Generative AI Video Specialist (R&D intern)

We are looking for a specialist who can push the boundaries of current AI models to create long-form, coherent visual narratives.

visual consistency pipeline

Key Technical Responsibilities

• Character Identity Preservation:

  Develop rigorous workflows using

  Midjourney (Cref/Sref)

  and

  Stable Diffusion (ControlNet/LoRA)

  to ensure a characters facial features, clothing, and body type remain 100% consistent across hundreds of generated shots.

• Generative Video Orchestration:

  Utilize advanced Image-to-Video tools (Runway Gen-3, Kling AI, Luma Dream Machine) to animate characters with precise physical movements (walking, running, sitting) without warping or "hallucinations."

• Audio-Visual Synchronization:

  Implement lip-syncing technologies (Hedra, LivePortrait, or Wav2Lip) to map AI-generated voiceovers (ElevenLabs) onto video characters with realistic emotive expressions.

• Scene Continuity:

  Ensure background environments (rooms, streets, landscapes) remain spatially accurate when cutting between different camera angles (Wide shot to Close-up).

• Post-Processing & Assembly:

  Stitch AI-generated assets into cohesive timelines using Adobe Premiere Pro or DaVinci Resolve, applying upscaling (Topaz Video AI) to meet HD/4K standards.

Required Skill Set (The "Must-Haves")

1. Image Generation & Control

• Mastery of Midjourney v6+:

  Deep understanding of parameters like --cref (Character Reference), --sref (Style Reference), --chaos, and --weird to lock in consistency.

• Stable Diffusion (ComfyUI/Automatic1111):

  Knowledge of ControlNet (OpenPose, Canny, Depth) to force AI characters into specific poses is a huge plus.

2. Video Generation Models

• Hands-on experience with

  Runway Gen-3 Alpha

  or

  Kling AI

  (specifically utilizing "Motion Brush" or "Camera Control" features).
• Ability to minimize "morphing" artifacts where faces distort during movement.

3. Audio & Performance

• Proficiency with

  ElevenLabs

  for generating emotive speech.
• Experience with

  Hedra (Character-1)

  or

  SadTalker

  for performance capture (making a static image talk and act).

4. Editing & Workflow

• NLE Proficiency:

  Adobe Premiere Pro / CapCut Desktop.

• Upscaling:

  Topaz Video AI or Magnific AI.

Mandatory Pre-Interview Assignment (Proof of Concept)

IMPORTANT:

The Task:

The Scenario to Build:

1. Character:

   A young male professional (Indian descent, glasses, casual hoodie).

2. Sequence of Events:

• Shot 1 (Wide):

  The character is walking down a busy Bangalore street (Daytime).

• Shot 2 (Close-up):

  The character stops, looks at a phone, and smiles.

• Shot 3 (Indoor):

  The character enters a modern office and sits at a desk.

• Shot 4 (Dialogue):

  The character looks at the camera and delivers a 10-second monologue (use any AI voice): "The neural network is finally stable. We are ready to launch."

3. The Requirement:

   The character's face and clothes MUST look identical in all 4 shots.

Submission Format:

• Upload the 1-minute video to Google Drive/YouTube (Unlisted) and attach the link to your application.
• Note: Applications without this proof of concept will be automatically rejected.