35+ Claude Fable 5 Prompts for Video Creators: Scripts, HeyGen Hyperframes, Remotion, SEO
38 battle-tested Claude Fable 5 prompts for video creators — scripts, HeyGen Hyperframes, Remotion animations, SEO, descriptions, and editing workflows.
P
PromptsRushJune 10, 2026
•
On this page
13 sections
35+ Claude Fable 5 Prompts for Video Creators: Scripts, HeyGen Hyperframes, Remotion, SEO
Read
0%
17 min read
1 views
Claude Fable 5 is the first writing-tier model that genuinely understands video. The scripts read on camera, the SEO copy ranks, and — most importantly — it can hold its voice across the dozen-step pipeline that takes an idea from a Google Doc to a published file. We have been running it as the writing layer in our production stack for the last month.
This post is the working library. 38 prompts across six categories — long-form scripts, HeyGen Hyperframes direction, Remotion code generation, SEO, descriptions and metadata, and editing workflows. Every prompt is something we either run weekly or have saved in our text-expander because the lift versus a naive version is worth the keystrokes.
Each section opens with the hero prompts in copyable blocks. The rest are numbered and ready to lift.
Why Fable 5 Fits Video Production Specifically
Three properties of Fable 5 matter here:
Voice retention. Paste in 200,000 tokens of your past scripts and Fable 5 keeps your voice across a fresh draft. Other frontier models drift by minute three.
Code-and-prose duality. Fable 5 writes both copy and code well enough that you can use it for the script and the Remotion component file in the same pipeline.
Long-context multi-document context. 1M tokens means you can dump your channel brief, your past scripts, your competitor analysis, and your shot list into one prompt — and get back a script that respects all of it.
Pro tip: The single biggest unlock for all the prompts below is pasting 3-5 of your top-performing transcripts at the start of the prompt. Fable 5 will infer your voice without you having to describe it.
Category 1 — Long-Form Video Scripts (Prompts 1–8)
For 8-25 minute videos. These are tuned for camera delivery, not for reading on a page.
Prompt 1 — Full Video Script With B-Roll Cues
Ready to use
I have pasted three of my top-performing video transcripts above. Identify my voice — sentence rhythm, openings, vocabulary, what I avoid. Then write a fresh {{N}}-minute video script on {{topic}} in that voice. Structure: 8-second cold open hook, 30-second value promise, 4-5 main beats with B-roll cues in [brackets], a single mid-roll re-engagement at the midpoint, and a CTA. Match my cadence. Open with the conclusion, earn it back. Maximum {{N×140}} words.
Prompt 2 — Educational Deep-Dive From Source Material
Ready to use
I have pasted {{N}} reference documents above (papers, transcripts, articles). Write a 12-minute educational video script that synthesises them. Structure: hook stating the counterintuitive insight, three foundational ideas (one per source where possible), one common misconception, an applied example, a one-sentence summary. Cite source material inline using bracketed [Source N] tags so I can pull citations into the description.
Story-first script. Write a 10-minute script on {{topic}} using the story-spine structure (once there was, every day, until one day, because of that x3, until finally, ever since). The story is the spine; the educational payoff hangs off it. First line under 10 words.
Tutorial / how-to script. Write a step-by-step tutorial script for {{task}}. State the finished result in the first 15 seconds. List requirements. Walk through 5-8 steps with verbal cues for screen recording moments. End with the most common mistake people make. 8-10 minutes.
Interview prep script. I am interviewing {{guest}} about {{topic}}. Write 15 questions ordered for narrative arc — origin, controversial middle, forward-looking close. Include 3 follow-up questions for the moments most likely to need a second pass.
Reaction / commentary script. I am reacting to {{source content}}. Write a 7-minute script that adds substantive commentary, not surface reactions. My hot take in the first 15 seconds. Three specific moments to call out with analysis. A final verdict. Voice: opinionated, willing to be wrong.
Mid-roll re-engagement insert. At the midpoint of a video on {{topic}}, write a 25-second re-engagement that resets attention without breaking flow. Cannot say "before we continue." Must reference something the viewer just learned and tease something coming next.
CTA writer that does not sound like a CTA. Write the final 30 seconds of a video on {{topic}}. The CTA is to {{ask}}. Treat the viewer like an adult. Single ask. No "smash that subscribe button." Feels like a continuation, not an outro.
