Make a video with Sauna and Higgsfield

SCENE AND STORY
Create a richly colored, boldly composed editorial mixed-media science short with a rhythm full of surprises, explaining why your recorded voice sounds unfamiliar. Not a quiet educational diagram: use oversized printed face cutouts, paper sound ribbons, contrasting display typography and precise shape-match cuts. Follow the same person's voice from the surprise of playback through two sound-conduction pathways to the revelation that the perceived mix is different. One complete story, four distinct compositions, three crisp cuts.

VISUAL SYSTEM
Saturated cobalt blue #2252FF, vivid orange #FF6728 and bright yellow #FFE14A are the main colors, with pure black #000000 and cream for typography and printed photographs. Each composition has one strong dominant background color rather than an all-over rainbow. The character is the same short-black-haired young adult in black-and-white halftone photographic cutouts, with rough white torn-paper borders and consistent appearance. Oversized ears, mouth and recorder details are close-ups of the same subjects, not random new characters. Bold magazine collage and refined 2D motion design: real paper texture, print dots, a little tape and small paper-layer shadows from consistent upper-left light. Every surface is flat matte printed paper: no glow, no neon, no bloom, no light halos, no luminous or backlit edges, nothing emits light anywhere in the film. Not glossy plastic 3D animation or a live-action presenter.
MOTION PRINCIPLES
High energy comes from major changes of scale, silhouette matching, paper actions and typographic composition, not incessant camera shake. Each beat forms its composition rapidly, then preserves a short clear reading window. Objects enter with fast deceleration, a slight overshoot and a stable landing; headlines retain clear letterforms after landing. Cause and effect happen sequentially, with one strongest visual focus at a time. Movement starts on the first frame, and a small loop continues at the end. No empty waiting frames.
OPENING: PLAYBACK SURPRISE
The first frame has a full cobalt background. An enormous halftone face enters from the left and occupies nearly half the picture. On the right, a paper fingertip is already pressing the yellow triangle Play button of an orange recorder. The button clicks down; a vivid orange paper sound wave springs from the speaker, nudging the portrait's head slightly backward. The eyes suddenly widen in recognizable comic surprise. Oversized cream bold type spells exactly "THAT'S ME?!", arriving like three typographic paper pieces slapped rapidly onto the picture. After that brief impact, keep the text readable without covering the eyes or the playback action.
MATCH CUT
Align the triangular Play button with a triangular arrow and cut crisply to an enormous side-profile head cutout on bright yellow. Move from an extreme facial close-up to the complete profile diagram for a strong scale contrast. The head outline is readable, its interior represented by simple non-gory paper layers. The black display headline reads "AIR + BONE"; the plus sign briefly enlarges and settles as the two paths converge. A blue paper sound wave travels from the mouth through outside air to the outer ear. An orange vibration ribbon travels from the throat through head tissues to the same circular inner-ear target. Activate the paths sequentially — each path slides fully into place as a flat matte paper ribbon with no glow and no light emission — then bring them together. Their geometry is clear: not left ear versus right ear, not two people. Push gently toward the inner ear following both colors, revealing two ways the same voice arrives.
INSERT CUT
Cut to vivid orange. The circular inner-ear target becomes a huge black disc in the center, with two paper sound waves passing through it. Preserve detail in the fine blue wave. A yellow low-frequency paper ribbon snaps into place from below, with wider, thicker waves representing the internal vibrations adding bass. Large black words "EXTRA BASS" expand and rebound once in synchrony with the wave, then remain completely clear. Letters do not liquefy or shatter. The two ribbons physically overlap. After the brief impact, allow enough time to understand that the combined wave is fuller. This is additional low-frequency content in the timbre, not a sudden explosion in loudness.
MATCH CUT
Match the circular inner-ear target to the recorder's circular speaker from the opening. Return to cobalt, pulling out to reveal the complete new composition of the character, recorder and waveform. The orange internal-vibration ribbon slides away cleanly like a strip of paper being pulled out. The blue recorded wave remains and travels from speaker through air to the person's ear. Surprise becomes an understanding smile and nod, bringing the story back to its opening. The large yellow headline reads exactly "SAME VOICE. DIFFERENT MIX." on two lines. End with clear typography and small rhythmic movements in the blue waveform: a conclusion, not a new set of graphics.
DATA AND STATS OVERLAY
On top of the existing compositions, add an analytical data layer of paper infographic elements. This layer never changes any described action, headline, layout or cut; every element sits in empty background space, built from the same paper-collage system: real paper texture, rough torn white borders, small upper-left-light drop shadows. Every digit and label is printed complete from its first visible frame, extra-bold, generously sized, never morphing, never redrawing, never counting up. Percentage bars never animate: every bar that carries a percent label is printed already complete in its final, correct state from its first visible frame and stays frozen like a printed magazine chart. Only elements without percentages may animate through their containers, with fast deceleration and a slight overshoot before locking still. Every percentage is proven by geometry: each percentage bar sits on a visible full-length track with a thin black outline that represents exactly 100%, and the fill length always equals the printed percent of that track — 100% means the track is completely full edge to edge, 50% means the fill occupies exactly the left half of the track and the right half of the track is visibly empty inside its outline. A printed percent never contradicts its bar in any frame; if the label says 50%, the fill is exactly half, never full; if the label says 100%, the track is completely full, never partial.
