InVideo AI Text to Video: Script to Finished Video in 2026

InVideo AI can turn a typed prompt or a finished script into an edited video, but in November 2026 there are two routes, and the default is not the one most older guides describe. The prompt page opens Agent Two, which runs a Script to Video or Explainer workflow as a conversation, or builds the whole video in one pass with Autopilot. The original stock-media generator, where you paste a script and get voiceover, subtitles and music back, still sits behind the Older versions button on the same page ^[1]. Both work. They suit different jobs.

Our read: the route matters more than the prompt. If you want stock footage with AI narration and captions, the classic flow is still the shortest path. If you want generated shots from models such as Seedance 2.5 or Veo 3.1, Agent Two is the way in, and the bill follows seconds of generated video and the number of retries, not the length of your script. Timeline edits and exports in the invideo Editor cost nothing, which changes how you should work.

Two routes from a script to a finished video

The classic flow trades control for speed. You set media type, music, subtitle style and language once, and the generator fills in the rest from stock or generated media. Agent Two works shot by shot: it plans, drafts a storyboard, picks a model for each job and assembles the cut on a timeline where every clip can be reshaped or undone. invideo does not say how long the older versions will stay reachable, so treat the classic route as available today rather than a long-term promise.

Script to Video with Agent Two, step by step

1. Open the prompt page. Agent mode is selected by default. Paste your script or describe the video, and attach anything useful: the agent reads uploaded videos, scripts, PDFs, links and rough cuts.
2. Pick the workflow. Script to Video and Explainer run as a back-and-forth conversation; Autopilot generates in one shot with no questions.
3. Pick the agent tier. Lite is fast and light on credits, Pro aims at premium quality, Ultra handles heavy work such as long scripts. Starter gets Lite only.
4. Set the approval mode. The default is to allow generation without asking. You can switch to ask before video or ask before any generation.
5. Review on the timeline, then export. Trim, reorder and recolor by hand for free, and export a 1080p MP4 or an editable project.

Step four deserves a minute. With approval on, a popup shows the prompt, the model, the duration and the aspect ratio before anything is spent, so a misread script costs a click instead of credits. invideo describes the same idea in its Always Ask mode, which pauses before every generation.

The agent keeps a project Context that grows as you work and is checked before each generation, so characters, locations and voices stay consistent across scenes. Text-to-video models listed for this job include Seedance 2.5, Gemini Omni Flash, Veo 3.1, Kling 3.0 and MiniMax H3, and you can lock which models the agent may use. Jobs run in the cloud and continue after the browser tab closes. invideo says the agent reviews each generation against your references and regenerates failures, a vendor claim worth checking on your own first draft.

The classic Use my script flow, for stock-footage videos

The older route still fits explainers, listicles and narrated social clips where stock footage is good enough. Click Older versions, choose Use my script, paste the text and set media type, background music, subtitle style and language. Each sentence becomes a segment. Adding a gap after a sentence inserts a voiceover pause and lets the media play longer, which is the main timing tool you get.

Voices are picked by naming them in the prompt: American regional accents, British, Indian English and Australian, plus Spanish, French, Italian, Japanese, Korean, Dutch, Polish, Portuguese and Russian voices, each male or female. A voice cannot be previewed before the video is generated, so name it precisely. A video can carry several speakers next to the narrator, all in the narrator's language.

After generation, Magic Box takes plain-language commands: change the voice or accent, switch music, change the pace, delete scenes, restyle subtitles or translate the video into Spanish. A slower pace adds less media and a faster one adds more. Subtitle styles include karaoke, word by word, bold popping and a Hormozi style, and you can move or resize them. After export, the subtitles download as an SRT file. Media choice is where this flow controls cost: a new chapter filled with generated media costs credits, while stock media for the same chapter saves them, and Text to clip, Text to image and Remix in the Media panel all cost credits.

Prompt, script or link: who writes the words

You do not need a finished script. invideo says its classic generator writes the script from your prompt, searches more than 16 million stock images and videos or generates visuals, then adds voiceover, music and transitions. That makes the prompt itself the script generator, and the quality of the outline you type decides the quality of what you get. Name the audience, the length, the tone and the call to action.

For longer source material, the agent route is the better fit. The docs describe turning a policy, SOP or handbook into a narrated explainer with b-roll, text slides and captions, with no avatar needed. Link to video makes a new video from the information on a page; it does not download videos hosted at that link, and not every image on the page carries over, so upload product photos that must appear exactly.

One limit surprises people. invideo does not fetch web pages in the agent workflows. When an outside assistant such as Claude or ChatGPT drives invideo, that assistant reads the URL with its own tools, then shows a script for approval before spending anything. For voice generation, a single call refuses text longer than the model accepts, so very long scripts get split into paragraphs.

What a text-to-video project costs in credits

Video models charge credits per second, and resolution, a video input and generated audio all change the rate ^[2]. The pricing compare table lists Seedance 2.5 at about 56 credits per 5 seconds at 720p and about 96 at 1080p, as read in November 2026 ^[3]. A quick Kling 2.6 Standard video can cost about 5 credits, while Veo 3.1 in 4K with audio can pass 100. invideo says the exact cost is shown in the app before you confirm; in Agent Two that confirmation appears only if you switch the approval setting away from the default.

Plan size sets the ceiling. Starter costs $20 a seat a month billed monthly and grants 400 credits a seat a month, with Agent Two Lite only and no Seedance 2.5. Plus costs $60 a seat a month billed monthly and grants 2,000 credits a seat a month. By our arithmetic, at about 96 credits per 5-second 1080p Seedance 2.5 clip, Plus covers about 20 clips a month and Max at 5,000 credits about 52, before any retries.

Retries are the real budget line. invideo's own documented productions averaged about three generations per usable shot, and invideo says iteration, not runtime, drives the credit budget. Its docs add that generated shots cost the most, stock is cheaper and text slides are free. Every timeline edit and every export is free ^[4], so fix pacing, cuts and captions by hand rather than regenerating.

Plan credits reset monthly on the subscription date, annual plans included, and unused credits do not roll over. If you run short mid-video, a paywall offers more credits, or you can finish with stock media. Fees are non-refundable; invideo's Terms allow a refund request within 24 hours of a first purchase or renewal only if no credits or premium features were used, and credits, including extra credits, are never refunded. The InVideo AI pricing guide covers that rule and the annual offer.

Resolution: what you generate is not what you export

Generation resolution and export resolution are separate numbers. Plus unlocks Seedance 2.5 at 1080p and Seedance 2.0 at 4K, and Veo 3.1 can generate up to 4K, but a finished timeline exports as an MP4 of up to 1080p. A 4K or HDR master needs a paid upscale or HDR model, charged per second of source video, or a finish in Premiere Pro or DaVinci Resolve. In Agent Two, only the Pro and Ultra tiers can upscale the export, and Starter gets Lite only.

That round trip is built in. The free invideo Editor exports an editable project for Premiere Pro, Final Cut Pro or DaVinci Resolve, though transitions, effects, captions and grades may look different in the other app. In the classic flow, changing between 16:9 and 9:16 by command regenerates the whole video, so settle the aspect ratio before you polish. For the full editor walkthrough, see our invideo Editor guide.

The bottom line

InVideo AI still turns text into video, but you now choose between a guided agent that generates and assembles shots on a timeline and a classic stock-footage generator kept behind Older versions. Use the classic flow when stock visuals and narration are enough; use Agent Two when you need generated shots and model choice. Either way, budget by seconds of generated video and expected retries, switch approvals on until you know the rates, and do the fine edits by hand in the Editor, because those cost nothing. For a cinematic finish, see InVideo AI for filmmakers.