Script to Video AI: The Full Workflow, From Script Length to Platform Rules

Script to video AI tools all promise the same thing: paste text, get a narrated video. The part they leave out is everything around that button. How long the script should be for the duration you want, where the footage comes from, which voice and subtitle settings to pick, which frame size each platform accepts, and what YouTube, TikTok and Instagram say about AI-made and unoriginal videos. This guide covers that whole workflow, using one measured run (a 441-character script that became a 32.6-second video) instead of rules of thumb, and it links to our tool comparisons wherever the choice gets detailed.

Short version: plan about 13–14 characters of English narration per second of video, choose footage by how much the topic needs to be shown rather than described, keep the export at 1080×1920 vertical, review every scene by hand, and label realistic AI content where the platform asks for it.

How long should the script be? Script length vs video duration

In the run we documented, the only reliable length was the exported file's, not the estimate. That gets expensive on credit-based plans, so it pays to size the script before you generate anything.

Here is the one measurement we can stand behind. On 10 October 2026 NoobClaw ran its own script-to-video workflow once and published the run record: a 441-character English script (73 words), read word for word by the Microsoft Edge voice "Jenny" (US English), exported as a 32.6-second MP4. That works out at about 13.5 characters, or 2.2 words, per second, roughly 134 words a minute. Scaling that pace gives the table below.

Treat these as a starting point, not a constant. The table comes from a single voice at default speed; a slower voice, added pauses or a different language will move every row. Two things from the same run are worth knowing even if you never use NoobClaw:

• Built-in estimates can be far off. The wizard estimated 98 seconds for that script because, per the run record, its estimate uses a Chinese reading rate. The real file was a third of that. Time the export, not the estimate.
• Scenes and subtitle lines multiply fast. With a 4-second shot pace, 32.6 seconds became 7 scenes and 28 subtitle lines, so each line was on screen for a little over a second. Short sentences read better than long ones at that rate.

Our read: write to the duration first, then cut. A 60-second script that ends at 810 characters is easier to review than a 1,200-character draft you hope the tool will compress. If you need a first draft, the free TikTok script generator lets you choose 30, 45, 60 or 90 seconds, and the hook generator drafts the opening line separately.

Stock footage, your own clips or AI-generated clips?

The visual source decides more about cost, review time and platform risk than the choice of tool does. A practical way to choose: ask whether the viewer needs to see something specific, or whether the footage only illustrates what the voice is saying.

Stock assembly is the route most exposed to sameness, because every user of a tool draws from the same libraries. The NoobClaw run record says some automatically selected footage "was generic or off-topic" and needed scene-by-scene review, and its limits section adds that stock-source licensing was not audited clip by clip. Plan for both.

Generated clips are a different purchase. On the plans we compared in our InVideo AI alternative breakdown, generation is metered per short clip or per second, while stock-footage videos are priced per minute or per video. That page and Fliki vs Pictory carry the dated price-per-minute maths, so we don't repeat the numbers here.

Own footage is the route most guides skip. If you already have screen recordings or product shots, an assembly tool that puts an AI voice and subtitles over your clips avoids both the off-topic stock problem and the generated-video label question. The faceless channel tools breakdown splits the stack into script, voice, visuals and upload if you are budgeting a whole pipeline.

AI voice and subtitle settings that hold up on a phone

Voice. Pick one voice per channel and keep it. Before you approve a voice, listen for names, brand terms and numbers, because, in our experience, those are where text-to-speech errors are most noticeable. The run record is honest about this: the audio track and a successful TTS log were present, but pronunciation was "not independently approved". A passing log is not a listening check.

Voice also touches disclosure. YouTube's GenAI disclosure page lists "Cloning one's own voice to create voice overs or dubs" among the uses that don't need a label, along with using AI for an outline or script and caption creation. It asks for disclosure when content makes a real person appear to say something they didn't. TikTok's AI-generated content page says labelling is required for "realistic images, audio, and video". TikTok's page doesn't say whether a realistic synthetic narrator counts as realistic audio; if the voice could pass for a real person, turn the label on.