Category 2 — HeyGen Hyperframes Direction (Prompts 9–15)
HeyGen Hyperframes is HeyGen's framed-scene system — you write a script and it renders the avatar across multiple scenes with different framings (close-up, medium, B-roll cut-aways, captions, transitions). Fable 5 is genuinely good at writing the direction layer here because Hyperframes rewards specificity at the scene level.
If you are not already on HeyGen, our full HeyGen review covers the platform itself.
Prompt 9 — HeyGen Hyperframes Script With Scene Breakdown
Ready to use
Write a 2-minute HeyGen Hyperframes script for {{topic}}. Output format: numbered scenes. Each scene has four fields — Spoken: (the avatar's line, max 30 words), Frame: (close-up / medium / wide / B-roll), On-screen text: (the caption that should overlay), Transition: (cut / dissolve / push). Aim for 6-10 scenes. Alternate framings to avoid the talking-head monotony.
Prompt 10 — Hyperframes Ad Script With Pattern Interrupts
Ready to use
Write a 60-second HeyGen Hyperframes ad script for {{product}} targeted at {{persona}}. Output: 8-10 scenes. Every other scene must be a B-roll cut-away or visual pattern interrupt — not the avatar. The avatar's lines must work as standalone soundbites if pulled into a Short. Include the on-screen text overlay for each scene. End on a single CTA.
Hyperframes explainer with chapters. Write a 4-minute Hyperframes explainer on {{concept}}. Structure as four chapters of one minute each, each with its own chapter title for the on-screen graphic. 4-5 scenes per chapter. Avatar framings rotate through close-up, medium, and over-the-shoulder.
Hyperframes product demo. Write a Hyperframes script for a product demo of {{product}}. Half the scenes are screen-recording B-roll with the avatar overlaid in a corner; the other half are direct-to-camera explanations. Specify which scenes are which.
Hyperframes faceless channel script. Write a Hyperframes script for a faceless YouTube channel on {{topic}}. Avatar appears only in opening and closing scenes. Middle scenes are stock B-roll with voiceover and large on-screen text. Specify the B-roll search term for each middle scene.
Multilingual Hyperframes pack. Write a single Hyperframes script for {{topic}} in {{language A}}, then translate the spoken lines into {{language B}} and {{language C}} while keeping the scene structure identical. HeyGen renders the avatar in each language from the same scene plan.
Hyperframes UGC-style ad. Write a Hyperframes script that mimics UGC creator content — first-person, handheld feel, conversational. The avatar plays a real-person creator reviewing {{product}}. Include cues for which scenes should look "phone-shot" vs "polished."
Hyperframes re-cut from existing footage. Here is the transcript of a long-form video I already published. Re-cut it into a 90-second Hyperframes script that preserves the strongest moments. Cut anything that does not earn its time. Specify which sentences become the avatar's lines and which become on-screen text.
Category 3 — Remotion Prompts for Animated Videos (Prompts 16–22)
Remotion is the React-based programmatic video framework. You write TSX components, the framework renders frames, FFmpeg stitches the final MP4. Fable 5 writes Remotion code fluently — composition files, animation primitives, dynamic data-driven scenes. This is the path for explainer videos, data visualisations, lyric videos, and any animation that has to scale across many topics.
Prompt 16 — Remotion Composition From a Script
Ready to use
I have pasted a video script above. Write a complete Remotion composition in TypeScript that renders it as a 1080p animated video. Use Sequence components to time scenes. Use spring() for natural motion. Add an opening title card with the video title, a per-scene caption that types on, and a closing CTA card. Output: a single Composition.tsx file plus the registerRoot setup. Include comments only where the timing math is non-obvious.
Build a Remotion template for {{video type}} that accepts a JSON props object with the following shape: {{schema}}. The template must validate the props with Zod, render N dynamic scenes from an array, and support optional fields gracefully. Output: Composition.tsx, schema.ts, and a sample props.json. Make the template render correctly for both 16:9 and 9:16 aspect ratios via a prop flag.
Remotion lyric video. Write a Remotion composition that renders a lyric video for {{song / spoken text}}. Use the audio file's waveform as a backdrop. Render lyrics word by word with springs. Provide the useAudioData hook setup and the Sequence math for word timing.
