One email each morning — yesterday's AI, sortedGet it in your inbox

METAL LAB

Higgsfield's First AI Feature Film Screened Alongside Pixar, Disney at SIGGRAPH

"KÖK BÖRÜ," based on a Kazakh folktale, was made with Seedance 2.0 and released with its full prompts

이미지: METAL LAB 생성

Summary

  • Higgsfield's first AI animated feature, "KÖK BÖRÜ," was unveiled at SIGGRAPH 2026 LA alongside screenings from Pixar, Disney, and NVIDIA
  • It was made by running Seedance 2.0, developed by ByteDance Seed, on the Higgsfield platform, with the full prompts and assets released as open source
  • The film is an entry in the $1 million Higgsfield Global Film Festival, and the published prompts include a six-shot emotional sequence with dialogue written in Kazakh
Video from the source
작품명
KÖK BÖRÜ (쾨크뵈뤼)
상영 무대
SIGGRAPH 2026, 미국 LA — 픽사·디즈니·엔비디아 상영작과 나란히 공개
사용 모델
Seedance 2.0(개발: ByteDance Seed), 힉스필드 플랫폼에서 구동
공개 범위
프롬프트·에셋 전량 오픈소스 공개, 중복 제거 후 프롬프트 2건
연계 행사
100만달러 규모 힉스필드 글로벌 필름 페스티벌 출품작
발표
Higgsfield AI, 2026년 8월 17일 X 게시

Higgsfield's first feature takes the SIGGRAPH stage

Higgsfield has unveiled its first AI animated feature, "KÖK BÖRÜ." According to an announcement posted on its X account, the film was screened at SIGGRAPH 2026. SIGGRAPH is the world's largest conference on computer graphics, where companies like Pixar, Disney, and NVIDIA unveil new technologies and works each year. Higgsfield said this screening took place on the same stage as works from these established studios.

Higgsfield is a platform that bundles video and image creation tools built on generative models from companies such as OpenAI, Google, and ByteDance. This particular work was made by running Seedance 2.0, developed by ByteDance Seed, on the Higgsfield platform. This is a different version from Seedance 2.5, which reportedly launched in early August, and the published prompt metadata explicitly labels the version as "seedance_2_0."

이미지: X — 미디어·생성AI

A Kazakh nomadic narrative rendered with Seedance

The story in the published prompt opens inside a ger (yurt) in a Kazakh nomadic aul (village) during a midwinter blizzard. Aksakal, the village elder, orders a dawn evacuation before the storm hits, but the protagonist Ana's young daughter Umai has gone missing in an earlier storm and has not returned. Ana, believing the evacuation means abandoning the place her daughter would return to, begs for just one more day, but Aksakal quietly yet firmly refuses and signals to Ana's husband, Mergen, to take his wife away.

The prompt breaks this entire scene into six hard-cut shots. The structure is built so the characters' emotions escalate step by step — from suppressed despair to protest, pleading, refusal, and finally restrained grief.

The published prompt: style and prohibition directives

The style directive states verbatim: "STOP-MOTION ANIMATION — stepped frame-by-frame, ANIMATED ON TWOS at a true 12fps." This is followed by "NO smooth interpolation, NO motion blur, NO morphing, NO AI-glide," a directive deliberately blocking the smooth transitions typical of generative video.

Notably, the prompt also instructs that background elements and character motion be handled separately. It flags this as "MOTION SEPARATION — CRITICAL," specifying that hearth flames, smoke, and breath should move smoothly, while characters, fabric, and the yurt walls should move in a choppy, stop-motion style. For painterly texture, the prompt also instructs that canvas texture and oil-brushstroke marks be pulled directly from the style reference images.

Dialogue and directing notes across six shots

The dialogue is written in Kazakh, with emotional directives attached before and after each line. In Shot 1, after a heavy silence, Aksakal issues the order: "Qıs — qatal... Aş böri azıq izdep qayta şabadı." followed by "Köşu qajet." (Winter is harsh... a starving wolf will strike again in search of food / We must migrate.) In Shot 4, Ana pleads with a cracking voice: "Bir-aq kün küteyik! Tağı bir tañ atqanşa... Murşa beriñizşi!" (Let us wait just one more day! Until dawn breaks again... Please show mercy!) In Shot 5, Aksakal ends the exchange, sad but resolute: "Auıldı ajalğa qimaymın. Tañmen jüremiz." (I cannot leave the village to death. We move at dawn.)

The prompt controls Ana's performance with particular precision. It repeatedly instructs that she not scream or fall to her knees, but remain standing, allowing only restrained emotion — welling eyes, a trembling jaw. Screen direction is also fixed: Aksakal always stands on screen right and Ana on screen left, keeping the axis consistent across all six shots.

Sequence structure table

ShotCharacterEmotional stageKey directive
1AksakalAnnouncing the orderA heavy silence, then a gaze that sweeps over the people rather than facing forward
2AnaQuiet despair, hesitationA subtle shift in expression as if about to protest, then stopping
3AksakalConsolingSoftening, with a sad and gentle tone
4AnaRestrained, desperate pleadingStanding, voice low and cracking, but not wailing
5AksakalRefusalA sad but unwavering final declaration
6AnaSuppressed griefNearly motionless expression, a single tear

How to replicate it

To adapt this prompt for a different scene, first place style reference images and character-specific reference images into the designated image slots (formatted like <<<image_1>>>). Then, keep the paragraph defining the stop-motion texture intact, and only swap out the numbered items in the STORY CONTEXT and DIRECTOR'S NOTES sections with the new scene's characters, conflict, and dialogue. If the resulting motion ends up smoothed out, the first thing to check is whether the "NO smooth interpolation"-type prohibition phrases are still present near the top of the prompt. If screen direction flips between shots, re-emphasizing the left-right fixing statement in the GEOGRAPHY & SCREEN DIRECTION section should fix it.

Higgsfield previously released a 110-minute AI feature film with a real cast, "The Cully Hill Boys," on August 16, along with details of its production process and prompts.

Editor's take

What stands out in this announcement is less "where it screened" than "what wasn't hidden." Higgsfield put its work on SIGGRAPH's stage — one of the industry's most prestigious venues — while simultaneously releasing the full prompt text behind it. It reveals exactly which prohibition terms were used to achieve stop-motion's characteristic choppiness, and how even a single tear from an actor's eye was directed. This isn't just showing off the finished product — treating the production process itself as a public asset is a pattern the company has repeated over the past several weeks. Whether in the K-pop action series "Zephyr" or the sci-fi short "Adiliada," Higgsfield has released the raw materials alongside the finished work each time.

Working with earlier generations of AI video tools in practice, the conclusion was always the same: the output looked convincing, but the instructions behind it were hidden, leaving other teams unable to learn from it. When a single directive like "separate motion into background and character layers" — which, as this prompt shows, accounts for most of the stop-motion texture — is documented, other creators attempting the same effect don't have to start experimenting from scratch. That's a real, tangible difference.

Domestic video production teams or solo creators could use this prompt structure directly as a reference. Separating style, character, and environment into image slots, and pinning down emotional stages numerically to give shot-specific directives, is a framework that can be reused across genres. However, elements tied to a specific language or cultural background — like the Kazakh dialogue — shouldn't simply be copied; they need to be adapted into a tone and language that fits one's own story for the result to feel natural.

It seems likely that follow-up entries referencing this structure will appear in the Higgsfield Global Film Festival in the coming weeks. A trend favoring those who don't hide their prompts is already taking shape.