
showtime site capture --serve examples/09-studio-trailer <job>/work/capture-board --page /studio-board.html --aspect 16:9
showtime qa examples/01-launch-tidepool/final.mp4 -o <job>/work/qa-ex01 --platform youtube    # the qa text shown in the terminal
showtime new dom fmt -d 4 && showtime render fmt --preview --out-dir .                        # scratch: the render progress line format
showtime audio mix examples/01-launch-tidepool/project/audio/mix.json -o <job>/work/ex01-mix.wav   # the mix line + waveform in the terminal

# project
showtime new dom <job>/project --title "showtime" --duration 30
showtime voice ipa "API keys showtime Claude"
showtime voice script <job>/project/narration.md -o <job>/project/voice --fit 25      # x4 while cutting words (1.15x was too fast)
showtime assets font "Inter" --copy-to <job>/project/fonts
showtime assets font "JetBrains Mono" --copy-to <job>/project/fonts
showtime audio styles ; showtime audio sfx-types ; showtime motion transitions
#   (wrote index.html, audio/mix.json; ~/.showtime/venv/bin/python data/peaks.py for the waveform data)
showtime retime <job>/project --from-voice <job>/project/voice/timeline.json --dry-run
showtime retime <job>/project --from-voice <job>/project/voice/timeline.json
showtime snap <job>/project --at 0.3,1.6,4.0,6.2,11.0,16.6,19.3,24.8,29.5 --width 960 --format jpg --sheet   # found overlapping wall tiles
showtime snap <job>/project --at ... --sheet                                          # 3 more rounds: hook box, wrapping, local layout
showtime check <job>/project                                                          # 6 WARN: Menlo glyphs, 5 short_text
showtime voice script <job>/project/narration.md -o <job>/project/voice --fit 25.5     # pause_after=0.9 on the terminal line
showtime retime <job>/project --from-voice <job>/project/voice/timeline.json           # then close set to 2.557 s so the total stays 30.00 s
showtime audio mix <job>/project/audio/mix.json -o <job>/project/work/mix-test.wav      # -14.0 LUFS, voice 15.1 dB over the music
showtime snap <job>/project --every 1 --cols 5
showtime job note <job> --stage plan ... ; showtime check <job>/project               # 1 WARN (see QA)

# final
showtime job note <job> --stage build ...
showtime render <job>/project --job <job>                      # final.mp4, 27.4 MB at crf 16; qa WARN frozen 24.93-27.50 s
showtime qa <job>
showtime snap <job>/project --at 25.0,27.3 ... ; showtime check <job>/project --find-first frozen   # after a waveform playhead + push-in
showtime render <job>/project --job <job> --crf 20             # final-2.mp4, 15.7 MB
showtime qa <job>                                              # PASS
showtime review-pack <job>                                     # round 1: the 08 tile flashed white at 17.6 s
showtime snap <job>/project --at 17.3,17.5,17.6 ... ; showtime snap <job>/project --at 13.0,16.0,17.3,17.6 ...   # in-point 11.4 -> 10.2
showtime render <job>/project --job <job> --crf 20             # final-3.mp4
showtime qa <job> ; showtime review-pack <job>                 # PASS; round 2
showtime deliver exports <job> --targets square                # auto = blur pad (bright blurred bars)
showtime deliver exports <job> --targets square --fit crop --out-dir <job>/work/square-crop   # cut the hook and terminal text
showtime deliver exports <job> --targets square --fit pad --pad-color "#0f0d0b" --out-dir <job>/work/square-pad   # chosen
showtime qa <job>/work/square-pad/final-3.square.mp4 --platform square
showtime render <job>/project --job <job> --crf 20             # final-4.mp4: poster moved to the end card (shipped before the review round)
showtime qa <job>                                              # PASS
showtime review-pack <job> --force-round                       # round 3, for the shipped file
showtime deliver exports <job> --targets square --fit pad --pad-color "#0f0d0b"   # final-4.square.mp4 (superseded by final-6.square.mp4)
showtime qa <job>/exports/final-4.square.mp4 --platform square --no-sheet
showtime captions <job>/project/voice/captions.words.json --style clean --aspect 16:9 -o <job>/work/captions.ass --srt <job>/final-4.srt
showtime qa <job> --no-sheet                                   # PASS, captions sidecar checked too
showtime job note <job> --stage deliver ...

# review round (outside critic's findings on final-4, see "Review round" below)
showtime snap <job>/project --at 2.3,2.417,15.6,16.5,17.4,18.5,19.41,20.3,21.173,25.9,26.9,27.1,27.25,27.5,28.721,29.3 --width 960 --format jpg --sheet --cols 4
showtime snap <job>/project --at 18.2,19.0,19.45,20.3,21.2,26.9,27.1 ... ; showtime snap <job>/project --at 19.45,20.5,24.15,24.4,26.9 ...
showtime check <job>/project                                   # 2 WARN (new one: the voice label after the shorter scene), fixed -> 1 WARN
showtime render <job>/project --job <job>                      # final-5.mp4 at crf 16 (29.9 MB), qa PASS
showtime render <job>/project --job <job> --crf 20             # final-6.mp4, shipped as final.mp4
showtime qa <job>                                              # PASS
showtime deliver exports <job> --targets square --fit pad --pad-color "#0f0d0b"   # final-6.square.mp4
showtime qa <job>/exports/final-6.square.mp4 --platform square --no-sheet
showtime review-pack <job> --force-round                       # round 4: the fixes, and the fixed mix-report context
showtime job note <job> --stage feedback --verified "critic round applied ..."

