A Teachable curriculum is a slide stack, not a course trailer
Lesson titles, module order, and progress UI are stills you stack — /for/online-courses is marketing video; Teachable's object is the curriculum board.
Lesson titles, module order, and progress UI are stills you can stack. /for/online-courses is marketing video. Teachable's object is the curriculum board.
A Teachable school already has the expensive part: the outline a student will click. Module names, lesson order, a progress bar that is this course. That board is a set of stills. Stack them. Do not spend a cinematic trailer row restaging a talking instructor who never taught the lesson.
The curriculum board is the product
A student who is already on the fence is matching titles to a syllabus they were promised. "Week 3 — the 12-bar blues in G" is a string. A golden-hour hero of "an inspiring online classroom" is a different school.
Screenshot the real Teachable UI you ship: sidebar modules, lesson list, the progress chip. Those JPEGs are the identity of the course. Do not generate a cousin dashboard and typeset your titles onto it. Generated chrome is a different LMS. Generated progress is a lie about completion.
If you need a title card the screenshot cannot carry — a module stamp, a lesson number at thumbnail size — that is a lettered still, not a restage of the school. Quote the string. Proof at 100%. One wrong lesson name is a regenerate, not a "vibe of week three."
Do not invent students in the UI. Do not generate a classroom of faces "so it looks enrolled." Empty board, real titles. Occupancy is a grant problem you do not need.
Qwen letters the lesson titles
The slide is judged on whether every word is the word you specified, at a size a phone can read. Send a curriculum still to a portrait model and you get text-shaped noise.
Qwen Image 3 Text to Image is the stills row when the letters have to read. Catalog facts: Qwen's third-generation image model, strong prompt adherence, complex text rendering in Chinese and English, intelligent prompt rewriting on by default, 4 credits, 2K max, qualities 1K and 2K. No audio. No duration. Aspects include 1:1, 16:9, 9:16, and 21:9.
Write the exact strings. If a module title must stay verbatim, say so in the prompt and inspect the still. Rewriting is a help for scene language, not a licence to paraphrase "Module 07 — Mixolydian" into "a jazz lesson." Set 1K while the words are in draft. Set 2K when the slide is the slide. There is no 4K chip on this row.
A video model is a bad typography engine. It will give you seven frames of almost-title and a classroom you cannot ship. Do the lock here. Approve the characters. Then stack. Do not ask a talking cinematic row to typeset the syllabus. Pause any such clip: the titles are the model's, not yours.
Plan the stack; stamp after lock
Generate a slideshow from a prompt is the named agent job when you need a planned set: theme, format, slide count, then a board per beat. Cost is priced per image generated, shown before you confirm. That planner is useful for structure — welcome, module one, module two, CTA. It is dangerous if you let it invent the syllabus.
Do not brief "a popular course about guitar." Brief the actual strings, or generate the lettered stills on Qwen and assemble them. A prompt-slideshow that fills "Lesson 04" because it sounded curricular is a cousin course.
The honest stack is mixed: Teachable screenshots you already own, then Qwen cards for titles the screenshot cannot carry. The slideshow maker is the surface that holds that set. It does not become a trailer because the landing page says video.
Add text to slideshow images is the stamp pass after the pictures are keepers. One overlay auto-expands across every slide, or N overlays matching N images. It does not generate the board. Caption-in-image paints letters into pixels you cannot edit; if a module name will change, use the stamp, not the bake. Lock order and count first. Convert to a reel only when the edited images are the ones you mean to ship.
/for/online-courses is the marketing excerpt
Online course creators convert with a real lesson excerpt, not a testimonial reel. Three posts a week, 16:9 for the YouTube lesson and a 9:16 crop of the same still for Reels. That page is teaching on camera: clip the lesson, prove the instructor. It is not a Teachable outline restaged as a cinematic trailer.
Do not rewrite that audience page as a slideshow of module titles. Do not paste a generated talking lecturer over the curriculum board. A synthetic instructor is a likeness you probably do not have.
Use one real frame from the curriculum: the screenshot, or the Qwen card you proofed. Do not generate a second curriculum for the vertical. If the type dies in the crop, the slide had no safe zone. Syllabus they were promised: Teachable board. Whether they will learn from this person: clip a real lesson on the courses /for page. One generate cannot be both.
FAQ
Is generate-a-slideshow-from-a-prompt the same as screenshotting Teachable?
No. The agent plans a new board from a theme, priced per image generated. A Teachable screenshot is identity you already own. Use the planner for structure and quoted titles. Do not let it invent a cousin LMS.
Can I put lesson titles in a video prompt so the trailer typesets the syllabus?
No. A talking cinematic row will invent plate math and a classroom. Letter the titles on Qwen Image 3 at 4 credits, 2K, and proof the characters. Stack screenshots of the real progress UI beside those cards.
When do I stamp overlay instead of baking type into the Qwen still?
Stamp when the line may change — CTA, campaign, a curriculum rewrite. Bake only when the type is the scene and you will not edit it. Do not chain a stamp on top of baked-in type or you double-print the module name.
Should we generate a talking instructor as the course trailer?
Not as a substitute for the curriculum board, and not as a synthetic lecturer. /for/online-courses is a real lesson excerpt. The Teachable object is stills you stack: titles, order, progress UI. A constructed mouth is a different job you probably should not buy.