Skip to content
SURGE AIFull-stack AI for enterprises
Menu

Free alternatives · Video, slides and media

AutoVideoStudio

Type a topic and get a researched, fact-checked YouTube video of 1 to 60 minutes, with captions, music, thumbnails and Shorts.

A free alternative to VidRush, built by Surge Software. Free. You pay only what third-party services cost us, at cost.

Live at autovideostudio.com

Walkthrough

See AutoVideoStudio in action

A narrated tour of AutoVideoStudio: the problem it solves and how a team uses it, screen by screen.

The problem

What it solves for a business

For: Faceless YouTube channel owners, education creators and coaching institutes, news and media teams, and agencies that run many channels.

  • Making a long researched video by hand takes days of research, scripting, footage hunting and editing.
  • AI video tools pick footage that does not match the narration, and credits get burned on renders that fail.
  • A wrong fact or an unlicensed clip in a video can hurt the channel.

How it works

AutoVideoStudio, step by step

  1. 01

    Write a brief

    Give a topic, a format (documentary, Top-10, explainer or story), a length of 1 to 60 minutes, an aspect ratio, audience and tone.

  2. 02

    Research and script

    The app pulls Wikipedia extracts as numbered sources and writes a script where every scene cites its source.

  3. 03

    Fact-check gate

    Each claim is matched to its cited source; unsupported claims or wrong numbers stop the render until you edit or approve them.

  4. 04

    Footage, voice and edit

    Licensed footage is gathered, the voice track is added, and the video is cut with retention-aware pacing, captions and a music bed.

  5. 05

    Thumbnails, metadata and Shorts

    You get two thumbnails, a title, a description with chapters and tags, and 9:16 Shorts cut from the best stretches.

  6. 06

    Fix by clicking or typing

    Swap a shot in the scene editor, edit the script, or type a change in the chat editor; only what changed is re-rendered.

Comparison

AutoVideoStudio vs VidRush

VidRush is a specialist tool that makes researched long-form faceless YouTube videos from a brief.
 AutoVideoStudioVidRush
PriceFree. You pay only what third-party services cost us, at cost.$99/month for 45 min to $639/month for 409 min (checked 24 Oct 2026)
Free tierThe whole product is freeNone; waitlist reported
Failed rendersNothing is charged for a failed renderTrustpilot reviews report export crashes that still used credits
Fact-check and sourcesEvery claim checked against its cited source; render blocked until fixedSourcing panel only in the enterprise "Atlas" offering
Footage relevanceRelevance score per shot, one-click swap, re-search or upload your ownIrrelevant footage is a reported complaint (Trustpilot, 31 Aug 2026)
ShortsAutomatic 9:16 cut-downs with karaoke captionsDescribes itself as "not built for Shorts"
APIPOST/GET /api/v1/videos with a workspace keyNone (planned, per founder)
Max length1 to 60 min6 to 60 min

VidRush details are taken from its public website and may have changed since ($99/month Starter (45 min) to $639/month Pro (409 min), about $2.20 to $1.56 per finished minute (checked 24 Oct 2026)).

Why switch

Advantages over VidRush

No plans, no credits

The product is free. Paid outside engines, such as a narration voice, are passed through at cost, and only for successful renders.

Facts checked before render

Every claim is matched to the source it cites. A wrong number stops the render until you fix it or explicitly approve it.

Licence record for every file

The asset ledger keeps the provider, licence, attribution and a SHA-256 hash per file. NC and ND licences are rejected.

Shot-level control

Each shot shows a relevance score. You can swap it, re-search the scene, upload your own clip or type the change in chat.

Shorts from every video

The best 15 to 58 second stretches are re-cut to 9:16 with captions, which VidRush says it is not built for.

Features

What it does today

  • Brief to finished video: documentary, Top-10, explainer or story, 1 to 60 minutes, in 16:9, 9:16 or 1:1.
  • Research from Wikipedia with numbered sources, and a script where each scene cites a source.
  • Fact-check gate: claims are matched to cited sources and wrong numbers block the render.
  • Footage from Wikimedia Commons (public domain and CC licences only) with an asset ledger that records the licence and a SHA-256 hash for every file, plus credits in the description.
  • Retention-aware edit: fast cuts in the hook, pattern interrupts at most 20 seconds apart, and a predicted retention curve.
  • Captions in three styles (bold-pop, karaoke, clean), a generated music bed, two thumbnails, and a title, description with chapters and tags.
  • Automatic 9:16 Shorts cut from the best 15 to 58 second stretches.
  • Scene editor with a relevance score per shot, a chat edit agent, a script editor, brand kits, copy-your-channel-style, team workspaces and an API (POST/GET /api/v1/videos).

Use cases

How teams use AutoVideoStudio

Faceless history channel

A creator types "Taj Mahal" and gets a sourced documentary, two thumbnails, chapters and a Short.

Coaching institute explainers

An education team turns a syllabus topic into an explainer where each claim is checked against a cited source.

Agency running several channels

An agency keeps one brand kit per channel, pastes each channel's past titles to match its style, and queues topics in a series.

Developer workflow

A media team creates videos from its own system through the /api/v1/videos endpoint.

Under the hood

Where AI is used, and where it is not

The model

Scripts and chat edits run on a local Qwen 27B model on Surge's own GPU first. Claude is used only if the local model fails or its output is rejected. If both fail, an extractive script built from verbatim source sentences is used (no AI).

Rules and code

Code rejects scripts with bad JSON, unknown source ids or a length outside 60 to 150 percent of target. The fact-check gate matches claims to sources by keyword share and exact numbers. Pacing rules set cut lengths. Licence filters accept only PD/CC0/CC BY/CC BY-SA. Chat edits go through a deterministic command parser before any model.

People in the loop

When the fact-check gate blocks a claim, the user must edit it or explicitly choose to render anyway. The user reviews and edits scenes, script and captions before re-rendering.

Book a working session

A working session with our architects on one of your use cases. You bring a process and sample documents; we leave you with a scoped plan: model, hosting, integrations, timeline and cost range.