4-Level AI Narration Instructions: Context to Segments

The quality of AI-generated narration depends heavily on how much context you give it. Bitcut provides a 4-level instruction system that lets you guide the AI at every stage — from broad video context down to individual segment tweaks. The more detail you provide upfront, the less editing you need afterward.

Each level builds on the previous one, giving progressively finer control over the narration output:

  1. Video Context Setup — describe what the video is about, choose a style and tone
  2. Per-Block Instructions — give specific direction for each group of clips
  3. Block Refinement — adjust the generated narration for an entire block
  4. Per-Segment Regeneration — fine-tune individual narration segments

Level 1: Video Context Setup

Before generating any narration, Bitcut opens the Setup screen where you describe your video. This is the single most impactful step for narration quality — a detailed description here means the AI understands your content from the start.

1

Describe what the video is about

Fill in the main text area with a description of your video. Include the topic, location, participants, and key events. For example: "A walking tour of the old town in Tallinn, Estonia. I visit the main square, a medieval pharmacy, and the cathedral overlook." The more specific you are, the better the AI can write narration that matches your footage.

2

Choose a style

Select the content style that best matches your video. The available styles are:

  • Personal Vlog — first-person, casual, diary-like feel
  • Educational — informative and explanatory
  • Storytelling — narrative arc with beginning, middle, and end
  • Documentary — observational, factual, measured pace
  • Comedy — lighthearted, witty, playful
  • News — concise, factual, broadcast-style delivery
  • Podcast — conversational, in-depth, discussion-oriented
3

Set the tone

Pick a tone that defines how the narration should feel:

  • Warm — friendly and approachable
  • Professional — polished and authoritative
  • Casual — relaxed and informal
  • Energetic — upbeat and fast-paced
  • Emotional — evocative and heartfelt
  • Neutral — balanced, no strong personality
4

Add additional instructions (optional)

Use the Additional Instructions field for anything the style and tone pickers don't cover. Examples:

  • "Avoid common travel cliches"
  • "Focus on the atmosphere and sounds of each location"
  • "Mention Sanskrit terms when describing the yoga poses"
  • "Keep sentences under 10 words for fast pacing"
5

Tap Start

When everything looks right, tap Start at the bottom. Bitcut sends your context along with the clip visual descriptions to the AI for script generation.

Use voice input: Both the video description and additional instructions fields have a microphone button. Tap it to dictate instead of typing — useful when describing something complex or when you want to capture thoughts quickly.

Level 2: Per-Block Instructions

After the AI generates the initial script, your timeline is divided into blocks — groups of clips that share a narration section. At this stage, you can type a specific instruction for each block using the text bar at the bottom of the screen.

This lets you steer different parts of the narration in different directions without regenerating the entire script. For example:

  • Block 1 (arrival clips): "Focus on the landscape and first impressions"
  • Block 2 (market clips): "Mention the local food and describe the colors"
  • Block 3 (sunset clips): "End with a reflective, calm tone"

Each block instruction is independent — adding one to block 2 does not affect blocks 1 or 3. The AI regenerates only the block you provide an instruction for.

Block instructions are additive: They work on top of your Level 1 video context. The AI still knows the overall topic, style, and tone — the block instruction gives it additional focus for that specific section.

Level 3: Block Refinement

Once a block has generated narration, you can refine it further. Type a refinement instruction — such as "Make it more dramatic" or "Shorten this section" — and the AI rewrites all unlocked segments within that block.

Key details about refinement:

  • Targets unlocked segments only — if you have locked specific segments you are happy with, refinement skips them and only changes the rest
  • History is preserved — each refinement instruction is kept in context, so the AI builds on previous adjustments rather than starting from scratch
  • Multiple rounds — you can refine the same block several times, each time narrowing in on the result you want

This is useful when the overall direction is right but the execution needs adjustment. Rather than rewriting the block instruction from Level 2, a quick refinement nudges the existing narration in the right direction.

Level 4: Per-Segment Regeneration

For the finest level of control, you can regenerate a single narration segment with a specific instruction. Select a segment and provide direction like:

  • "Make this sound more excited"
  • "Add a question to hook the viewer"
  • "Mention that this was filmed at sunrise"
  • "Keep the same meaning but make it shorter"

Only the selected segment is regenerated — everything else stays untouched. The AI receives your full context (video description, style, tone, block instruction, and refinement history) along with the segment-specific instruction, so the result stays consistent with the rest of the narration.

Lock before you refine: If a segment is exactly right, lock it before running block-level refinement (Level 3). Locked segments are preserved during any regeneration above them.

Tips for Better Results

  • Be specific in the video description — "Travel vlog" tells the AI very little. "Three-day hike through the Scottish Highlands, starting at Fort William, ending at Inverness, solo trip in October rain" gives it real material to work with.
  • Match style to content — a cooking tutorial works better with the Educational style and Professional tone than with Comedy and Energetic. Let the content lead the choice.
  • Use additional instructions for exceptions — if your video is mostly educational but has a funny moment near the end, note that in the additional instructions rather than changing the overall style.
  • Work top-down — start with Level 1 context, then use Level 2 block instructions only where needed. Save Level 3 refinement and Level 4 per-segment regeneration for final polishing.
  • Keep refinement instructions short — "More energy" or "Less formal" is enough. Long, complicated refinement instructions can confuse the AI.
  • Regenerate rather than over-refine — if a block needs more than 2-3 refinement rounds, it may be faster to rewrite the Level 2 block instruction and regenerate from scratch.