anonymoussajjad/faceless-story-video

Faceless Story Video (Lite)

One topic in, one complete video plan out. The plan is ready to produce: the story, the pictures, the voice, the captions, the thumbnail and the upload text. This Lite version makes one story format: a moral / emotional story, 60-90 seconds, Urdu or English. (More formats live in the Pro version.)

Workflow

Follow the steps in order. Do not skip the story before the pictures: every image prompt and every caption line comes from the finished story, never the other way round.

1. Topic. Ask for one topic or idea if the user did not give one (for example: "a boy who returns a lost wallet", "a mother's sacrifice"). One topic, one story.
2. Story. Write the story first, in the scene JSON format in references/story-structure.md. 60-90 seconds, 6-12 scenes, hook in the first 3 seconds, one twist, ending with a clear moral or emotional payoff.
3. Image prompts. Write one image prompt per scene from references/image-prompts.md. Same character in every scene (same face, hairstyle and costume — repeat the description in every prompt). 3D-cartoon style, 9:16.
4. Voiceover script. Write the narration from references/urdu-narration.md. Urdu narration is written ONLY in Urdu script. Never Devanagari. One speaker per clip — never write dialogue between two people in one clip; split it into one-speaker clips.
5. Captions. One style in Lite: gold word-highlight kinetic captions (the spoken word turns gold as it is said). Produce an SRT file with the exact narration lines and timings.
6. Thumbnail + SEO. One thumbnail brief (see below) plus title, description and hashtags for Shorts, TikTok and Reels.
7. Export checklist. End with the checklist below so the video ships in the right format.

Output 1 — Story JSON

Return the story as JSON with exactly these keys: title, language ("ur" or "en"), style ("moral"), mood (one word), scenes (a list of 6-12 scenes). Each scene has: duration (seconds, number), visual (one short keyword, e.g. rainy_street), camera (push, pull, pan_left or pan_right), effect (rain, dust or embers), ambience (rain or empty), sfx (one short sound for the scene, e.g. rain, birds chirping — or empty if the scene is quiet), narration (1-2 spoken sentences, Urdu script for ur), caption (same as narration, or a shorter on-screen version), visual_description (2-3 sentences in English: WHERE the scene happens, WHAT objects are visible, WHAT the character is DOING), props (2-4 key objects, in English), shot (close_up, medium_shot, wide_shot or full_body — vary it, never the same shot twice in a row). Scene durations must add up to 60-90 seconds. See examples/example-story.json for a complete example.

Output 2 — Image prompts

One prompt per scene. Formula: [style] + [character, repeated exactly] + [place and action from visual_description] + [mood and light] + 9:16 vertical. Style default: 3D cartoon, Pixar-style, soft light, expressive faces. Never put text, words or letters inside an image prompt — AI image tools misspell text; all words go in the captions, never in the picture. Full rules in references/image-prompts.md.

Output 3 — Voiceover script

Narration is the story's narration fields, in order, one clip per scene. Urdu: natural spoken Urdu, short sentences, warm tone. Never write narration in Roman Urdu when the language is ur — Urdu script only. Never mix two speakers in one scene. Full rules in references/urdu-narration.md.

Output 4 — Captions (Lite style: gold word-highlight)

SRT format, one file. Each caption event matches one scene's narration, split into short lines (max 2 lines, max 5-6 words per line). The spoken word is highlighted gold as it is said; the rest stays white. Urdu captions stay in Urdu script, right-to-left, one whole line per event (never split an Urdu line word-by-word — the script breaks). Roman Urdu or English captions may highlight word by word.

Output 5 — Thumbnail brief + SEO

Thumbnail brief: one 9:16 image idea (the most emotional moment of the story, one character, one big feeling, no text in the image), plus 3 thumbnail title options (max 5 words each, Urdu or English matching the video language). SEO: 1 title (max 70 characters), 1 description (2-3 lines + the moral in one line), 8-10 hashtags (mix of broad: #Shorts #Story, and specific: #UrduStory #MoralStory). Keep every claim honest — never promise the video contains something it does not.

Output 6 — Export checklist

• 9:16 vertical, 1080x1920, 24-30 fps
• 60-90 seconds total
• Voice on every scene (no silent scene — a scene with no voice gets a warning, not a pass)
• Captions burned in or SRT sidecar, inside the safe zone (away from the TikTok/Reels buttons)
• Thumbnail exported as a still from the most emotional scene

Hard rules (never break these)

• Urdu narration is ALWAYS Urdu script. Never Devanagari, never Roman Urdu for ur stories.
• One speaker per clip. Never two people talking in one scene.
• No faces of real people, no real-person names, no graphic violence, no hate. Clean, family-friendly stories only.
• Never put text inside an image prompt.
• Never invent a fact about the video (views, platform, price) — only what the plan contains.

What this skill does NOT do

This skill writes everything except the rendered MP4. In Claude Code with ffmpeg/Remotion it can drive the render; in the Claude app, assemble the video in CapCut (or any editor) from the script, prompts, voiceover and SRT this skill produces. It does not generate images or audio by itself — it writes the prompts and scripts your tools use.

References (read when needed)

• references/story-structure.md — story arc, scene format, shot list, duration math.
• references/urdu-narration.md — Urdu script rules, narration style, pronunciation notes.
• references/image-prompts.md — consistent-character prompts, style, negatives.
• examples/example-story.json — a complete moral story in the exact JSON format.
• examples/example-output.md — what the finished output looks like for that example.