Rebiha

Home › Datasets › Creative Writing

Creative Writing Fine-Tuning Dataset

Creative writing, storytelling, narrative techniques

Fine-tune a model on this dataset

$19 for the dataset. Training is priced separately by model, starting at $47.

What is in this dataset

943 instruction-response pairs covering creative writing, storytelling, narrative techniques. Each pair is a realistic question with a full written answer, in the JSONL format used for supervised fine-tuning. No synthetic filler, no duplicated rows.

Example pairs

Real rows from the dataset, shown in full.

How do I write an opening line that hooks the reader immediately? Can you show me a few examples and explain what makes them work?

A compelling opening line creates an instant contract with the reader: it promises tension, mystery, or a compelling voice. Focus on three techniques: 1) In medias res (plunge into action), 2) Anomaly (something off-kilter that begs explanation), 3) Voice-driven intrigue (a narrator whose perspective is irresistible). Example: 'The last time I saw my father, he was trying to sell a used coffin to a widow who hadn't ordered it yet.' This works because it combines anomaly (selling a coffin to a widow) with voice (the narrator's dry, almost morbid tone) and tension (the relationship with the father is fraught). Another: 'When the elevator doors opened, the man inside…

I'm struggling to make my characters feel real and not like cardboard cutouts. What techniques can I use to develop a three-dimensional character?

Three-dimensional characters emerge from the interplay of three layers: the public self (how they appear), the private self (what they believe about themselves), and the hidden self (what they refuse to face). Start with a contradiction. A character who is both fiercely independent and secretly terrified of being alone is more interesting than a purely brave one. Then, give them a wound—a specific past event they're still compensating for—and a desire that conflicts with their need. For example, a retired boxer (public: tough) who volunteers at a library (private: wants to be seen as gentle) but secretly hopes his estranged daughter will walk in (hidden: need for forgiveness). Show…

How do I write dialogue that feels natural but also carries subtext? My characters just say what they mean and it's boring.

Realistic dialogue is a dance of what is said and what is unsaid. The rule: characters should rarely answer a direct question directly. Instead, they deflect, evade, or answer a different question entirely. Subtext lives in the gap between intent and expression. Example: A husband and wife at a dinner party. She says, 'You're drinking quickly tonight.' He could say, 'I'm stressed about work,' but that's flat. Instead: 'The wine's better than I remembered.' Or: 'Am I? I hadn't noticed.' Or silence, then a change of subject. Each response tells us more about his actual state (avoidance, denial, tension) than a literal answer…

Who this is for

Writing assistants covering structure, dialogue, and character.

Which models this works with

The dataset is plain JSONL, so it works with any instruction-tuned open model. On Rebiha you can train it directly on:

Training uses LoRA, or QLoRA on a 4-bit base for larger models. Your base model's weights are never modified — training produces a separate adapter.

How it works

  1. Pick this dataset and a base model.
  2. Rebiha runs the job on a managed GPU. You set up nothing.
  3. Download your model: a ready-to-run GGUF, plus developer assets (adapter, tokenizer, configs).

Fine-tuning or RAG?

Use RAG when the model needs facts that change — your prices, your documents, today's inventory. Use fine-tuning when you want consistent style and structure without prompting for it every time. This dataset teaches the shape of a good answer in this domain, not facts to look up.

Related datasets

Start fine-tuning