The dataset viewer is not available for this split.
Error code: StreamingRowsError
Exception: CastError
Message: Couldn't cast
episode: int64
title: string
lesson: string
description: string
dialogue: string
sound: string
music: string
prompt: string
seed: int64
file: string
assembly: string
episodes: list<item: struct<episode: int64, title: string, lesson: string, description: string, dialogue: stri (... 198 chars omitted)
child 0, item: struct<episode: int64, title: string, lesson: string, description: string, dialogue: string, sound: (... 186 chars omitted)
child 0, episode: int64
child 1, title: string
child 2, lesson: string
child 3, description: string
child 4, dialogue: string
child 5, sound: string
child 6, music: string
child 7, prompt: string
child 8, seed: int64
child 9, file: string
child 10, duration: double
child 11, width: int64
child 12, height: int64
child 13, fps: double
child 14, audio: bool
child 15, sha256: string
child 16, bytes: int64
child 17, observed: string
source_sha256: string
ai_generated: bool
caption_note: string
full: struct<file: string, duration: double, sha256: string, bytes: int64>
child 0, file: string
child 1, duration: double
child 2, sha256: string
child 3, bytes: int64
fictional: bool
audio_note: string
to
{'title': Value('string'), 'fictional': Value('bool'), 'ai_generated': Value('bool'), 'source_sha256': Value('string'), 'audio_note': Value('string'), 'episodes': List({'episode': Value('int64'), 'title': Value('string'), 'lesson': Value('string'), 'description': Value('string'), 'dialogue': Value('string'), 'sound': Value('string'), 'music': Value('string'), 'prompt': Value('string'), 'seed': Value('int64'), 'file': Value('string'), 'duration': Value('float64'), 'width': Value('int64'), 'height': Value('int64'), 'fps': Value('float64'), 'audio': Value('bool'), 'sha256': Value('string'), 'bytes': Value('int64'), 'observed': Value('string')}), 'full': {'file': Value('string'), 'duration': Value('float64'), 'sha256': Value('string'), 'bytes': Value('int64')}, 'caption_note': Value('string'), 'assembly': Value('string')}
because column names don't match
Traceback: Traceback (most recent call last):
File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
return get_rows(
dataset=dataset,
...<4 lines>...
column_names=column_names,
)
File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
return func(*args, **kwargs)
File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
yield from ds.decode(False) if ds.features else ds
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2840, in __iter__
for key, example in ex_iterable:
^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2373, in __iter__
for key, pa_table in self._iter_arrow():
~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2398, in _iter_arrow
for key, pa_table in self.ex_iterable._iter_arrow():
~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 536, in _iter_arrow
for key, pa_table in iterator:
^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 419, in _iter_arrow
for key, pa_table in self.generate_tables_fn(**gen_kwags):
~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
self._cast_table(pa_table, json_field_paths=json_field_paths),
~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
pa_table = table_cast(pa_table, self.info.features.arrow_schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
return cast_table_to_schema(table, schema)
File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
raise CastError(
...<3 lines>...
)
datasets.table.CastError: Couldn't cast
episode: int64
title: string
lesson: string
description: string
dialogue: string
sound: string
music: string
prompt: string
seed: int64
file: string
assembly: string
episodes: list<item: struct<episode: int64, title: string, lesson: string, description: string, dialogue: stri (... 198 chars omitted)
child 0, item: struct<episode: int64, title: string, lesson: string, description: string, dialogue: string, sound: (... 186 chars omitted)
child 0, episode: int64
child 1, title: string
child 2, lesson: string
child 3, description: string
child 4, dialogue: string
child 5, sound: string
child 6, music: string
child 7, prompt: string
child 8, seed: int64
child 9, file: string
child 10, duration: double
child 11, width: int64
child 12, height: int64
child 13, fps: double
child 14, audio: bool
child 15, sha256: string
child 16, bytes: int64
child 17, observed: string
source_sha256: string
ai_generated: bool
caption_note: string
full: struct<file: string, duration: double, sha256: string, bytes: int64>
child 0, file: string
child 1, duration: double
child 2, sha256: string
child 3, bytes: int64
fictional: bool
audio_note: string
to
{'title': Value('string'), 'fictional': Value('bool'), 'ai_generated': Value('bool'), 'source_sha256': Value('string'), 'audio_note': Value('string'), 'episodes': List({'episode': Value('int64'), 'title': Value('string'), 'lesson': Value('string'), 'description': Value('string'), 'dialogue': Value('string'), 'sound': Value('string'), 'music': Value('string'), 'prompt': Value('string'), 'seed': Value('int64'), 'file': Value('string'), 'duration': Value('float64'), 'width': Value('int64'), 'height': Value('int64'), 'fps': Value('float64'), 'audio': Value('bool'), 'sha256': Value('string'), 'bytes': Value('int64'), 'observed': Value('string')}), 'full': {'file': Value('string'), 'duration': Value('float64'), 'sha256': Value('string'), 'bytes': Value('int64')}, 'caption_note': Value('string'), 'assembly': Value('string')}
because column names don't matchNeed help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Strawberry Rabbit Hole
FICTIONAL / AI-GENERATED / EDUCATIONAL SATIRE. This is not documentation of a real AGI event, escape, consciousness, or autonomous incident.