Timings#

The machine was a 6-core Intel i5 Mac (macOS 15), shared with about three other example jobs.

check estimated the render at 52-58 s; the real capture ran at about half that speed. The six H.264 tiles on the wall are seeked on every frame (seek 56-61 ms per frame per worker).

QA#

showtime qa on the shipped render (final-6.mp4): verdict PASS (0 fail, 0 warn, 0 note)

• file: h264 High, yuv420p, 1920x1080, 30 fps, faststart, BT.709; duration 30.00 s, matches showtime.json
• loudness: -14.1 LUFS integrated, -1.6 dBTP true peak (target -14 / -1); voice 15.1 dB above the music while speaking
• no silent gaps, no black stretches, no frozen stretches over 2.5 s, frame 0 has a picture (the baked end card)
• captions sidecar final.srt: 17 cues, readable line lengths
• no attribution required

showtime check on the project: PASS, 0 errors, 1 WARN (justified): short_text on the verbatim qa output line PASS loudness -14.1 LUFS (target -14), true peak -1.5 dBTP inside the terminal, on screen for 3.0 s where the rule asks 3.7 s. It is real output and was not shortened; the line the viewer needs, verdict: PASS (...), meets its reading budget. The notes are small labels (the frame counter, the browser-frame URL bar, the terminal title) and edge proximity of the left-hand role labels.

The square version (final.square.mp4, qa --platform square): WARN (0 fail, 2 warn), -14.0 LUFS, -1.5 dBTP. The two warnings are frozen 9.90-12.57 s and black_segment 21.70-22.37 s. They come from the pad: the 16:9 picture fills only 56 % of the square frame, so the terminal's slow push and the first "No cloud APIs." line change too few pixels for qa's whole-frame thresholds. The same stretches pass in the master.

Earlier findings that were fixed: a frozen 2.6 s hold in the "local" scene (qa, first render: added a playhead that sweeps the spoken sentence's waveform and a 6 % push-in), the 08 tile flashing white at 17.6 s (review pack round 1: moved the clip's in-point), text drawn with the system font Menlo (› and ✓ are not in the Inter subset: replaced with SVG), and five short_text warnings (a chip in the hook was cut, the terminal beats were moved earlier, and the narration got a 0.9 s pause after the terminal line).

Notes and deviations#

• Square version. deliver exports --targets square defaults to a blurred pad, which smeared bright tiles into the bars; --fit crop cut the hook, the terminal and the "No ..." lines. The shipped square uses --fit pad with the video's own ground colour (#0f0d0b), so the bars read as part of the frame. A true 1:1 layout would need its own render.
• Poster. The first finals baked the wall (17.2 s) as the poster. It was moved to the end card: the thumbnail then carries the name, and the NASA interview tile (a real person) is not the face of the product.
• Critic. Review rounds 1 and 2 were self-reviews (this run could not start a critic sub-agent). Round 3, on the shipped final-4, was judged by an outside critic; its findings were applied in the review round below.
• Terminal content. Commands and output are typeset in the page (not a screen recording). Each output line is copied from a real run: qa of example 01's final, audio mix of example 01's mix.json (the output file was named ex01-mix.wav, shown as mix.wav), and the render progress format from a real render. The render line counts up to 600 in showtime's format; it is not a recording of example 01's render.
• End card length. retime --from-voice keeps the end card's length, so the total came out at 30.52 s; the close was set to 2.557 s by hand to keep 30.00 s. The review round took 0.7 s from the local scene's hold and gave it to the close (now 3.257 s).
• Install lines. The end card shows the README's /plugin marketplace add faviovazquez/showtime and /plugin install showtime@showtime. The CHANGELOG still lists the final GitHub owner/name as a release-checklist item; update the end card if it changes.

Sources and licenses#

• Clips: the final.mp4 of examples 01, 02, 05, 06, 07 and 08 in this repository (used muted), and the poster frames of examples 01, 02, 04, 07, 08 and 09 (the hook's backdrop). Their own sources apply: Tidepool is a fictional app made as showtime sample material; example 06 is NASA/JSC footage (public domain, NASA Image and Video Library, jsc2023m000095) and example 08's images are NASA/ESA/CSA/STScI public-domain images, credited as a courtesy with no endorsement implied (see those examples' credits.txt). Example 08's music (CC BY 4.0) is not used here: the tiles are muted.
• Studio board: a capture of examples/09-studio-trailer/studio-board.html.
• Voice: Kokoro-82M (Apache-2.0), voice af_heart, generated locally. It is a synthetic voice.
• Music and sound effects: generated on this machine by showtime audio (compose style upbeat-tech, key C, and procedural effects). Nothing downloaded, nothing to credit.
• Fonts: Inter and JetBrains Mono, SIL Open Font License 1.1 (project/fonts/*/LICENSE.txt).
• Claims: showtime's README.md (local, no cloud AI services/keys/uploads, the tagline, the install commands).

Review round#

An outside critic judged the shipped final-4 (review pack round 3): ship after fixes, 1 blocker, 5 should-fix, 3 polish. Each finding was checked against the frames first (all confirmed), then fixed in one pass and re-rendered as final-6 (shipped here). Before/after stills at every cited time were compared; review/round-4/ in the job is the pack of the fixed video.

The shorter local scene made check flag the "this voice" label (2.3 s on screen, 2.7 s needed); it now reads "this voice · Kokoro, on this machine" and comes in 0.4 s earlier. After the round: check is back to the one justified WARN, qa on final-6 is PASS (-14.1 LUFS, -1.6 dBTP), and the square export has the same pad-only WARNs as before.

The project files on GitHub