Remotion animated explainer scene. Write a single Remotion scene component that animates {{concept}} for an explainer video. Use interpolate for opacity and Y-position, spring for snap-in motion, and absolute-positioned SVG for the diagram elements. 5-second scene at 30fps.
Remotion product showcase from a list. Build a Remotion composition that takes an array of {{N}} products with name, image URL, and price, and renders a 30-second showcase. Each product gets 2.5 seconds. Include a counter ("1 of N") in the corner and a final summary scene.
Remotion social-media render queue. Set up a Remotion project that renders the same composition in three aspect ratios (16:9, 9:16, 1:1) and three audio tracks (full, no music, captions-only) from a single command. Provide the package.json scripts, the render.ts file, and the composition prop hooks.
Remotion captioned interview clip. Write a Remotion composition that takes a transcript with word-level timestamps and renders animated captions in TikTok-style — two words at a time, large bold typography, highlighted current word in {{accent color}}. Include the timestamp parsing logic and the per-word Sequence wrapping.
Remotion programmatic series. I want to produce 50 short videos from a CSV of topics. Build a Remotion + Node.js setup that reads the CSV, renders each video with the topic injected as a prop, and outputs to /out/{topic-slug}.mp4. Include the render-all script and the error handling for individual failures.
Category 4 — SEO Prompts for Video Discovery (Prompts 23–28)
YouTube SEO and Google Video SEO have meaningfully diverged in 2026. These prompts target both surfaces.
Prompt 23 — YouTube Title + Tag Pack From a Script
Ready to use
I have pasted my finished video script above. Generate: 10 title candidates (under 70 chars, lead with the search-intent keyword), 25 tag candidates ranked by search-volume potential and competition, and a one-line video summary for the description hook (under 150 chars). Mark the title and three tags most worth A/B-testing. Skip generic SEO sludge.
I am planning my next 12 videos on the niche {{niche}}. Produce a search-intent cluster map: 4 pillar topics, 3 video ideas per pillar, the primary keyword and one long-tail variant per video, and the cross-link strategy between videos. Mark which videos serve "I want to learn" intent vs "I want to buy" intent.
Title rewrite for CTR. The current title is "{{title}}". Score it on clarity, curiosity, and honest delivery. Rewrite if any axis scores below 7. Provide 5 variants under 70 characters. Mark the strongest.
Tag research from a competitor. Here is the title and description of a competitor video that ranks for {{keyword}}. Identify the tag patterns they likely used. Suggest 20 tags I should try, ranked by signal-vs-noise.
Google Video SEO description. Write a description for a video on {{topic}} optimised for Google Video search specifically. Include the primary keyword in the first 150 characters, a structured FAQ-style middle section, and timestamped chapters. Total length 400-600 words.
Search-intent gap finder. I have pasted my last 20 video titles. Find the search-intent gaps — topics my channel is positioned to rank for but has not covered. List 8 candidates ranked by competitive opportunity.
Prompt 29 — Full Description With Chapters & Links
Ready to use
I have pasted the script and the YouTube affiliate links for the tools I mention. Write a complete YouTube description: 150-char hook with the primary keyword, 3-paragraph summary of what the video covers, timestamped chapters (auto-generated from the script transitions), a "tools mentioned" list with the affiliate links, and a single CTA at the bottom. Keep it under 1,500 characters total.
Chapter markers from a transcript. Here is the timestamped transcript of my video. Extract 6-10 chapter markers. For each: the timestamp, the chapter title (under 40 chars), and a one-line description for accessibility tools. Skip chapters under 30 seconds.
Pinned-comment writer. Write a pinned comment for a video on {{topic}}. Should invite engagement, reference a specific moment in the video, and ask a question that creates a comment thread. Under 250 characters.
End-screen script. Write the spoken script for the final 15 seconds of a video on {{topic}}. The end screen has two video tiles and a subscribe button. Tease both videos without summarising them. Voice: the viewer just enjoyed the video, treat them as a friend.
Localized description pack. I have pasted my English description. Translate it into {{language A}}, {{language B}}, {{language C}}. Keep the affiliate links unchanged. Adapt the hook to feel native in each language, not literal.
