Vox Agent

Vox Agent

Run the complete workflow from this folder. Do not read scripts, credentials, templates, or assets from another installed Skill.

Turn a one-line topic into a finished Vox-style paper-collage video: a bold, punchy, narrated explainer/ad where each beat is a torn-paper collage poster that comes alive, with captions. Keyframes use GPT Image 2 in Codex with a whole-project Liblib fallback, motion runs on Liblib/Kling, and assembly runs on local ffmpeg.

The look is the modern editorial paper-collage popularized by Vox explainers and creators like Stav Zilber / rom1trs: hand-cut paper cut-outs, torn edges, tape, halftone dots, newspaper clippings, bold flat color per beat, big cut-out headlines.

The core idea (read this first)

The Vox collage look and the collage motion are two different steps:

1. The look is born in the IMAGE step. Each beat is a finished collage poster made by a

text-to-image model. All the collage DNA (torn paper, cut-outs, halftone, bold color, headline text) lives in that image. If the image isn't a rich collage, nothing downstream will save it.

1. The motion is added after. By default an AI video model animates the whole poster (the

"living poster" path — simple, automated). For dramatic piece-by-piece assembly you cut the poster into parts and drive them with the local keyframe engine (advanced path).

Everything hinges on the prompts. Before writing any image or video prompt, read references/prompt-guide.md — it has the exact prompt structures that make the difference between "a real Vox collage" and "a moving PowerPoint".

First-use setup (mandatory; check before any production step)

• Run python3 scripts/configure_credentials.py --check at the start of the first task. If it

reports missing credentials, pause generation and direct the user to run python3 scripts/configure_credentials.py. It requests exactly three values in hidden terminal prompts: Liblib AccessKey, Liblib SecretKey, and Fish Audio API Key. Get the Liblib pair at https://www.liblib.art/apis and the Fish key at https://fish.audio/zh-CN/app/api-keys/. README screenshots show both locations. Never request that users paste secrets into chat.

• Save the three values only in the skill-local .env (mode 0600, git-ignored). Codex chat image

generation uses the current Codex session and needs no separate OpenAI API key. Never put credentials in beats.json or commit .env.

• Run python3 scripts/configure_credentials.py --check before a production job. It reports only

configured/missing status and never prints key values. Environment variables with the same names take precedence over the skill-local .env.

• Read references/liblib-api.md before changing motion routing. It documents the local-PNG

upload that bridges GPT Image 2 keyframes into Kling image-to-video.

• command -v ffmpeg ffprobe — required for assembly (brew install ffmpeg on macOS).
• python3 -c "import PIL" — Pillow, for captions/watermark overlays.

Standard workflow (topic → film)

This is the default, most-automated path. Every stage is one script, all driven by a single beats.json per project under out//.

1. Topic → beat map. First read references/beat-layer.md (the story layer) and pick a

narrative arc that fits the topic (timeline for history, pas/bab for ads, how_it_works for explainers, man_in_hole for transformations, …). Then write out//beats.json following that arc: beat-1 headline must be a ≤3s hook; beat count per duration (30s→6–8, 60s→10–12); split each beat into 2 shots (wide+detail) with per-shot camera_move VARIED across adjacent beats (never repeat; static on the payoff) and rich element_motion (see step 4). Each beat: narration, title_cn/title_en, scene, bg, feel, hook. This draft is the first mandatory approval gate — show the user the beat map before generating (the aspect-routing approximation in step 4 is the other one). Examples in examples/.

1. Pick the visual style (hybrid — do this BEFORE keyframes). Do not reuse one house style

for every topic. Read references/prompt-guide.md (§5 theme presets); pick 3–4 theme presets (styles.THEME_PRESETS: american-retro, swiss-modern, punk-zine, soviet-constructivist, wpa-propaganda, 70s-groovy, chinese-ink, atomic-age, newsprint-editorial) that fit the topic's era/culture/tone — or compose a custom theme by mixing the prompt-guide dimensions (medium/era/palette/type/finish) when none fit. Match the topic, not the language (an English film on Chinese history should look Chinese). A theme bundles the whole LOOK layer (idiom+palette+type+finish+mood+motion). Run a bake-off and let the user pick by eye — AI proposes, the library is the quality floor, the human decides. Set the pick as "theme": python3 scripts/style_bakeoff.py out/ american-retro,swiss-modern,punk-zine,atomic-age Set the chosen name as "collage_style" in beats.json (keyframes.py reads it).

1. Keyframes (the collage look). python3 scripts/keyframes.py out/

Set image_provider: "auto" (the default) and image_model: "gpt-image-2".