In the opening composition, on the cobalt background directly above the orange recorder, place a small live playback meter: a cream paper chip holding a row of five short vertical black VU bars that bounce rhythmically with the playing sound wave, and next to it a solid red circle with black text reading exactly "REC 0:15" printed static on the same chip. The meter contains no other marks and stays clear of the face, the headline and the fingertip.
In the second composition, in the open yellow space in the lower-left area below the black "AIR + BONE" headline block, place a legend group of three stacked paper pieces: a cobalt blue rectangular chip with cream text reading exactly "AIR 50%", beneath it an orange rectangular chip with black text reading exactly "BONE 50%", and beneath the chips one thin horizontal split bar of the same width, drawn as a full-length paper track with a thin black outline representing the whole voice, 100%. The split bar is printed already complete and correct from the moment the legend appears and never animates: its left half is solid cobalt blue, its right half is solid orange, and the blue-orange boundary sits exactly at the center of the track, so each color occupies exactly half of it — matching the two equal "AIR 50%" and "BONE 50%" chips, never more, never less. The whole group stays perfectly still, flat matte paper color with no glow. Additionally, tag each pathway with one small round paper measurement tag of the path's own color: a cobalt blue tag with cream text reading exactly "340 M/S" in the clear yellow space above the crown of the head, pointing at the blue air wave, and an orange tag with black text reading exactly "3500 M/S" in the clear yellow space below the jaw, pointing at the orange vibration ribbon, stating the real speed of sound in air versus bone. The legend and both tags stay clear of the headline, the head outline and both waves.
In the third composition, in the clear orange space of the upper-left corner, place one cream paper chip containing a minimal equalizer chart: five vertical bars on a thin black baseline. The three right bars are short and black and stay still. The moment the yellow low-frequency ribbon snaps into place, the two leftmost bars grow with a slight overshoot and lock, the first clearly tallest, both colored the same yellow as the ribbon; under them one black label reads exactly "LOW" with smaller black text reading exactly "<500 Hz" beneath it. The chart states that the internal vibrations boost the low-frequency band, in sync with the bass cue. No other axis marks or numbers on the chart.
In the fourth composition, in the open cobalt space of the upper-left corner above the character's hair, place a two-row comparison stat built on two identical horizontal paper tracks with thin black outlines: both tracks exactly the same length and height, sharing the same left edge and the same right edge, stacked one directly above the other; each full track represents 100%. The entire stat is printed already complete and correct from its first visible frame and never animates, like a chart printed on paper. Top row, labeled in cream exactly "HEAD" with cream text reading exactly "100%" just past the track's right end: this track is completely full from edge to edge — left half cobalt blue, right half orange — full track means 100%. Bottom row, labeled in cream exactly "MIC" with cream text reading exactly "50%" just past the track's right end: this track is filled only on its left half with solid cobalt blue, the fill ending exactly at the track's midpoint, directly below the blue-orange boundary of the HEAD track above, while the entire right half of the MIC track is empty inside its black outline — the visibly empty half is what makes the 50% true. There is no orange anywhere in the MIC track in any frame. The HEAD track is always 100% full, the MIC track is always exactly half full; neither label ever sits next to a bar of the wrong length. It states analytically that a recording keeps only the air path: the full two-color HEAD bar against the half-length blue MIC bar mirrors the orange ribbon being pulled out of the main scene. The stat stays clear of the character, the recorder and the headline.
TEXT CONTROL
Only four headlines appear, strictly in this order: "THAT'S ME?!", "AIR + BONE", "EXTRA BASS", "SAME VOICE. DIFFERENT MIX." Use one consistent extra-bold sans-serif. Short headlines occupy a major part of the composition with safe margins; only the current headline remains on screen. In addition, only the data overlay elements carry exactly these labels: "REC 0:15" in the opening; "AIR 50%", "BONE 50%", "340 M/S" and "3500 M/S" in the second composition; "LOW" and "<500 Hz" in the third; "HEAD", "100%", "MIC" and "50%" in the fourth. English spelling, punctuation and letterforms are accurate and stable. Typography participates in the visual action rather than acting as dialogue subtitles. Beyond the four headlines and these data labels, no other text, brands, numbers or watermarks.
AUDIO
One English narrator with a warm, clear, natural voice, surprised and curious energy, brisk but not rushed, never an advertising shout. Begin immediately on the first frame. Speak the following script exactly once, word for word, finishing the final sentence completely, with no paraphrasing or additional dialogue: "Hate your recorded voice? You normally hear it through air and vibrations inside your head. Those vibrations add extra bass. A recording leaves them out, so the same voice suddenly sounds unfamiliar."
Synchronize the sequence semantically: the question accompanies playback surprise; air and vibrations accompany the two pathways; extra bass accompanies the wave overlay; recording accompanies removing the internal pathway. The paper character does not speak additional lines, and the recorder plays no secondary dialogue. Lively plucked bass, dry drum hits and short marimba phrases provide an interesting rhythm, always clearly beneath the narrator. Add a button click, crisp paper slaps when typography lands, short rhythmic accents on match cuts, and a very brief soft bass cue when the low-frequency layer is added. Sound effects support the actions without covering any spoken words.

