Video summary

I Made Bhagat Singh Movie using AI (Full workflow)

Main summary

Key takeaways

Technology

Summary of the AI Bhagat Singh Movie Workflow (from the subtitles)

Overview / What the video claims to teach

  • A complete 9-step “masterclass” workflow for making a full AI-generated movie (example: Bhagat Singh), including:
    • Research → Scriptwriting → Storyboarding → Character sheets → Video generation → Music & pacing → Editing/polish
  • It also covers how to monetize, including an AI-movie submission/contest with rewards.
  • The host argues most AI films become “junk” due to repeating issues, then explains fixes.

Common problems the tutorial targets

  1. Character consistency

    • The AI may change the hero’s identity (e.g., different gender/person) across scenes.
    • Fix: use proper character sheets and reuse them consistently.

  2. Audio / dialogue inconsistency

    • Text-to-voice tools can create different voices each run, breaking continuity.
    • Fix: generate voice once and dub consistently (mentions 11Labs).

  3. Visual continuity / camera logic

    • Shot descriptions can drift, causing unrelated changes (e.g., camera pans while hair detaches/floats).
    • Fix: consistent scene design and a more controlled pipeline for camera/lens/style.

Tools and platforms mentioned

  • Sponsor / tool platform: Akul AI
    • Used for AI images/videos; also linked to the contest.

  • Research automation: AsiAI AI Browser

    • Described as a browser that reads articles and compiles a research folder of sources (including letters/photos/jail notes/FIR mentioned).
    • Concern raised: chat assistants can “elucidate/lie,” compounding misinformation.

  • Alternative references for voice/camera (conceptual)

    • Claude / ChatGPT (mentioned for extracting facial geometry, via image-to-description concept—subtitle text is unclear).
    • Script/story prompting via AI (general AI usage; no single tool name beyond general ChatGPT mention).

  • Image generation / scene frames

    • Mentions Video to Frames for reference frames.

  • Video generation models (cost/behavior tradeoff)

    • Seedance / “Seet Dunn” variants:
      • Seedance 2 and Seedance 2.5 (expensive; used for higher movement like fighting/running)
    • “Cling 3”
      • Cheaper; used for lower movement scenes

  • Music tools

    • Artlist and YouTube for copyright-free music selection
    • Sono AI to generate a flute/similar version when copyright blocks a real song

  • Video editing / polish concepts

    • Mentions an editing learning channel: Create Slabs (spelled unclearly in subtitles)
    • Mentions LUTs and color matching
    • Mentions credits/app and “apologia” for credits/follow (unclear exactly which apps)

Step-by-step workflow taught

Step 1: Research & ideation

  • Use AsiAI AI Browser to automate research:
    • The workflow implies giving access to the computer.
    • The AI compiles a folder containing:
      • story outline, real photos, fellow freedom fighters
      • letters, jail notes, FIR, and historical context
  • Example misinformation warning:
    • AI research can get details wrong (e.g., a slogan attribution claim about “Inquilab Zindabad”).

Step 2: Scriptwriting & storytelling (Hero’s Journey)

  • Uses Hero’s Journey structure:
    • normal life → problems → lowest point → bounce back → transformation
  • Suggests:
    • decide the ending first
    • plant thorns/problems throughout the story
  • Notes:
    • “Good scripts” have uncertainty/rhythm, and AI can assist in drafting.

Step 3: Character sheets & scene design (consistency fix)

  • How to create reusable character sheets:
    • Start from a reference image (real Bhagat Singh photo)
    • Extract facial geometry (via Claude/ChatGPT conceptually)
    • Create character sheets on a grey background to improve results
  • Emphasizes:
    • Don’t assume any random reference image will preserve continuity.
    • Create separate sheets for:
      • different life stages
      • other characters

Step 4: Storyboarding + “end film looks” (cinematic framing cheat codes)

Cheat codes (2 methods)

  1. Picture Method (cinematography reference-based)

    • Pick a film whose cinematography you like (example: Odyssey)
    • Download trailer and use Video to Frames to extract still frames
    • Build a prompt framework with letters:
      • P = photographic style
      • C = camera composition
      • T = timing & lighting
      • U = use of film & effects
      • R = render style
      • E = extra details

  2. Kinesthetic skills (prompting into “cinematic skill”)

    • Use AI + reference frames to identify:
      • best camera angles/lenses (mentions brands/types conceptually like IMAX/Canon/Red)
      • per-frame camera/lens choices

Continuity / quality improvements

  • Repeated image edits can cause AI-face artifacts (e.g., “square face” look).
  • Fix:
    • reapply the character sheet and instruct AI to replace from the sheet.
  • To avoid flat/same-looking results:
    • recommend LUTs / color grading filters
    • apply LUTs per scene with different variants (“grades”/looks)
  • Wrong aspect ratio:
    • fix by adding cinematic letterboxing (bars)

Step 5: Music and pacing (pre-edit planning)

  • Generate music before creating the full video timeline
    • This avoids wasting limited AI video generations (credits are limited).
  • Music selection:
    • prefer copyright-free tracks from Artlist/YouTube
  • Key concepts:
    • Pacing: align image cuts to beats/tempo
    • Needle drop: cut music at key moments to increase tension/emotion
  • Copyright issue example:
    • Using Vande Mataram triggered copyright during YouTube checking.
  • Solution:
    • use Sono AI to generate a flute/similar version after providing an mp3 and requesting a comparable style.

Step 6: Video generation with controlled camera motion + emotion

  • Generate video clips from storyboard images:
    • upload scene image → add emotion and camera movement
  • Camera movement sourcing:
    • mentions websites like:
      • “I Can Be”
      • “ai camera moments.com”
  • Model selection based on motion needs:
    • Cling 3 for subtle movement / cheaper runs
    • Seedance 2 / 2.5 for heavy action (running/fighting) despite higher cost
  • Logical editing/rhythm:
    • use cut logic for continuity (example given about why someone doesn’t catch the character and how the next scene should be sequenced)
  • Audio limitation handling:
    • the host notes AI video generation can produce wrong/mismatched audio, so dialogue is replaced later.

Step 7: Voice generation & dubbing (consistency)

  • Claims AI video models produce different audio each run, so:
    • use 11Labs to create consistent voice output
    • dub scenes manually for consistent dialogue

Step 8: Editing and polish (lip sync + color matching)

  • Editing turns clips into a “movie”:
    • Fix lip-sync mismatch after dubbing using a workaround:
      • overlay a policeman’s B-roll/insert shot when lips don’t match
  • “Polished” look through:
    • matching fonts, colors, and color palette
    • LUT-driven consistency
  • Credits mention:
    • the host credits another editor/producer (e.g., Ad Bhai / Ashad Bhai) for generation/editing/pacing work.

Monetization / contest mention

  • Video is sponsored by Akul AI.
  • Mentions a contest:
    • upload your AI movie
    • rewards:
      • $3000 for the winner
      • $100 for each qualified submission (host claims guaranteed)
    • deadline: September 15
  • If not selected:
    • host offers an alternate route via Instagram DM/tag for another subscription reward.

Main speakers / sources (as referenced)

  • Main speaker/host: creator of the tutorial video (referred to as “I”)
    • also mentions Ashad Bhai as editor/producer
    • “Love” appears as a name for kinesthetic skills
  • Referenced collaborators/crew: Ashad Bhai / Ad Bhai (editor/producer credited)
  • Referenced tools/vendors: Akul AI, AsiAI AI Browser, 11Labs, Sono AI, Artlist, Video-to-Frames, camera-movement reference sites (e.g., I Can Be, ai camera moments.com), and generic video editors

Original video