• Inside Codex: the script writes keyframes/gpt-image-2-manifest.json. Read

references/codex-parallel-keyframes.md, count manifest items as N, and request exactly N logical subagents: one image per agent. Start all available agents concurrently; when the runtime concurrency cap is lower than N, queue the rest and refill slots immediately. The root agent must orchestrate and validate, not generate manifest images sequentially. Each worker calls Codex image generation once, saves only its assigned PNG at the exact dest, and never edits JSON. Treat the project as one provider batch. If every item succeeds, rerun the script to register paths. If any item has a network/service failure, interrupt unfinished image workers, do not retry, and immediately run python3 scripts/keyframe_fallback.py out/ --all. That command creates every keyframe with Liblib and points the project at those files; earlier Codex files stay local but are not used. Never mix Codex and Liblib keyframes in one finished video.

• Outside Codex: auto resolves to Liblib and the script generates the complete keyframe

batch directly through Liblib. Video remains on Liblib/Kling. A content-quality rejection is not a network failure; stop for creative review instead of silently switching providers. Explicit image_provider: "openai" remains only as a legacy unattended API mode and is not part of the default three-key setup. Compose prompts with the 5-part structure in references/prompt-guide.md. Verify each poster looks like a real layered collage before animating. For Chinese or factual labels, prefer "title": false and add post_title/captions locally; generative lettering is not reliable enough for facts.

1. Motion. python3 scripts/clips.py out/

Animates each poster with Liblib Kling image-to-video. Two independent axes (see references/beat-layer.md §3, tested on our stack): • camera_move — ONE move per shot. Safe/default: {static, push_in, pull_out, pan, tilt, parallax}. Bold/experimental {orbit, dolly_zoom, roll, whip} are available, not banned — they can warp the flat art, so pair with constraints: loose and re-roll. Any custom phrase also passes through. • element_motion — where the energy lives; AI writes it per beat to fit that scene (not a template). Make it RICH (several elements moving) — be bold. A hero element flying across the frame (paper bird/plane/coins) is a great occasional punch on a key beat, not every shot (a flyer in every frame reads as a formula). motion_style = amplitude calm | punchy | max (the theme sets a default). constraints = strict (default: defect guards on — flat-2D, one-way, no-morph; best for clean text-heavy explainers) or loose (let the model explore 3D/bold moves; re-roll the misses). Headline text is hard-protected only on shots that have a title (detail shots without a headline are free to go wild). The default uses kling-v2-1, 5- or 10-second clips, and follows the GPT Image 2 keyframe aspect. Account submission limits are handled by a bounded queue with rate-limit/concurrency backoff.

1. Voice + music. python3 scripts/audio.py out/

Liblib's documented visual workflow API does not supply TTS or BGM. Background music defaults to the free local ACE-Step 1.5 server: if bgm_path is absent or invalid, audio.py calls scripts/music.py, generates a narration-friendly instrumental to audio/bgm.wav, and writes its absolute path back to beats.json. Install/start ACE-Step once with the commands in README; it needs no cloud music key. To reuse owned/licensed music instead, set a valid local bgm_path and the generator is skipped. For narration, either set one local narration_audio file per beat, or configure voice.provider: "fish" with a Fish library/clone reference_id. Fish credentials come only from FISH_API_KEY. If no voice block and no complete local narration are provided, default to Fish s2.1-pro-free, voice “历史故事·清晰”, reference_id: 6fc59d2b56cf402eb572934114c8d8aa. Use a detailed music.prompt for deliberate era/instrument/emotional control; otherwise the local prompt is derived from the topic and beat feelings.

The provider contract is intentionally split: one project uses one keyframe provider (Codex or Liblib), Liblib/Kling owns image-to-video, and Fish Audio owns narration.

1. Assemble. python3 scripts/assemble.py out/

ffmpeg: normalize + concat all shots, lay narration ducked under the music, burn captions, add the watermark. Narration defaults to continuous timing: the next sentence starts about 0.1s after the previous one ends, and visual cuts/captions move to those handoffs instead of waiting for a fixed shot slot. Set narration_timing.mode: "beat_locked" only when deliberate pauses between beats are required. Output out//final.mp4.

1. Verify. You can't read an mp4 directly — extract frames to jpg and look:

ffmpeg -ss -i final.mp4 -vf "scale=640:-1,format=yuvj420p" -frames:v 1 f.jpg

Cadence — how long shots should be

A common mistake is one long shot per beat. On a 9:16 / social piece especially, a static 10s shot reads as dead air. Aim for a cut every ~4–6 seconds:

• Shots run 3–6s; never let a single shot exceed ~7s — beyond that the AI motion has

nowhere to go and it feels static.

