UX: make the creation flow accessible for marketing, redactors, and product managers #1

New issue

Open

opened 2026-05-19 14:11:54 +00:00 by casper-stevens · 0 comments

casper-stevens commented

2026-05-19 14:11:54 +00:00

Member

Problem

The current workflow was designed with a developer/AI-operator mental model. For a marketing team or redactor to use it independently, several friction points block them.

1. No way to bring in context from your own files

The only input today is a single free-text "Intent" field (+ voice). There is no way to:

Upload or paste a campaign brief, script, or storyboard
Attach a brand style guide or visual reference images
Reference an existing document to seed the scene generation

For marketing, context is the work. A brief might be a PDF, a Word doc, or a Google Doc excerpt. Without a way to bring that in, users must manually summarise everything into a single text box — lossy and time-consuming.

2. Intent field is too narrow and unstructured

Marketing briefs have structure: audience, key message, tone of voice, call to action, platform/format. A single textarea discards all of that. The AI generates generic scenes because it lacks the context that lives in those fields.

Minimum structured brief fields needed:

Target audience
Key message / CTA
Tone / style
Platform & aspect ratio (Instagram Reels 9:16, YouTube 16:9, LinkedIn square…)
Duration target

3. Scene prompts are exposed as raw AI prompts

In the Planning step, each scene shows an Image Prompt and a Video Prompt textarea — raw AI generation prompts. A redactor or PM does not know what makes a good Stable Diffusion or Kling prompt. They think in terms of what the scene should show and feel, not model directives.

The interface should let them describe the scene in plain language and handle prompt engineering internally, or at minimum label/hide the technical fields behind an "Advanced" toggle.

4. No brand/campaign context at the collection level

Collections have only a Name. There is no place to attach:

A brand kit (logo, color palette, font)
A default tone/voice
Recurring visual style rules

Every project inside a collection starts from scratch, so brand consistency requires repeating the same context in every Intent field.

5. Pipeline step naming is technical

The steps Planning → Imaging → Video → Assembly map to the system's internal architecture. For non-technical users the natural language is closer to:

Current	Suggested
Planning	Brief & Storyboard
Imaging	Visuals
Video	Clips
Assembly	Final Edit

This is a label change but it meaningfully lowers the cognitive load for new users.

6. No per-scene editorial notes

A redactor reviewing AI-generated scenes needs to leave written feedback per scene before approving. Right now the only options are "Approve" or edit the raw prompt. There is no notes/comments field, so feedback has to happen out-of-band (Slack, email).

7. No script / storyboard import

Marketing teams often arrive with a pre-written script (numbered scenes, voiceover copy, shot descriptions). There is no way to import that structure — they would have to manually create each scene and fill in the fields one by one.

A plain-text import ("one scene per paragraph" or a simple numbered format) would unlock a large share of real-world workflows.

Acceptance criteria

Project creation modal supports file attachment or multi-field brief input alongside or instead of the free-text Intent field
Collection record can store brand context (name, style notes, reference) that pre-populates new projects
Scene planning view hides raw AI prompts from the default view; users describe scenes in plain language; prompts accessible via "Advanced"
Script / plain-text import: paste or upload a structured scene list to pre-populate scenes
Platform/format selection (aspect ratio + duration) available at project creation
Per-scene notes field available in planning step
Step labels reviewed for non-technical audiences

Out of scope for this issue

Collaboration / multi-user review flows (separate issue)
Video export format options beyond current assembly step

## Problem The current workflow was designed with a developer/AI-operator mental model. For a marketing team or redactor to use it independently, several friction points block them. ### 1. No way to bring in context from your own files The only input today is a single free-text "Intent" field (+ voice). There is no way to: - Upload or paste a **campaign brief**, script, or storyboard - Attach a **brand style guide** or visual reference images - Reference an existing document to seed the scene generation For marketing, context *is* the work. A brief might be a PDF, a Word doc, or a Google Doc excerpt. Without a way to bring that in, users must manually summarise everything into a single text box — lossy and time-consuming. ### 2. Intent field is too narrow and unstructured Marketing briefs have structure: **audience**, **key message**, **tone of voice**, **call to action**, **platform/format**. A single textarea discards all of that. The AI generates generic scenes because it lacks the context that lives in those fields. Minimum structured brief fields needed: - Target audience - Key message / CTA - Tone / style - Platform & aspect ratio (Instagram Reels 9:16, YouTube 16:9, LinkedIn square…) - Duration target ### 3. Scene prompts are exposed as raw AI prompts In the Planning step, each scene shows an **Image Prompt** and a **Video Prompt** textarea — raw AI generation prompts. A redactor or PM does not know what makes a good Stable Diffusion or Kling prompt. They think in terms of *what the scene should show and feel*, not model directives. The interface should let them describe the scene in plain language and handle prompt engineering internally, or at minimum label/hide the technical fields behind an "Advanced" toggle. ### 4. No brand/campaign context at the collection level Collections have only a **Name**. There is no place to attach: - A brand kit (logo, color palette, font) - A default tone/voice - Recurring visual style rules Every project inside a collection starts from scratch, so brand consistency requires repeating the same context in every Intent field. ### 5. Pipeline step naming is technical The steps **Planning → Imaging → Video → Assembly** map to the system's internal architecture. For non-technical users the natural language is closer to: | Current | Suggested | |---|---| | Planning | Brief & Storyboard | | Imaging | Visuals | | Video | Clips | | Assembly | Final Edit | This is a label change but it meaningfully lowers the cognitive load for new users. ### 6. No per-scene editorial notes A redactor reviewing AI-generated scenes needs to leave **written feedback** per scene before approving. Right now the only options are "Approve" or edit the raw prompt. There is no notes/comments field, so feedback has to happen out-of-band (Slack, email). ### 7. No script / storyboard import Marketing teams often arrive with a **pre-written script** (numbered scenes, voiceover copy, shot descriptions). There is no way to import that structure — they would have to manually create each scene and fill in the fields one by one. A plain-text import ("one scene per paragraph" or a simple numbered format) would unlock a large share of real-world workflows. ## Acceptance criteria - [ ] Project creation modal supports file attachment or multi-field brief input alongside or instead of the free-text Intent field - [ ] Collection record can store brand context (name, style notes, reference) that pre-populates new projects - [ ] Scene planning view hides raw AI prompts from the default view; users describe scenes in plain language; prompts accessible via "Advanced" - [ ] Script / plain-text import: paste or upload a structured scene list to pre-populate scenes - [ ] Platform/format selection (aspect ratio + duration) available at project creation - [ ] Per-scene notes field available in planning step - [ ] Step labels reviewed for non-technical audiences ## Out of scope for this issue - Collaboration / multi-user review flows (separate issue) - Video export format options beyond current assembly step