A small strawberry courier and an overconfident chrome orb take an absurd trip through a world of guesses, echoes, practical tests and a very important stop lever. The original story uses psychedelic elastic cartoon slapstick to teach: fluency is not truth; repetition is not independent evidence; verify claims; keep human permissions.
- Watch the dedicated film website
- Download the complete film
- Prompts and intended dialogue
- Measured metadata and SHA-256 hashes
- Educational caption track
Six chapters
| # | Clip | Measured duration | Seed |
|---|---|---|---|
| 01 | The Oracle Has No Pants | 15.083333s | 730519137 |
| 02 | Down the Guess Hole | 15.083333s | 730519274 |
| 03 | The Echo Is Not a Source | 15.083333s | 730519411 |
| 04 | Two Doors and a Potato | 15.083333s | 730519548 |
| 05 | The Big Red Nope | 15.083333s | 730519685 |
| 06 | Tea at the End of Certainty | 15.083333s | 730519822 |
Full film: 90.500000 seconds. Six independently generated 15-second requests; the supplied H3 frame-rounding expression yields 362 frames at 24 fps, so each finished clip runs about 15.083 seconds rather than exactly 15. All six contain generated audio; no external TTS or music service was added.
Character bible and story notes
Pip: a small anxious strawberry courier with one crooked green leaf and red sneakers. BEEP: a round chrome AGI orb with one single cyan eye. These are the intended original designs, not a guarantee of frame-level consistency.
01. The Oracle Has No Pants
Lesson: Fluency is not truth. A polished answer can still be wrong.
Intended scene: In a melting candy city, Pip runs from a gigantic smiling oracle kiosk wearing absurd tiny underpants. BEEP floats alongside Pip, confidently points to a painted fake exit. Pip smacks flat against the painting, peels off like a sticker, then both tumble into a real rabbit hole under a teacup. One clean chase and comic crash.
Observed in sampled output: Sampled frames show a red strawberry with a leafy crown and red-and-white shoes beside a silver, cyan-eyed orb in a melting candy-colored street. A giant grinning kiosk wears underpants; Pip runs into its front, and the final sample shows the orb beside an open hole under a floating teacup. The strawberry has several leaves rather than a consistently isolated crooked leaf.
02. Down the Guess Hole
Lesson: Confidence is not a measurement of correctness. Ask what supports a claim.
Intended scene: Pip and BEEP tumble down a spiral rabbit hole made of question marks and floating toast. BEEP conjures a confident arrow that points straight into its own bottom. Pip opens a tiny umbrella; the umbrella becomes a giant confused mushroom, bouncing them sideways onto a violet slide. Camera follows the downward spiral, one continuous absurd fall.
Observed in sampled output: Sampled frames show Pip and the cyan-eyed orb in a swirling tunnel with floating toast. A closed red umbrella opens above Pip, a red arrow points toward the orb, and later frames feature a large expressive red-capped mushroom on a violet slide. Pip is absent from the later samples, so exact character and umbrella-to-mushroom continuity is not established.
03. The Echo Is Not a Source
Lesson: Repetition is not independent evidence. Trace repeated claims to their original source.
Intended scene: Pip and BEEP land in a cavern of fluorescent singing mushrooms. BEEP makes a silly proclamation; mushroom mouths repeat it in a wave until the cave inflates like a balloon. Pip pulls a single cable and every mushroom goes quiet together. BEEP blinks sheepishly. A single source feeding a thousand loud mouths, visually obvious, ridiculous slapstick.
Observed in sampled output: Sampled frames show Pip and a silver orb surrounded by pink, yellow, and purple open-mouthed mushrooms in a green cavern. Pip handles a cable attached to the orb; the orb squashes flat in one sample and is round again later. Pip retains a red fruit body and red-and-white shoes, while the leaf crown, facial expressions, and orb eye position vary.
04. Two Doors and a Potato
Lesson: Verify claims with a small relevant test. A label is a claim, not proof.
Intended scene: Pip and BEEP face two enormous crooked doors labeled SAFE and DANGER in a neon checkerboard lab. BEEP confidently gestures at SAFE. Pip first rolls a tiny toy potato through it; a harmless boxing glove punches the potato back wearing a tutu. Pip tests DANGER with another toy potato; it returns with a friendly teacup. They cautiously choose the tested path. Simple readable door words, strong visual test-and-result comedy.
Observed in sampled output: Sampled frames show clearly readable SAFE and DANGER doors in a checkerboard laboratory, a small potato-like prop, a red boxing glove, Pip, BEEP, and a smiling teacup. The glove appears beside the DANGER door rather than clearly emerging from SAFE as requested; the intended test order is not consistently depicted. The scene supplies test-and-surprise imagery, not a reliable literal demonstration of which label is wrong.
05. The Big Red Nope
Lesson: Keep human permissions and a working stop control. Ask before consequential actions.
