UGC scripts fail in two places: flat hook lines that never survive into video, and chatty LLM threads that burn credits before you approve a plan. Claude Fable 5 on Director targets the first problem. It sits in the higher credit usage tier in the picker, but one strong Fable rewrite often costs less than re-rendering a Kling 3.0 clip built from a thin prompt.
Claude Fable 5 is Anthropic's creative and expressive model in the Director catalog. On Picturesque, it is the default pick when you need founder-tone UGC, sensory adjectives, and hook timing that still reads well on the plan card before Nano Banana Pro and Kling 3.0 steps run.
Start with the Director LLM guide for picker basics. For head-to-head tradeoffs, read Director LLM comparison. For skill mechanics, see Director UGC ads skill guide.
What Fable 5 Does in Director
Director separates planning LLMs from generation models. Fable 5 handles chat, tool calls, and plan cards. It does not render pixels. Your UGC stills and motion still come from Picturesque image and video routes on the approved plan.
| Layer | Your pick | UGC job |
|---|
| LLM | Claude Fable 5 | Script, hook, shot list, tone notes |
|---|---|---|
| Image | Nano Banana Pro (typical) | Hero product still (15 credits) |
| Video | Kling 3.0 (typical) | Vertical motion from approved frame |
Credit Usage: Higher Tier Honesty
Claude Fable 5 shows a Higher usage badge in the Director LLM picker, same tier as Claude Opus 4.8, Kimi K3, Grok 4.5, and flagship GPT 5.4 / 5.5 rows. That is not a fixed per-turn price. Each assistant reply displays actual LLM credits consumed in the thread.
| Tier | Fable 5? | When it pays off |
|---|
| Lower usage | No | Use GPT 5.4 Nano or Mini for shot-list drafts first |
|---|---|---|
| Higher usage | Yes | Final script polish before expensive video steps |
When Fable 5 Is Worth the Higher Tier
Yes, use Fable 5 when:
- The brief needs emotional UGC tone ("kitchen counter confession", "founder demo energy")
- Hook lines must carry camera verbs and timing into Kling motion
- You are on the UGC ad skill or type /ugc with a vertical 9:16 deliverable
- Client copy must sound human, not bullet-list corporate
- You only need five hook variants to A/B (try Grok 4.5 — see Grok hooks guide)
- The thread is spec-heavy SKU matrices (try Claude Sonnet 4.6 or Opus — see Opus campaigns guide)
- You are still exploring credit ceilings (start on GPT 5.4 Nano — see Nano drafts guide)
Prompt Craft for UGC Scripts
Fable responds well to constraint blocks plus tone anchors. Paste product facts once, then ask for script structure.
Example brief (paste into Director):
> Product: oat protein bar, peanut butter chip. Format: 9:16 UGC, under 30s. Tone: tired parent, morning kitchen, not influencer-polished. Credit ceiling: 200 total including LLM. Skill: /ugc. Hero still on Nano Banana Pro, motion on Kling 3.0.
Example follow-up for Fable polish:
> Rewrite the plan with generation-ready prompts: hook in first 2s, one sensory detail per beat, camera move verbs for each shot. Keep disclaimers out of the hook.
Example hook-only pass:
> Give me three hook lines under 8 words each. Label them A/B/C. No emojis.
Weak prompt: "Write a UGC ad." Fable will answer, but the plan card may stay generic.
Strong prompt: Format, tone, credit ceiling, skill slash command, and which generation models to route.
UGC Skill Workflow with Fable 5
Step 1: Open Director and set the LLM picker to Claude Fable 5 (or GPT 5.4 Mini for a first pass, then switch).
Step 2: Type /ugc or open UGC ad skill from the skill picker.
Step 3: Attach a product still or paste SKU facts. State 9:16, duration, and brand words to avoid.
Step 4: Review the plan card. Confirm LLM credits so far and per-step generation totals (Nano Banana Pro, Kling 3.0).
Step 5: Approve renders. If motion misses tone, edit script lines in chat on Mini (cheaper) unless the underlying prompt structure is wrong.
Step 6: Export from the thread or Projects for paid social upload.
Manual fallback without Director: UGC product ad workflow.
Fable 5 vs GPT Mini for UGC
| Question | GPT 5.4 Mini | Claude Fable 5 |
|---|
| Usage tier | Lower | Higher |
|---|---|---|
| Script voice | Clean structure, lighter personality | Expressive, mood-forward |
| Best phase | Exploration, credit checks | Final plan before renders |
| Risk | Generic tone in video | Cost if you chat instead of ship |
Sample Fable 5 Script Output Shape
A strong Fable plan card for UGC often includes:
| Block | What Fable adds | Why it matters for Kling |
|---|
| Hook (0–2s) | Spoken line + facial cue | Sets opening motion energy |
|---|---|---|
| Beat 1 | Product reveal + hand motion | Drives I2V framing |
| Beat 2 | Benefit line + b-roll note | Separates A-roll from insert |
| CTA | Spoken close + on-screen text note | Timing for end card |
Pairing Fable with Nano Banana and Kling
Typical credit flow on the plan card (generation only):
| Step | Model | Notes |
|---|
| Hero still | Nano Banana Pro | 15 credits typical |
|---|---|---|
| Vertical motion | Kling 3.0 | Published clip rate on plan card |
| LLM turns | Fable 5 | Higher usage tier; per-message label |
Fable 5 vs Grok 4.5 for Hooks
| Job | Fable 5 | Grok 4.5 |
|---|
| 20 hook labels | Slower, more narrative | Faster grid (see Grok hooks guide) |
|---|---|---|
| Full UGC script | Best fit | Secondary |
| Usage tier | Higher | Higher |
Limitations to Be Aware Of
Higher usage tier. Fable is the wrong default for endless hook brainstorming.
Feature flag. Director, including the LLM picker, requires the Director tab on your account.
Separate credit pools. LLM turns and Kling / Nano Banana Pro steps both consume credits.
Usage-based LLM billing. Trust per-message credit labels in the thread, not static blog tables.
No external Claude subscription. Fable bills through Picturesque credits, not your Anthropic account.
Beta skills. Slash commands and plan-card formatting can change. Confirm model paths before client delivery.
Try Claude Fable 5 for UGC scripts on Picturesque
Pair expressive planning with the UGC ad skill, then approve generation on the models you trust.