Subtitles. Burned-in subtitles travel with the file to every platform; platform captions have to be added on each one separately. Our run used burned-in subtitles. The tested setting in our run was medium white text with a black outline at the bottom of the frame. Check the first upload preview on each platform: if the app's own caption text or buttons cover your subtitles, move them up and export again.

Keep the spoken script and the post caption as two separate texts. The script is what the voice says; the caption introduces the video on the feed. The free caption generator is built for the second job.

Aspect ratio and export settings for TikTok, Shorts and Reels

One vertical export can serve all three short-video platforms, provided the length and frame rate sit inside each one's limits. These are the limits the platforms publish themselves.

Our measured export, 1080×1920 at 30 fps, H.264 video with AAC audio, 32.6 seconds and about 7.6 MB, sits inside every row of that table. If you post the same file to all three, 9:16 at 1080×1920 and 30 fps is the safe common denominator. Note the TikTok row comes from TikTok's developer documentation for uploads through its API; it is the most precise spec we found from TikTok, but it describes the API path rather than the app.

Instagram's page also says you can turn on "Upload at highest quality" under Media quality in settings. If your Reels look softer than the export, check that switch before you change export settings. For Edits' export quality options, see our Edits vs CapCut comparison.

Platform rules on AI-generated and unoriginal content

None of the three platform pages we read bans AI-made video as such. All three care about two separate things: whether realistic AI content is disclosed, and whether the video adds something original. A label answers the first question and does nothing for the second.

YouTube: inauthentic content and disclosure

YouTube's channel monetization policies carry a notice dated 15 July 2025 renaming the "repetitious content" policy to "inauthentic content". The current section, headed "Generic or Repetitive Content", lists "AI-generated content made with generic or unoriginal templates" and "Image slideshows, templated storylines, or scrolling text with minimal or no narrative" as not allowed, while a shared intro and outro is fine if "the bulk of your content is different". The reused-content section applies "even if you have permission from the original creator". On disclosure, YouTube's GenAI disclosure page says that "Disclosing AI content alone won't limit a video's audience or impact its eligibility to earn money". Our YouTube AI video monetization article walks through review, notices and appeal windows.

TikTok: label realistic AI content, add something new

TikTok's Integrity and Authenticity guidelines (released 25 August 2026, effective 24 September 2026) require creators "to label AI-generated or significantly edited content that shows realistic-looking scenes or people"; unlabelled content "may be removed, restricted, or labeled". The same page makes content ineligible for the For You feed "if it includes unoriginal or reused material without anything new". To add the label, TikTok's AI-generated content page gives the path: before posting, tap More options and turn on AI-generated content (on a computer: Upload, See more, then the same switch). It also says the label "won't affect the distribution of your video" and cannot be removed after posting.

Instagram: originality decides reach to non-followers

Instagram's Original Content Guidelines say unoriginal content "may be ineligible to appear in recommendations to people who don't already follow you". Borders, watermarks and speed changes count as low-effort edits; narration or a storyline that "meaningfully re-contextualizes" the source counts. The page's own test is useful for any AI video: if your contribution could be removed and the post "would virtually be the same", it needs more of you in it. Eligibility returns once most posts in a rolling 30-day period are original. For AI labels, Meta renamed its "Made with AI" label to "AI info" in July 2024 and applies it from industry signals or from the uploader's own disclosure.

Our read across all three: a script-to-video tool makes the file, but originality comes from the script. Stock clips and a synthetic voice are shared by everyone using the same tool; the angle, the facts and the example in your script are not.

Review checklist before you post a script-to-video export

Run this on every export, not just the first. It takes a few minutes on a 30-second video and catches the problems that cost reach or trigger a notice.

1. Play the whole file once with sound. Does every point in your script survive in the narration?
2. Check each scene against the sentence it sits under. Swap any clip that is generic, off-topic or contradicts the voice.
3. Listen for names, numbers and niche terms. Re-generate the voice or respell the word if one is wrong.
4. Read the subtitles on a phone screen. Lines should be short enough to read at about one second each and clear of the app's own buttons.
5. Check the length against the platform table above, and that the file is vertical 9:16.
6. Note where each stock or uploaded clip came from and the licence that covers it.
7. Remove any claim in the script you can't support with a source.
8. Decide on the AI label: realistic synthetic scenes or people need one on TikTok and YouTube.
9. Apply the "remove your contribution" test from Instagram's guidelines. If the video would be the same without your script's angle, rewrite before posting.
10. If you post to several accounts, make a different script per account rather than the same video with a new caption; our Repurpose.io alternatives page explains when reposting is the job and when creating new videos is.

