Self-hosted · $0/video possible · 21 agents

Type a topic.Get a video.

VigenFlow is an AI pipeline that turns any topic into a publish-ready YouTube video: research, script, voice, video per scene, consistent characters, subtitles, SEO, thumbnail and shorts. Fully automated. Runs on your machine.

21 production agents Multi-language YouTube · Facebook · TikTok · Instagram
vigenflow

The Accountant Who Found Two Sets of Books In The CEO's Safe

0views
0likes
+0subs

How it works

Three steps to a finished video

No UI needed. No cloud subscription. Your machine, your keys, your content.

Step 01

Type your topic

One command. VigenFlow researches angles, pain points, and trending hooks for your topic automatically. No templates, no manual input.

python main.py "The Accountant Who Found Two Sets of Books"
Step 02

21 agents run end-to-end

10 pipeline steps, 21 specialized agents. Each step has a dedicated AI reviewer with retry logic. The pipeline stops if quality falls below your threshold.

Topic Review → Research → Script → Character
→ Images → Voice → Video → Subtitles
→ Effects → SEO → Thumbnail → Shorts
→ Final Review (score gate)
Step 03

Get publish-ready output

Final video, 3 short clips, thumbnail, SEO metadata and chapters. All scored. Ready to upload or schedule.

✓ Score: 8.6/10 - Ready to upload
✓ video:     final_video.mp4  (36 MB)
✓ shorts:    hook / climax / moral
✓ thumbnail: thumbnail.png
✓ seo:       seo.json

Provider Chain Demo

One command.
Four visual tiers.

Same topic:python main.py "The Accountant Who Found Two Sets of Books"Four visual outputs. VigenFlow auto-selects based on your API keys.

Free tier: $0/video
Paid tier: bring your own keys
Free Stock$0 / video

Pexels · Pixabay · Unsplash

Real footage from Pexels, Pixabay, and Unsplash. No API key needed.

Free AI$0 / video

Gemini Flash · MLX Flux · Pollinations

Gemini Flash free tier, MLX Flux on-device (Apple Silicon), or Pollinations cloud. No API key required.

AI + Character~$0.30 / video

Gemini img2img · Flux Kontext · Imagen 3

Character seeded once per channel. Gemini img2img, Replicate Flux Kontext, or Imagen 3 keep the same face every scene.

AI Video~$1–2 / video

Seedance · Kling · Wan2.1 · Hailuo · Runway

Scene-level video with actual motion. Swap provider in one config line.

Same topic · Different provider · No manual editing

Features

One command. Publish-ready video.

21 agents, each independently reviewable, retryable, and replaceable. Review and regenerate any scene before final export.

Unique

AI Review Loop

Every step has a dedicated reviewer agent. Script, images, audio, thumbnail, all scored and auto-retried up to 3×. No other tool does this.

Unique

Video Per Scene

Each script scene gets its own image sequence, voice clip, subtitle block, and timing, assembled into a full cinematic video.

Unique

Character Consistency

Seed a character once per channel. Same face, same visual style, across every scene of every video, automatically.

Unique

Score Card

Every run scored on hook, CTA, arc, atmosphere, tension. Monetization risk score included. Engagement prediction per video.

Also included

Multi-LLM Chain

Claude → Gemini → GPT-4o → Grok. Auto-fallback per step with no provider lock-in. Free tier first.

Voice Chain

ElevenLabs → OpenAI TTS → Kokoro (local, no API key, runs on-device) → Edge TTS (free cloud) → gTTS. Pick your quality tier or let the chain auto-select.

Image Chain

Imagen 3 → DALL-E 3 → Pollinations → Pixabay → Pexels. 8-provider cascade. Always has an image.

Video Chain

Seedance · Kling · Wan2.1 · Hailuo · Runway. Swap providers in one config line. New providers added regularly as the market evolves.

Auto Shorts

3 vertical clips (hook / climax / moral) with karaoke-style yellow word highlight. Sized for TikTok, Facebook Reels, Instagram Reels, and YouTube Shorts.

Market Opportunity

$3.4B
AI video market by 2033
from $788M in 2025
12h+
Saved per video
vs. manual editing
$80K–300K
Monthly revenue
top faceless channels
38%
New channels faceless in 2025
up from 12% in 2022

Why VigenFlow

Not another clip tool.

Most tools clip, caption, or repurpose existing footage. VigenFlow creates from scratch: a full production pipeline in one command.

Review loop nobody else has

Every agent has a dedicated reviewer. Script scored on hook, arc, and CTA. Images checked for watermarks, duplicates, and visual consistency. Failed steps auto-retry up to 3×.

Review & regenerate any scene

Not happy with a scene's image or voice? Review the output, regenerate just that scene, and re-assemble. Full manual control, no need to restart the whole pipeline.

Character consistency across videos

Seed a character once per channel profile. Same face, same visual style, every scene of every video. Works with Imagen 3 and Pollinations FLUX.

Your keys. Your machine. Zero subscription.

$0/video is real: Gemini free tier for LLM, Kokoro (local) for voice, Pollinations (free) for images. Paid providers optional, and you bring your own keys.

Compare

How we stack up

No other tool is open-source, self-hosted, and has a full AI review loop at feature parity.

ToolPriceOpen-sourceSelf-hostedFull pipelineAI review loopMulti-LLMShorts genScore cardVietnameseFree tier
VigenFlow
$0*
In*** AI
$28/mo
Sy***ia
$29/mo
Auto***ts
$19/mo
Pi***ry
$19/mo
He***en
$24/mo

* $0/video possible using Gemini free tier + Kokoro (local TTS) + Pollinations (free images). Paid APIs optional.

