Will Smith eating spaghetti is the official benchmark of AI evolution
Dark Narr · 2026-02-08 · ai-made · 519,473 views
Made by AI
Model: various (compilation) · Series: Will Smith spaghetti benchmark
Evidence: Description: the meme 'accidentally turned into the most accurate benchmark for AI evolution', stepping through 2023, 2024, 2025 and 2026 outputs.
Human role: Compilation and narration by the uploader.
Pipeline: Compilation of AI video outputs 2023 → 2026
Lore: will-smith-spaghetti
What's in the video
Description written by Gemini, which watched and listened to the whole video.
Summary
This short video, presented by a synthetic narrator on the channel "Dark Narr," surveys the evolution of generative AI video from 2023 to 2026 using the famous meme benchmark of "Will Smith eating spaghetti." It contrasts the uncanny, morphing outputs of 2023–2025 models with a high-fidelity 2026 scene highlighting Kling 3.0's multi-shot cinematic cuts and integrated audio generation.
What is shown
- [00:00 - 00:08]: Early 2023 AI video clips featuring Will Smith with severe spatial inconsistencies, distorted hands, and morphing noodles.
- [00:08 - 00:13]: 2024 AI generation clips demonstrating clearer facial features and smoother motion, including a clip shouting into a pasta bowl.
- [00:14 - 00:19]: 2025 AI video showing photorealistic beach lighting and anatomy, but still displaying slightly awkward, uncanny chewing dynamics.
- [00:20 - 00:57]: A 2026 cinematic scene depicting Will Smith and a young man eating spaghetti on an outdoor balcony overlooking a city skyline, showcasing shot-reverse-shot dialogue editing, lip-synchronization, and realistic food physics.
- [00:58]: Outro image reading "2023 → 2026" depicting a cyborg Will Smith eating noodles.
Claims & numbers
- The narrator states that "Will Smith eating spaghetti is the official benchmark of AI" [00:01].
- The narrator states that in 2023, "AI couldn't even get hands right" [00:04].
- The narrator states that in 2024, "faces improved, movements smoother" [00:09].
- The narrator states that in 2025, AI was "almost real, but still uncanny" [00:15].
- The generated characters state that Kling 3.0 "can create multiple scene cuts like this with a single prompt" [00:29].
- The generated characters state that the model "knows when to cut to whoever is talking" [00:39].
- The generated Will Smith states that "all this audio was also generated with the same prompt" [00:43].
Notable quotes
- "Will Smith eating spaghetti is the official benchmark of AI." (Narrator, [00:01])
- "I heard it can create multiple scene cuts like this with a single prompt." (Young man character, [00:29])
- "Study harder, kid. Eat your spaghetti." (Will Smith character, [00:53])
Assessment
This is a social media showcase highlighting recent progress in generative video models, specifically spotlighting Kling 3.0's multi-scene and native audio capabilities. While it faithfully tracks real-world milestone clips from the community timeline, it presents a curated generation without showing the prompting UI or generation runtime.
Lyrics & themes
The narration and dialogue humorously trace the history of generative video through internet lore:
- "Will Smith eating spaghetti is the official benchmark of AI" [00:01]
- "Uncle Phil, come try this!" [00:12]
- "Study harder kid. Eat your spaghetti." [00:53]
Lore & references
- Will Smith eating spaghetti: The definitive 2023 viral video meme (originally produced via ModelScope) that became the universal running joke and de facto progress benchmark for AI video.
- "Uncle Phil": A reference to Philip Banks, Will Smith's uncle in the television sitcom The Fresh Prince of Bel-Air.
- Kling 3.0: Kuaishou's 2026 video foundation model featuring native multi-shot "AI Director" scene cutting and synchronized voice/audio generation directly from text prompts.
Visual style & craft
- The video combines historical short-form AI generation clips edited together with burned-in subtitles and synchronized background sound effects.
- The 2023 footage features characteristic early-diffusion artifacts: fluid melting, floating pasta, and morphing digits.
- The 2026 sequence demonstrates modern world-model coherence, cinematic focal blur, stable lighting across different camera angles, and natural mouth/hand interaction with cutlery and noodles.
Described by gemini-3.8-flash on 2026-09-29 from the video's audio and frames.