Community tab post. Write a Community tab post to tease an upcoming video on {{topic}}. Under 200 characters. Includes a question. Designed to drive a poll-style engagement.
Category 6 — Editing & Production Prompts (Prompts 34–38)
Prompt 34 — B-Roll Shot List From a Script
Ready to use
I have pasted my finished video script above. Extract a B-roll shot list. For each shot: timestamp range, what is on screen, the search term to find it on a stock library (Storyblocks / Artgrid / Envato), and a fallback search term if the primary returns nothing useful. Group shots that could share a single longer take.
Caption pass for accessibility. Here is the raw transcript with auto-generated punctuation. Rewrite as accessibility captions: max 2 lines per caption, max 32 chars per line, sensible break points, sound effects in [brackets]. Output as SRT format.
Editing brief for an editor. Write an editing brief for my editor for a video on {{topic}}. Include: the energy curve (where to cut tight vs let breathe), the moments to add zoom-in punches, the on-screen text overlay style guide, and the three specific edits that will make or break the video. Under 600 words.
Music brief. Write a music brief for a video on {{topic}}. Genre, BPM range, instrumentation, energy curve across the runtime, three reference tracks (with timestamps in those tracks where the vibe is closest), and the moments where music should drop out for impact.
Thumbnail concept brief. Write a 3-option thumbnail concept brief for a video on {{topic}}. For each option: the visual idea, the text overlay (under 5 words), the emotion the thumbnail communicates, and which audience segment it skews toward. Mark which I should A/B-test first.
The Full Video Production Pipeline With Fable 5
Single prompts are useful. Pipelines are what change your weekly output. This is the chain we run end to end.
Research and intent (Prompt 24). Map the search-intent cluster for your niche to pick which video to make.
Script (Prompt 1 or 2). Draft the full script in your voice.
Hyperframes or Remotion (Prompts 9-15 or 16-22). Convert the script into the production format your channel uses — HeyGen Hyperframes for talking-head/avatar, Remotion for animated/data-driven, both for a mixed stack.
SEO pack (Prompt 23). Generate the title, tag, and summary candidates from the finished script.
Description and chapters (Prompts 29-30). Pull the chapters from the script transitions; write the description around them.
B-roll list (Prompt 34). Hand the editor the shot list so they can pull stock or shoot before the edit starts.
Thumbnail brief (Prompt 38). Three thumbnail concepts to A/B test before publish.
Distribution copy. Pinned comment (31), end-screen script (32), Community tab tease (33).
That is one video. Done in 90-120 minutes of Fable 5 time, instead of a working day of manual writing. For animated videos through Remotion, the same pipeline plus the code-generation stage produces a finished MP4 with zero manual After Effects work. We covered the wider workflow in How to Create an AI Avatar for YouTube.
The Two Production Lanes
Most video creators in 2026 fall into one of two production lanes. Pick yours; the right prompts above will differ.
Lane
Best fit
Core prompt set
Render layer
Talking-head / avatar
Personal-brand creators, course creators, faceless channels with consistent host avatars
Explainer channels, data-driven content, lyric videos, high-volume series
Cat 1 + Cat 3 (Remotion) + Cat 5
Remotion + Lambda renders
Most serious channels run both lanes — talking-head for evergreen content and animated for high-volume topical content. The prompts above are designed to compose into either lane.
Patterns That Make These Prompts Work
If you read the 38 prompts carefully, they all share the same set of moves. Steal these:
Output format spelled out. Every prompt that produces structured output (Hyperframes scenes, Remotion code, SEO packs) names the format. Vague prompts produce vague outputs.
Bracketed cues for production. Asking for B-roll cues, on-screen text, transitions in [brackets] makes the script directly usable by an editor or a code generator. Do not skip this.
Constraints kill filler. "Under 70 chars", "under 30 words", "max 32 chars per line" — every constraint removes a model's default lazy move.
Past performance as context. Almost every prompt is stronger if you paste in 3-5 transcripts or descriptions from your top videos. Voice inference beats voice description.
One ask per prompt. Script, SEO, and description in one prompt produces mediocre versions of all three. Three prompts produce three good outputs.
Reverse-engineering existing wins. Several prompts above start with "here is something that worked — replicate the pattern." This is the most underrated technique for sustained channel growth.
