Correction, 2026-10-01: the narrated-video table listed "Seedream 5" at 6 credits (that is Seedream 5 Lite) and used Merge Videos to assemble narration, music and stills. Merge Videos (the Montage node) joins video clips only; it takes no audio and no stills. The table now shows what each tool does, and the voice and pacing notes no longer state model behaviour we have not measured.
Correction, 2026-09-30: we priced scripts pro rata. Billing is per started 1,000 characters, so a 90-second script is 26 credits, not 18, and anything under 1,000 characters is 13. The table and totals below are fixed.
ElevenLabs TTS on this canvas is 13 credits per started 1,000 characters, up to 5,000 characters in one generation, which makes narration one of the cheapest things you can generate. A 90-second script costs 26 credits. A full-length book, split into 5,000-character generations, costs about $65.
The arithmetic is simple. The craft is punctuation and a proofread.
TL;DR
- 13 credits per started 1,000 characters, about $0.13, up to 5,000 characters per generation
- 90-second script (~1,400 chars): 26 credits. 3-minute script: about 39. Anything under 1,000 characters: 13
- Punctuation is your pacing control, because synthetic delivery follows it literally
- Read it back once. Names, jargon and numbers are where it fails, and those are the words that matter
- Library voices, not cloning. This is TTS, not a voice cloning platform
The arithmetic
Canon 8frame prices, a credit is $0.01 at pack rate:
| Script | Characters | Credits | Approx |
|---|---|---|---|
| 15-second social clip | ~230 | 13 | $0.13 |
| 30-second ad | ~450 | 13 | $0.13 |
| 90-second explainer | ~1,400 | 26 | $0.26 |
| 3-minute video | ~2,800 | 39 | $0.39 |
| 10-minute video, two generations | ~9,000 | 117 | $1.17 |
| 90,000-word book, 5,000-character generations | ~500,000 | 6,500 | $65 |
Speaking pace is roughly 150 words a minute, and a word averages about 6 characters with spaces. So characters divided by 900 gives you approximate minutes.
The started-block rule changes how you batch. A 1,001-character script bills two blocks, 26 credits. Several short lines generated one by one bill 13 each; the same lines in one generation, up to 5,000 characters, bill by their combined length. Batch short reads together, and split long scripts at 5,000-character boundaries, not at 5,001.
Punctuation is the interface
Synthetic delivery follows punctuation more literally than a person does, which is a control rather than a limitation once you know it.
Commas are short breaths. Use them where you want the voice to pause, even where a grammar checker would not.
Full stops are longer breaths. A long sentence without any will be read without air and sounds rushed.
Paragraph breaks are the longest pauses. Use them between beats.
Dashes and semicolons: if one doesn't give the pause you want, rewrite around it.
Write numbers as words where pronunciation matters. "$19" is ambiguous; "nineteen dollars" is not. "1985" can come out as "one thousand nine hundred eighty five".
The proofread that matters
Listen to the whole track once. Not a spot check.
What fails predictably, and it is always the important words:
- Proper nouns. Your product name, your company name, people's names.
- Numbers and units. Prices, dates, measurements.
- Domain jargon. Any term the model has not heard often.
- Acronyms. Read as letters or as words, inconsistently.
A 90-second track takes 90 seconds to check. Shipping a video where your product name is mispronounced for the whole runtime is a self-inflicted wound.
The fix for a stubborn word: spell it phonetically in the source text, or use the tool's pronunciation dictionary if you have one.
When to use your own voice instead
Synthetic narration suits informational delivery, where the register is flat anyway. Think twice where the voice has to carry conviction.
Use synthetic for: documentation, training, tutorials, explainers, product walkthroughs, localised versions of your own material.
Use your own voice for: anything about your judgement, your company's position, a founder message, an apology, or a video where the audience is deciding whether to trust you. The audience is responding to presence, and presence is the thing TTS removes.
Scope, stated plainly
This is text to speech with library voices. It is not a voice cloning platform. If you want narration in your own voice, that is a dedicated service, and the generated audio can then be used in your work here.
And the rule that applies wherever cloning happens: your own voice is legitimate, anyone else's needs documented consent for the specific use, and never a public figure.
The full narrated-video cost
| Element | Route | Credits |
|---|---|---|
| Narration, 1,400 characters | ElevenLabs TTS | 26 |
| Twelve illustrative stills | Seedream 5 Lite at 6 | 72 |
| Two b-roll plates, 5s | Wan 3.0 at 480p | 48 |
| Music bed, 30s looped | Lyria 2 | 13 |
| Join the two video plates | Merge Videos (Montage node) | 10 |
| Total for 90 seconds | 169, about $1.69 |
Merge Videos joins video clips only. Lay the narration, music and stills over the clips on the timeline of 8frame's separate Video Editor mode, which has sound layers, or in your own editor.
FAQ
How much does AI narration cost? 13 credits per started 1,000 characters, so anything under 1,000 characters is 13 and a 90-second script is 26.
How do I control the pacing? Punctuation. Commas and full stops are breaths, and synthetic delivery follows them literally.
Does 8frame clone voices? No. Library voices only.
Do I need to check the output? Yes, all of it. Names, numbers and jargon are where it fails.
Why did a short line cost 13 credits? Billing rounds up to the next 1,000 characters, with 13 as the minimum. Batch short lines into one generation.
Sources
- ElevenLabs TTS price, block rounding and the 5,000-character limit: the 8frame app, checked 2026-09-30
- Other prices: the 8frame app, checked 2026-09-30
- Merge Videos inputs (2 to 10 video clips, no audio input) and the Video Editor timeline with sound layers: the 8frame app, checked 2026-10-01
Punctuate for breath, listen to all of it once.
Try it without signing up: the 8frame canvas opens on a real board. The canvas is free and unlimited; generation is paid from $19/month.