Techhalla publishes prompts behind the roughly three-minute film “They Already Saw It”

• Techhalla published production prompts for “They Already Saw It,” made with Midjourney, H3, Suno and Claude, as credited in the release post.
• Beat timing is written into the video generation: the H3 prompt supplies a song excerpt and assigns repeated gestures to its pulse.
• The soundtrack combines ritualistic darkwave at 126 BPM with raspy rap, whispers and filtered chants, according to techhalla's Suno recipe.

A cassette-fed puppet deity gets its own image prompt. The video breakdown specifies a hard cut at exactly 00:03.810, using the image-and-audio references supported by H3's reference-generation mode.

“They Already Saw It” runs about 3:05. Claude appears in the tool credits, but the shared replies do not specify its contribution.

They Already Saw It

Techhalla shared this complete image prompt, including the settings suffix:

Dozens of hands feeding long ribbons of brown magnetic tape from cassettes into the glowing single eye of a puppet deity on an altar of speakers, red candlelight, grainy Hi8 camcorder still --ar 16:9 --raw --profile 8wo3daj --stylize 500 --hd

The recipe carries a personalization dependency alongside its scene description: --profile 8wo3daj. Midjourney's documentation explains that personalization profiles apply aesthetic preferences learned from image selections and likes.

The published song recipe separates musical style from lyrics, matching the separate inputs available in Suno's Custom mode.

• Tempo and rhythm: 126 BPM, with a sparse, broken pulse.
• Production: heavy drones, low distorted bass, sparse drums, clipped drum-machine transients and a restrained dark mix.
• Voices: low, raw, raspy male rap; hushed close-mic whispers; filtered chants; soft chorus harmony pads.
• Arrangement: verse, chorus, second verse, repeated chorus, spoken-word bridge, then a fading outro with distant chanting.

The lyrics carry the image props into the song: “Speakers for an altar, cassettes for the last dream.” Section tags also direct delivery, including [Chorus | Low Energy | Chant vocals | Filtered Vocals].

The H3 example specifies a ten-second sequence built from three reference assets:

1. _Image1: hooded ravers kneeling in shallow rice-field water under a red moon.
2. _Image2: a line of acolytes holding smoking incense burners.
3. _Audio1: the song excerpt, identified as the full soundtrack to reuse unchanged.

At 126 BPM, each beat lasts approximately 0.476 seconds. The requested cut falls after eight beat intervals:

• 00:00–00:03.810: the kneeling ravers slap both palms onto the water together on every beat, then lift them back up.
• 00:03.810–00:10.000: a hard cut introduces the acolytes, who jerk their incense burners upward together on every beat, then let them drop.

Both shots use locked-off cameras, with movement already underway in the first frame. The beat-by-beat choreography is the gem here.

The two-shot H3 example accompanying the prompt

The prompt separately requests water slaps, rattling chains, rustling robes and rhythmic breathing. Its preservation instructions cover the people, costumes, setting and Hi8 look from both reference images, alongside the unchanged song excerpt.

Techhalla also published a roughly 2:30, 1970s-styled performance video titled “Rock is dead (1973, full version).” An accompanying workflow reply links to an H3 music-video process and explicitly names Hailuo AI.