5 Mistakes Killing Your AI Realism in 2026 (Easy to Fix)

What this lesson covers

Thomas Lundström details five common pitfalls in AI filmmaking and walks through how he produced a realistic hot sauce commercial. He demonstrates using Claude for structured prompting, creating standardized character reference sheets with Nano Banana, and managing Seedance 2.5 generations using draft mode.

Key takeaways from the creator

AI-extracted notes, not independently verified product claims. Timestamp links let you check each point in the original video.

1. 01:45 ↗

   Use an LLM like Claude to co-develop concepts and convert plain language into structured prompts rather than writing complex prompt code manually.

2. 03:21 ↗

   Standardize character reference sheets with a facial close-up alongside front and back full-body views to preserve identity consistency across AI generations.

3. 04:05 ↗

   Evaluate image generation models carefully for reference assets, preferring models that avoid telltale synthetic noise patterns like those produced by GPT image 2.5.

4. 05:47 ↗

   Structure Seedance 2.5 video prompts into style, scene/character descriptions, and timestamped shot-by-shot cuts with explicit dynamic camera moves.

5. 07:08 ↗

   Break multi-scene scripts into separate generation clips (e.g., 15-second and 21-second segments) to maintain directional clarity and prevent the video model from getting confused.

6. 07:38 ↗

   Generate initial video tests using Seedance 2.5 Draft Mode to confirm camera motion and staging before committing full rendering credits.

Workflow outlined in the video

1. Set up an AI assistant workflow to brainstorm concepts and generate structured video prompts with style prefixes and cut breakdowns.
2. Create character reference templates containing facial close-ups and dual full-body angles using photorealistic models like Nano Banana 2.1.
3. Specify camera movement, transitions, and speed ramps explicitly in video prompts to avoid static shots.
4. Divide sequence prompts by location or scene changes rather than attempting a single long generation.
5. Execute initial test shots in draft mode to evaluate composition and movement before running final high-credit generations.

Before you use this workflow

These notes describe the source video at its publication date. Model access, pricing, connectors and interfaces may have changed. Check the original source and the provider’s current documentation before installing an add-on, connecting an account or spending credits.

ReelStack has not independently tested this workflow. Preview one representative shot and check motion, continuity and output quality before applying it to a full production. No result, cost saving or model capability is guaranteed.

Explore the tools

How to interpret the numbers

1,262 views captured 30 October 2026. ReelStack recommendation score: 39/100. Formula: radar-v5. This is ReelStack’s calculation, not a YouTube rating or a measure of factual accuracy.

Read our methodology and limitations →