Which script to video AI tool for which job

We have compared these tools in separate articles with prices checked on the day. Rather than copy prices that change, this table points you to the right comparison.

Where NoobClaw fits in a script-to-video workflow

NoobClaw's script-to-video feature covers steps 2 to 5 of the workflow above inside its Windows and macOS desktop app. The input is a pasted script: 200–800 characters in word-for-word mode, which at the measured pace is roughly 15 to 60 seconds of narration. Footage comes from online stock or your own clips, narration from an AI voice, and subtitles are burned into the frame; the result is a 9:16 MP4 saved locally or sent to connected accounts. Each account keeps its own profile with niche, persona and keywords, which matters if you run several channels and want a different script for each.

The documented output is 9:16 only, and the product docs describe no avatars or timeline editing; for frame-level edits, finish the MP4 in a separate editor. Even a local-only export requires an internet connection and a connected account profile; scenes are matched automatically, and each video spends AI credits (new accounts get 1M free credits; paid plans start at $9.90/month, see pricing). The step-by-step setup guide shows the exact settings from the measured run, including the local-only output option so nothing is posted before you review it.

FAQ

How many words is a 60-second script?

About 130–135 English words, or roughly 810 characters, at the pace we measured with a standard US English AI voice. Slower voices and pauses need fewer words. Time your own export once and adjust.

Is CapCut script to video free?

CapCut's AI script to video maker calls itself free, runs in the browser after you sign in, and says it "does not stamp your work" with watermarks. CapCut's paid plans are priced by region; our Edits vs CapCut page has the details.

Do I have to label a video made with an AI voice and stock footage?

It depends on what the video shows, not on the tool. YouTube and TikTok ask for a label when content looks or sounds realistic and could be mistaken for real people or events; YouTube lists captions, AI-written scripts and cloning your own voice as not needing disclosure. Check each platform's page for your case.

Can AI-made videos be monetized on YouTube?

They can be eligible, but AI use does not decide it. YouTube looks at whether videos are templated, repetitive or reused without a meaningful difference. See our YouTube AI monetization rules article before you build a channel around one format.

Sources

1. Script-to-video run settings and measurements — NoobClaw — 10 Oct 2026 — noobclaw.com/examples/script-to-video-workflow-2026-10-10.json
2. YouTube channel monetization policies — YouTube Help — notice dated 15 Jul 2025, read 23 Oct 2026 — support.google.com/youtube/answer/1311392
3. Disclosing use of GenAI content — YouTube Help — read 23 Oct 2026 — support.google.com/youtube/answer/14328491
4. Upload YouTube Shorts — YouTube Help — read 23 Oct 2026 — support.google.com/youtube/answer/12779649
5. Integrity and Authenticity, Community Guidelines — TikTok — released 25 Aug 2026, effective 24 Sep 2026, read 23 Oct 2026 — tiktok.com/safety
6. AI-generated content — TikTok — read 23 Oct 2026 — tiktok.com/tns-inapp/pages/ai-generated-content
7. Content Posting API: media transfer guide — TikTok for Developers — read 23 Oct 2026 — developers.tiktok.com
8. Reel size and aspect ratios on Instagram — Instagram Help Center — read 23 Oct 2026 — help.instagram.com/1038071743007909
9. Original Content Guidelines — Instagram for Creators — undated, read 23 Oct 2026 — creators.instagram.com/original-content-guidelines
10. AI Script to Video Maker — CapCut — read 23 Oct 2026 — capcut.com/tools/script-to-video-maker
11. Our Approach to Labeling AI-Generated Content and Manipulated Media — Meta — 5 Apr 2024, updated 1 Jul and 12 Sep 2024 — about.fb.com