• A beat's narration is ~8–10s, so give each beat 2 shots (a wide establishing shot with

the headline + a detail cut-in without it). The narration plays continuously across both; the visual cuts mid-sentence. This is the single biggest rhythm win.

• So a ~60s film is typically ~6 beats × 2 shots × ~5s = 12 shots, not 6 × 10s.
• Reuse the wide keyframe as shot a; generate a tighter detail scene for shot b.

keyframes.py skips any shot that already has a keyframe_url, so adding b shots and re-running only generates the new ones.

Add a shots array to each beat (see schema). Give each shot its own short scene and motion; set "title": true only on the wide shot so the headline shows once per beat.

A-roll mode (talking-head → collage)

The default Liblib direct endpoint does not provide video-to-video restyling. Do not run the legacy A-roll scripts with the default provider; first configure an API-enabled Liblib custom video_workflow whose documented inputs accept the source video.

The standard workflow above is B-roll: a topic becomes AI-generated collage posters that get animated. A-roll is the reverse case — the user already has a real recorded talking-head video (a presenter speaking to camera) and wants it itself turned into the collage look, keeping their actual performance (face, lip movement, gestures) intact. There is no poster to generate; the "keyframe" is the presenter's own footage. Use A-roll when the user gives you a video file of themselves/a presenter talking, not a topic to write from scratch.

1. Transcribe + auto-segment. python3 scripts/asr_beats.py

Runs xai/stt-v1 on the source's own audio and cuts it into beats at sentence-ending punctuation or natural pause gaps (never exceeding ~9.5s, under Omni/Kling video-edit's 10s per-call cap). Writes beats.json with each beat's start/end/text — this is the same mandatory approval gate as the B-roll beat map: review it, set "theme" (run style_bakeoff.py the same way — the presenter's segment works fine as the bake-off source), and optionally fill in a content_beats string per beat (a sticker/stamp idea to layer in) before generating anything.

1. Generate. python3 scripts/aroll_clips.py [only_ids]

Cuts each beat's time range out of the source, uploads it, and re-styles it with a photographic paper-cutout sticker treatment on the presenter — her real likeness, lip movement, eye-line and gestures follow the source frame-for-frame; only the silhouette edge and the world around her are paper-collage. Default model is google/gemini-omni-flash/video-edit; any beat it rejects automatically retries on bytedance/seedance-2.0/reference-to-video (set via video_model/video_model_fallback in beats.json). Never ask the model to redraw or halftone-texture the face itself — that gets rejected regardless of how the prompt is worded (tried both a strong and a softened phrasing; both failed). Uses the same aspect-routing confirm gate as clips.py.

1. Assemble. python3 scripts/aroll_assemble.py

Muxes each generated clip with the original beat segment's own audio (never whatever audio the video model produced) so lip-sync is guaranteed regardless of which model handled that beat, normalizes every beat to one canvas, and concats into final.mp4.

C-roll mode (one photo → collage)

C-roll uses Codex GPT Image 2 editing with a local anchor image; it does not use Liblib image generation. Liblib is first used after the edited poster exists, when Kling animates it. This is the one exception to the non-Codex B-roll routing rule: outside Codex, stop and request a configured Liblib reference-image workflow instead of silently degrading identity/product fidelity.

The third input modality — "cutout roll". A-roll re-styles a talking-head VIDEO; B-roll generates everything from a topic; C-roll takes a single still PHOTO (a selfie, an avatar card, a product shot) and anchors it inside the collage world: the subject is cut out as a PHOTOGRAPHIC sticker — never redrawn — and per-beat posters are generated around it with an image-EDIT model, then animated through the normal clip stage. Use C-roll when the user gives you one photo and a topic: a personal explainer fronted by their own face, or a collage ad built around a real product shot (validated on both, 2026-07-17).

1. Beat map. Same as B-roll (references/beat-layer.md, same approval gate), plus the

C-roll fields in beats.json: "mode": "croll", "anchor_photo", "croll_subject" (portrait | product), and subject_wardrobe (portrait — lock the outfit or the paper-doll body drifts) or subject_desc (product). Set "title": false on shots — C-roll posters carry no headline; text belongs to captions. If there is no separate script, transcribe/derive narration first and let the audio's ASR timestamps define the beats (audio-first, like A-roll — not text-first like B-roll).

1. Anchored keyframes. python3 scripts/croll_keyframes.py

Writes

…

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: hongtao520
• Source: hongtao520/vox-agent
• License: MIT

Install and usage instructions live in the source repository linked above.