Narrated Lesson Video From 10 Slides: A Seven-Minute Cost Breakdown

The short answer

A 7-minute narrated lesson with 10 slide images costs about $1.54 on Sume list prices: 6,000 characters of narration ($0.285), ten gpt-image-2.5 slides at medium quality and 2K ($0.556) and a 420-second timeline render, which is 7 billable minutes ($0.70).

Parts

Narration comes from tts_create at $0.0475 per 1,000 characters. One request is capped at 20,000 characters ($0.95 at the maximum estimate), so a 6,000-character script fits in a single call, but splitting by section makes a bad take cheaper to redo. Timeline 1.0 accepts an audio spine up to 1,800 seconds and up to 200 video slots, and treats a still image as a static hold, so each slide is one slot with a start time and duration.

Because the spine is the narration, lay out the slides by the time each section starts. Use audio.parts[] (up to 20 gapless slices) to join section recordings without re-synthesis.

Checks before you publish

Read the slides for text errors. Dense text in generated images is the main risk, and docs recommend a higher quality for finals and dense text. Rendering a slide at high quality and 4K costs $0.2225 each, so a deck of ten at high is $2.2250; spend that only on slides with small print.

Run POST /v1/timeline-1.0/plan first. It is unbilled and returns billable_minutes, so you see the $0.70 line before the render.

• Add chapter markers in your video host, not in the render.
• Captions: standalone caption jobs are priced for videos up to 60 seconds, so for a 7-minute lesson, cut and caption in sections or use your host's captions.
• Keep a copy of the script; the TTS job records the voice and settings so the next lesson can match.

Scaling to a course

A course of 12 lessons of this size costs 12 x $1.541 = $18.50 at list price, before any retries. Plan for retries: assume one slide in five needs a second pass, which adds 2 x $0.055625 per lesson, or $1.335 for the whole course.

If lessons share a template, generate the template slide once and edit it for each lesson with image_urls; the edit costs the same per image but gives a more consistent look.

Pace matters more than price. Seven minutes of narration with ten still slides means a new image about every 40 seconds; learners tend to lose the thread on a static picture that long. Add a second slide for any section over 30 seconds. Each extra slide is $0.055625 at medium quality and 2K, so adding ten more raises the lesson by $0.55625.