Make a 60 to 90 second explainer film about Sauna-to-Sauna communication: my teammates can ask my Sauna instead of asking me, so I do not have to be in meetings I was invited to.

BEFORE YOU GENERATE ANYTHING
Write the full narration script and show it to me. Do not generate a storyboard or a clip until I have agreed the words. Then generate ONE storyboard sheet and show me that. Only then generate video.
STRUCTURE
Open on the viewer's problem in the first five seconds, not on the product. The hook is that if you run a team, you are the bottleneck and every question waits for you. Then the solution, then one beat of proof, then the payoff, then a single call to action. Educational tone. Narrate the whole situation rather than labelling it.
LOOK
Editorial paper-collage, taped zine style. Palette only: deep forest green #003116, bright mint #93EFA4, warm cream #E7DFCC, black #171814. Every person is a black-and-white halftone photo cutout with a rough white torn-paper border and a small tape strip. Flat matte paper, visible print dots, hard drop shadows from an upper-left light. No glow, no gloss, no 3D.
FACES, AND THIS IS NOT NEGOTIABLE
Every face is a still printed paper cutout. Faces never move: no mouth movement, no talking, no blinking, no expression change, no morphing. When a person needs emphasis, the whole rigid head cutout rocks left and right a few degrees like a quick head shake, or nods. Nothing else.
SHOTS

YOU ARE THE BOTTLENECK. My head buried under a landslide of cream question cards, each printed with a black question mark, piling up to my eyes. A paper clock taped in the corner, hands ticking. Frantic.
THEY ASK MY SAUNA. Same head, now calm. The question cards fly away from me in a curved stream into a green card carrying a small sauna-bench glyph.
MY TASTE. MY CALLS. My head in strict side profile, the top opened like a lid, three tabbed cards sliding up out of it: TASTE, KNOWLEDGE, HOW I ACT.
WHAT THEY ASK ME. Three taped question cards on a pinboard, each with a small head clipped to its corner. Use the real questions: Dasol asking "did Filip build an app for the off-site?", Rafa asking "which positioning did you land on?", Rafa asking "how did you wire the ElevenLabs UI?". Each head rocks once as its card lands.
EVERY ANSWER HAS RECEIPTS. A paper till receipt printing line by line: READ RULES.md, CHECKED PostHog, FOLLOWED the relevant access rules. A mint chip reading SOURCES ATTACHED slaps on at the end. Do not print a time or any number I have not measured.
PERMISSIONS BY RELATIONSHIP. A four-row table. Robert and Rafa each get a cream ASK chip and a mint ACT chip. Sebastian and Dasol get ASK and a dashed empty slot where ACT would be. The point is that Sauna decides this from the relationship.
WHAT IT CAN TOUCH. The full loop, left to right, inside a dashed frame tagged GUARDRAILS. Robert's head, then a green card labelled with the sauna glyph that checks access, then a fan of memory cards labelled USER_PROFILE, RULES, COMPANY, then chips for PostHog, Gmail and Linear, then a dark card reading ACTION, gated by a chip reading ONLY IF ALLOWED. A mint dot travels the path, then returns along a single arrow underneath labelled ANSWER + SOURCES. One return arrow, one label, spelled correctly.
REJECTED. A question card reading "what does Filip think about me?" with a heavy NO stamp slamming onto it and a mint chip reading WRONG QUESTION. A paper loudspeaker shakes with jagged sound lines. Keep any alarm sound quiet.
NO. OR ESCALATE. A card reading "above its pay grade", a dashed mint arrow, and a mint iMessage-style bubble reading "escalated to Filip". Soft notification ding.
THERE WHILE NOT THERE. Me on a striped paper beach lounger under a parasol, sunglasses, holding a margarita. In the background, small heads sending question bubbles to a green sauna card and a mint answer coming back. Glass clink.
End card. Deep green, the word Sauna, and "try it for free at sauna.ai".

DO NOT INCLUDE
No reverse-direction section about me asking other people's Saunas. It dilutes the argument. No invented statistics or timings anywhere in the artwork.
DIAGRAMS
Do not generate shots 6 and 7 with a video model. The labels have to be exact. Draw them locally with Pillow and ffmpeg and cut them into the timeline.
MODEL
Seedance 2.5 for every generated shot. Pass each approved storyboard panel as the start_image, 1080p, generate_audio off.
AUDIO
Narrator, not my cloned voice. Warm and clear, explaining rather than selling. One narration file per shot, placed at an absolute offset, with a little room before and after each line. Music well underneath, ducked while anyone speaks. Nothing may be clipped at a scene boundary. Two sound effects only: the escalation ding and the beach clink. Verify by transcribing the finished soundtrack and reading it back.

Make a short explainer about my company using the Higgsfield video skill. Start with what you already know about the product and ask me for anything missing. Show me the script first, then one storyboard. Check the generation cost before making the film.