> [!abstract] Summary > Overview and analysis of generative AI tools entering the audiovisual and CG industry — covering the current tool landscape, potential impacts on workflows and jobs, and the ethical, legal, and ecological questions the industry must face. --- ## Source **GÉNÉRATION IA — L'arrivée de l'IA générative dans l'industrie audiovisuelle** Channel: [nextage ai](https://www.youtube.com/@nextageai) — Published February 2026 <iframe width="100%" height="500px" src="https://www.youtube.com/embed/nVh_UrSuZs4?si=t6o20tKMIRukZhFX" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe> --- ## What the video covers A popularization piece and state-of-the-art assessment of generative AI in the audiovisual sector. Three main areas: 1. **Landscape of public-facing generative AI tools** — what's available, what each tool category can do. 2. **Impact analysis on the industry** — how these tools change production workflows, roles, and creative processes. 3. **Ethical, legal, and ecological questions** — the harder problems the industry needs to address. --- ## Generative AI Tool Landscape (as of early 2026) ### Image Generation | Tool | Notes | | ---------------- | --------------------------------------------------------------------- | | Midjourney | High-quality stylized image generation, widely adopted in concept art | | Stable Diffusion | Open-source, highly customizable, can be fine-tuned on studio assets | | Adobe Firefly | Integrated into Creative Cloud, trained on licensed content | | DALL·E (OpenAI) | Accessible via ChatGPT, good for rapid ideation | ### Video Generation | Tool | Notes | | ---- | ----- | | Sora (OpenAI) | Text-to-video, cinematic quality but limited control | | Runway (Gen-3) | Image/video-to-video, widely used in post-production | | Kling / Pika / Luma | Rapid iteration tools, growing in quality | | Stable Video Diffusion | Open-source video generation | ### 3D & CG-specific | Tool | Notes | | ---- | ----- | | Luma AI / NeRF tools | Photogrammetry replacement, scene reconstruction | | Gaussian Splatting | Real-time 3D scene representation from photos | | AI denoising (NVIDIA, Intel) | Already production-standard in Houdini/Karma, Arnold, Renderman | | Generative texturing (Substance AI) | AI-assisted material generation inside existing pipelines | --- ## Impact on the Audiovisual Industry > [!important] The impact is not uniform — it depends heavily on the role, the studio size, and the type of production. ### What changes - **Concept & Previs** — speed increases dramatically. Mood boards, storyboards, previs iterations that took days can take hours. - **Texture & Lookdev** — AI-assisted material generation reduces repetitive work; artists focus on direction and refinement. - **VFX & Compositing** — rotoscoping, cleanup, and some FX tasks are partially automatable. - **Advertising & Short-form** — the most disrupted segment. Full AI-generated spots are already produced at commercial quality. ### What doesn't change (yet) - Complex character animation and simulation still require artists. - Art direction, shot design, narrative decisions remain human. - Pipeline integration, technical problem-solving, and quality control still need skilled generalists. > [!tip] The most resilient profiles are those who combine technical pipeline knowledge with creative judgment — not pure execution specialists. --- ## Ethical & Legal Questions > [!warning] These questions are unresolved at the industry level as of 2026. ### Training data and copyright - Most generative models were trained on scraped internet data, including copyrighted artwork and film. - Artists and studios are challenging this legally (ongoing lawsuits: Getty vs. Stability AI, class actions from illustrators). - Some tools (Adobe Firefly, Getty's Generative AI) offer **commercially safe** outputs trained only on licensed content. ### Consent and attribution - Artists whose work was scraped without consent have no current legal recourse in most jurisdictions. - Some platforms (ArtStation, DeviantArt) now offer opt-out — but opt-out is the default, not opt-in. ### Job displacement - Entry-level and mid-level execution roles (matte painters, junior compositors, concept artists at certain scales) are under pressure. - Debate continues on whether AI creates new roles faster than it eliminates existing ones. --- ## Ecological Questions > [!note] Often overlooked in the excitement around capabilities. - Training large AI models has a significant carbon footprint (GPT-4 estimated at thousands of tonnes of CO₂). - **Inference** (using the model) is cheaper, but at scale — millions of daily generations — the cumulative cost is substantial. - Data centers running these models require large amounts of water for cooling. - The industry has not yet established standards for reporting AI-related environmental costs in production.