Pro tip: Save your three highest-performing transcripts as Project Knowledge inside a Claude Project named "Voice." Every script-class prompt from there forward is one message away from your real voice.
Where to Run These
Claude.ai web app. The most natural surface. Save each category as a Project with your past transcripts in project knowledge.
Anthropic API. For wiring into a production tool. System prompt holds the voice instructions; user message holds the topic.
Claude Code. If you are running the Remotion lane, this is the right surface — Fable 5 writes the TSX, runs the render, fixes the type errors, all in one loop.
Multi-model aggregators like Genspark. Useful when you want to A/B Fable 5 against GPT-5.5 or Gemini Flash for a specific prompt — see our model comparison for when each one wins.
The Verdict
Fable 5 is the first model where the writing layer of video production stops being the bottleneck. Scripts in your voice. SEO copy that ranks. Hyperframes direction at scene granularity. Remotion components that compile. Descriptions, chapters, briefs — every text artifact a video pipeline produces, in one model, with a consistent voice across the chain.
The 38 prompts above are not the ceiling. They are the floor. Run them for thirty days, save the variants that worked, and you will have your own production library that beats anything you can find on a prompt-pack site. That is the actual unlock — not the prompts themselves, but the habit of treating prompts as a working library you grow.
Most work on Opus 4.7 and Sonnet 4.6 too — the structure carries them. But Fable 5 is specifically tuned for prose quality, voice retention, and creative output, which is exactly the workload these prompts target. The Hyperframes direction prompts and the long-form script prompts show the largest gap. Remotion code-generation prompts run well on both Fable 5 and Opus 4.7.
Hyperframes is HeyGen's scene-based composition system. Instead of one continuous avatar talking head, your video is structured as a sequence of framed scenes — close-up, medium, wide, B-roll cut-aways — with on-screen text and transitions between them. It is the feature that makes HeyGen avatar videos feel produced rather than generated. Fable 5 writes the direction layer (the scene plan) particularly well.
Remotion is a React-based framework that lets you build videos in TypeScript. Components render frames, FFmpeg stitches the MP4. The win for AI workflows is that the entire video is code — meaning you can parametrise it. One template renders 50 videos from a CSV. One animation component reuses across every video. Fable 5 writes the Remotion code fluently, so the AI handles both the script and the render layer.
If you have a talking-head channel, start with Prompt 1 (full video script) followed by Prompt 9 (Hyperframes scene breakdown). If you run an animated channel, start with Prompt 1 followed by Prompt 16 (Remotion composition from a script). Either way, save your past transcripts inside a Claude Project first so the voice work happens automatically.
Just the script — and the production direction, SEO pack, descriptions, and (for the Remotion lane) the code that renders the video. Fable 5 outputs text and audio, not video. The video itself is rendered by HeyGen, Hedra, Remotion, or a similar tool fed by Fable 5's output.
Save 5-10 of your top-performing transcripts as standing Project Knowledge inside a Claude Project. Every script prompt from there forward inherits the voice automatically. Update the project knowledge every quarter with your newest top performers — the voice drifts subtly over time and you want the model to track that drift.
If you are publishing more than one talking-head video per week, almost always yes. Hyperframes eliminates the per-video editing overhead that kills small creators' consistency. The cost-per-video amortises quickly against the alternative of hiring an editor. If you publish less often than weekly, the cost-benefit is closer — we covered the full breakdown in our HeyGen review.
Yes. Remotion Lambda renders compositions on AWS at roughly $0.0001 per second of output. A 5-minute video costs about $0.03 to render. No local hardware required beyond a laptop that can run the dev server. The setup is documented well — Fable 5 will write the IAM and config scaffolding for you if you ask.
The SEO prompts above are better at intent mapping and copywriting — title rewrites, search-intent clustering, description copy. They are worse at keyword volume data — you still want a TubeBuddy / VidIQ subscription to verify the tag volumes. The right workflow uses both: Fable 5 for the writing layer, the SEO tool for the data layer.
Prompt 24 — the search-intent cluster map. Most creators write videos one at a time. The cluster map prompt makes you plan twelve videos at once around four pillar topics, which is what actually builds channel authority. Run it once a quarter. The compounding effect on watch time is large.