Use Cases

Any channel. Any niche.

Drama, finance, wellness, animation, documentary. VigenFlow adapts to your channel through a YAML profile, not a template picker.

Drama / True Crime

Faceless storytelling channels

Input a story angle → 15-scene cinematic video, consistent characters, male voice, karaoke subs. 3 viral shorts auto-clipped and sized for TikTok.

  • 15 scenes / video
  • ~4 min production
  • 3 shorts auto-clipped
Finance, History, Wellness...

Any topic, any visual style

Finance explainer with an animated host, history documentary with archival-style images, wellness channel with a warm narrator. Each channel gets its own YAML profile that controls voice, visual style, review thresholds, and language.

  • YAML profile per channel
  • EN, VI, and 10+ languages
  • Unlimited channels
Agencies & Power Users

Multi-channel production at scale

Run parallel pipelines per channel. Each with its own voice, visual style, LLM chain, and review thresholds. Score card tracks quality across every run.

  • Parallel channel runs
  • Per-channel profiles
  • Full score card

Testimonials

What creators are saying

The reviewer loop alone saved me 3 hours per video. Scripts that used to go through 4 manual rounds now pass automatically or get flagged with exact fixes.

Marcus T.
Marcus T.
Drama Channel · 127K subscribers

I was spending $400/month on freelancers for voiceover and editing. VigenFlow replaced all of it. First month I made 12 videos, cost me less than $8 in API calls.

Sarah K.
Sarah K.
Finance Channel · 43K subscribers

Character consistency is a game changer. My audience recognizes the narrator across every video now. Retention went up 22% after switching.

James W.
James W.
History Documentary · 89K subscribers

Running 4 channels in parallel. Each has its own profile and voice. The score card tells me exactly which video to post and which one to redo.

Linh P.
Linh P.
Agency Owner · 6 active channels

The free stack is real. I used Gemini free tier plus Kokoro for voice and Pollinations for images. Ran 30 videos, spent $0. Quality surprised me every time.

Omar A.
Omar A.
Wellness Channel · 31K subscribers

I do Vietnamese content. FPT TTS integration works natively, subtitles are accurate, and the SEO output is localized. Nothing else on the market does this.

Nguyen H.
Nguyen H.
Lifestyle Channel · 58K subscribers · VI

The reviewer loop alone saved me 3 hours per video. Scripts that used to go through 4 manual rounds now pass automatically or get flagged with exact fixes.

Marcus T.
Marcus T.
Drama Channel · 127K subscribers

I was spending $400/month on freelancers for voiceover and editing. VigenFlow replaced all of it. First month I made 12 videos, cost me less than $8 in API calls.

Sarah K.
Sarah K.
Finance Channel · 43K subscribers

Character consistency is a game changer. My audience recognizes the narrator across every video now. Retention went up 22% after switching.

James W.
James W.
History Documentary · 89K subscribers

Running 4 channels in parallel. Each has its own profile and voice. The score card tells me exactly which video to post and which one to redo.

Linh P.
Linh P.
Agency Owner · 6 active channels

The free stack is real. I used Gemini free tier plus Kokoro for voice and Pollinations for images. Ran 30 videos, spent $0. Quality surprised me every time.

Omar A.
Omar A.
Wellness Channel · 31K subscribers

I do Vietnamese content. FPT TTS integration works natively, subtitles are accurate, and the SEO output is localized. Nothing else on the market does this.

Nguyen H.
Nguyen H.
Lifestyle Channel · 58K subscribers · VI

FAQ

Common questions

The pipeline is free and open source. With Gemini (free tier) + free stock providers, you can generate full videos at $0. Optional: ElevenLabs voice (~$0.30/video), DALL-E 3 images (~$0.50/video).

If you can run `pip install` and `python script.py` in a terminal, you can run VigenFlow. No web dev or ML background needed.

YouTube long-form, YouTube Shorts, TikTok, Facebook Reels, Instagram Reels. Aspect ratio, duration, LUFS target, and subtitle style auto-adjust per platform.

10-40 min depending on providers. ElevenLabs + Imagen 3 = 15 min. Free-only stack = 35 min. Fully unattended.

The voice chain tries in order: ElevenLabs (premium, most natural), OpenAI TTS (paid, solid quality), Kokoro (free, runs locally on your machine with no API key), Edge TTS (free cloud, no key needed), gTTS (basic fallback). For zero cost, Kokoro gives the best quality. Runs on Apple Silicon or CPU. Edge TTS works on any machine with no setup.

Yes. Vietnamese uses FPT AI TTS natively. Any language Gemini supports works. ElevenLabs multilingual or Edge TTS as fallback.

All three. Requires Python 3.11+ and FFmpeg. Tested on macOS Apple Silicon, Intel, and Ubuntu. Windows works via WSL2.

Yes. A desktop GUI app (Mac, Windows, Linux) is in development for users who prefer not to use a terminal. Same pipeline under the hood, visual topic input, live progress view, and one-click export. No coding required.

Yes. A cloud-hosted version is planned so you can type a topic in the browser and receive your video without installing anything locally. It will use the same pipeline and scoring system, billed per video. Early access list is open above.

Auto-upload via the YouTube Data API is nearly complete. When enabled, videos that pass the score gate upload automatically with the generated title, description, tags, chapters, and thumbnail. No YouTube Studio needed.

Early Access

Request a Demo

We're onboarding select creators and agencies. Tell us about your channel and we'll reach out to set up a live walkthrough.

No spam. No sales pitch. Just a real walkthrough of the tool.