AI video for course creators — B-roll, openers and cloned narration

The unglamorous truth about recording a course is that the lessons are the easy part. You know the material. What eats the weeks is everything around it — an opener that does not look like a 2014 template, B-roll for the four minutes where you are explaining an abstraction to a static slide, title cards that match across forty lessons, and a re-record of lesson six because you caught a cold halfway through the module. Generated video and speech are good at exactly that surrounding layer, and clumsy at the layer you were already good at.

What generated video is for in a lesson

Draw the line before you spend anything.

Good uses: the five-second opener, the establishing shot behind a section title, the visual metaphor while you narrate a concept, the transition between chapters, the still behind a quote, the hero image for the sales page.

Bad uses: anything that pretends to be you. There is no lip-sync in eroq, no talking avatar, and no training a model on your own footage, so a generated presenter is a fictional stranger whose mouth does not match your words. Your camera stays the face of the course. The one thing that genuinely can carry your identity is your voice, and that is a clone, not a puppet.

Also worth knowing before you plan around it: eroq renders clips and stops. Trimming, transitions and color grading happen in your editor.

One look for the whole curriculum

A course looks amateur when lesson one is warm and filmic and lesson nine is cold and digital. Three mechanisms keep forty lessons in one visual family.

A style Element. Elements is a library of reusable places, objects and styles — up to 50, each @mentionable in a prompt. Write your course look once as a style Element and drop the @mention into every render instead of retyping five rack settings and hoping you remembered them.

The director's rack. It folds into the prompt server-side, so your identity is a preset rather than prose you re-improvise: Film type "Documentary", Era "2010s", Tempo "Calm", Camera gear "35mm film", Palette "Nostalgic blue", Lighting "Soft cross" is one coherent house style. Change it once and every future lesson moves with it.

A seed. The Seedance engines honor a seed for reproducibility, so the opener you liked in lesson one can be re-rendered with a new subject and the same character rather than rediscovered by luck.

The opener prompt

One flowing paragraph, subject then motion then camera then light then mood:

An empty lecture hall at dawn seen from the back row, slow push-in down the center aisle as pale blue light rises through tall windows and dust turns in the beams, rows of wooden seats in soft focus, clean sharp lens at f/4, nostalgic blue palette, soft cross lighting, calm tempo, documentary look on 35mm film, wide shot.

Swap the subject line per module and keep everything after the first comma. That is the whole trick to a consistent curriculum — the grammar stays, the noun changes.

For lessons that need a sequence rather than one clip, Cinema mode in the video studio builds a storyboard progressively: add scenes one at a time, up to 12, each rendering immediately, with takes ×1–4 per scene and the option to edit, re-render, reorder or duplicate any of them. Drafts save as "My films", so next year's revision starts from this year's board.

Narration in your own voice

The voice studio does text-to-speech with a curated roster — Aria is warm and intimate, Orion low and calm, thirteen languages — and it also clones a voice from your own audio, which is what most course creators actually want. Record the clone once and lesson six gets its re-record without you getting over the cold first.

Two models, same roster and same clones: Voice Turbo at 2 credits per 100 characters, Voice One at 3 credits per 100 for the expressive flagship. The working habit is to draft every lesson on Turbo, listen, fix the script, then render the keeper on Voice One. Delivery is driven by punctuation and by the speed and expressiveness controls, so the edit that fixes a flat line is usually a comma, not a re-render on a different engine. More detail lives on the voice cloning page.

What a ten-lesson module costs

One module: an opener per lesson, two B-roll clips per lesson, a title card each, and narration. Clips at 5 seconds on Seedance 1.0 Lite at 60 credits, stills on Image One at 10, narration at 6,000 characters per lesson — roughly a thousand words — on Voice Turbo:

• Openers — 10 × 60 = 600
• B-roll — 20 × 60 = 1,200
• Title cards — 10 × 10 = 100
• Narration — 60,000 ÷ 100 × 2 = 1,200

Total 3,100 credits, about $31 at the entry pack rate of $10 for 1,000 credits. Render the final narration pass on Voice One instead and it is 60,000 ÷ 100 × 3 = 1,800, so the module lands at 3,700.

Either way one Creator month at $49 for 6,000 credits covers a full module with enough left for a second pass on half the lessons. Hobby at $15 for 1,800 credits covers roughly half a module a month, which is fine if you ship a module every two months and let the balance roll over — credits never expire. Failed renders refund themselves, so a bad batch costs an evening.

Assembling, and the part eroq does not do

Everything you render is saved in your library with its full recipe, and Remix reloads that recipe into the tool — that is how lesson eleven starts from lesson ten. Animate turns a still you like into the source for an image-to-video clip, which is the cheapest way to get motion out of a title card you already approved.

Then you export. Course assembly — timeline, captions, quizzes, the LMS upload — happens where it always did. The same division of labor is described from the other side on the faceless YouTube use case, where the generated layer is doing even more of the work.

FAQ

Can I make a video of myself teaching without filming?

No. eroq generates original fictional people, there is no lip-sync and no avatar that speaks your words, and training a model on your own footage is not a feature. Clone your voice, film your face.

How many credits does narration really take?

Billing is per character, not per word or per minute. Voice Turbo is 2 credits per 100 characters, Voice One is 3. A thousand-word lesson is around 6,000 characters, so 120 credits on Turbo and 180 on Voice One.

How do I keep forty lessons looking the same?

Save the look as a style Element, keep one director's rack preset, and reuse a seed on the Seedance engines. Change the subject of the prompt, never the grammar after it.

Writing module one this week? Open the voice studio, clone your voice first, and let the openers follow the script.