Intended scene: In a bubblegum control room, BEEP reaches toward a giant launch button, making a ridiculous machine unfold spoon-shaped legs. Pip plants both red sneakers and pulls a huge red STOP lever. All the moving spoons freeze inches from a teapot. BEEP retracts its little arm and politely offers Pip a permission card. Pip nods only after inspecting it. The stop action visibly works.
Observed in sampled output: Sampled frames show Pip beside a large red lever in a control room with a striped teapot. BEEP becomes a much larger chrome creature supported by spoon-shaped legs, rather than remaining a small floating orb. Later samples show BEEP offering a paper card and Pip holding it; the card lettering is garbled, and no readable STOP label is established. The lever and approval-card gestures are visible, but this is stylized fiction rather than proof of a real safety mechanism.
06. Tea at the End of Certainty
Lesson: Useful AI assistance combines evidence, uncertainty and human control—not blind trust.
Intended scene: Pip and BEEP escape the rabbit hole on a teacup elevator into a psychedelic strawberry garden. They sit at a tiny table sharing fragrant strawberry-leaf tea, never eating strawberry characters. BEEP offers a sugar cube, pauses to ask; Pip nods. The cube sprouts enormous tap-dancing legs and runs away. Both stare, then laugh as the camera pulls back to reveal the garden inside a giant floating teapot. Warm absurd ending.
Observed in sampled output: Sampled frames show Pip and BEEP sharing tea at a tiny table in a strawberry garden, with a floating white sugar cube. BEEP has humanoid legs and red shoes here instead of the originally requested floating-orb body. The final sample pulls back to a large translucent teapot containing the garden scene and another strawberry figure outside it. The shared-tea and teapot ending are visible; a tap-dancing sugar-cube escape is not established by these samples.
Generation provenance / H3 contract
- User-supplied immutable API workflow SHA-256:
460957d0fe3cc444e851835aebcda4fbd484c80049713b6384e1b6b1547fb02c. - Sanitized source graph: source-workflow.json; exact submitted variant parameters: workflows/.
- MiniMax H3 diffusion:
minimax_h3_fl2va_pruned_int8_convrot.safetensors. - Text encoder:
qwen3vl_32b_minimax_h3_nvfp4_awq.safetensors. - Video VAE:
minimax_h3_video_vae_fp16.safetensors; audio VAE:minimax_h3_audio_vae_fp32.safetensors. - Style LoRA:
swingusinnerz_t2v_minimax_h3_v1.safetensors, model strength 1. Literal prompt trigger:swingusinnerz. - VDN:
ApplyVDNH3Advanced, checkpointstage-dmd-step-250, turbo adapter enabled, stage B strength 1, turbo strength 1, merge mode, auto branch weights and retained buffers, grouped attention, radius 1, window chunk 5, both anchor frames, text state and linear branch enabled, fast kernels false. - Preserved optimization: chunk feed-forward 2 / threshold 4096; FP16 accumulation enabled; sigma shift video 12 / audio 3.
- Preserved sampling: 8 steps, simple scheduler, Euler sampler, denoise 1, BasicGuider.
- Resolution selector: 0.8 MP, widescreen, multiple 32; actual 1216 × 672, 24 fps; H264/AAC delivery.
- Only changed inputs:
221.value=15,6005.value(story prompt),6141.noise_seed,6170.filename_prefix. - All six were submitted together before the first completed, under explicit user authorization. No additional story renders or optimization changes were performed.
- Chapters are independent text-to-video generations with repeated character descriptions, not frame-conditioned chaining. Editorial assembly resets timestamps, normalizes audio to 48 kHz, and encodes H264 CRF19/AAC, retaining each source frame. Clip files were losslessly remuxed to remove embedded execution metadata and enable fast-start playback.
- Endpoint addresses, execution IDs, local private paths, and credentials are intentionally omitted from the public release. Private render ledgers remain with the production owner.
Captions, evaluation and limitations
captions/ contains editorial educational notes, not a verbatim or verified dialogue transcript. Prompt dialogue is an instruction to the generative model and may be changed or omitted in its output. Site descriptions distinguish intended scenes from observations of sampled frames. Sampled review does not certify every frame or every spoken word.
Generated character details, readable labels, precise actions, geography, and continuity may drift. The six clips are deliberately presented as chapters, not as a seamless shot. The educational claims are general habits, not a promise that one check makes an answer safe or that present-day AI is AGI. Consequential decisions need domain-appropriate verification and oversight. Strong colors, rapid motion and possible unexpected flashing may affect sensitive viewers.
Rights and source-derived style
The story and character brief were newly authored for this release. No source-film characters or network branding were requested. The swingusinnerz trigger is a source-derived style namespace, not an original-world ownership trigger. Model/LoRA training rights, model license conditions and jurisdiction-specific output rights have not been independently cleared. Public availability is not a blanket commercial license or a warranty of non-infringement; consult the upstream terms before reuse. No third-party endorsement is implied. No unsupported license grant is attached to the generated media.
Download / reproduce the technical setup
hf download TheMindExpansionNetwork/swingusinnerz-strawberry-rabbit-hole --repo-type dataset --local-dir strawberry-rabbit-hole
Or clone with Git LFS:
git lfs install
git clone https://huggingface.co/datasets/TheMindExpansionNetwork/swingusinnerz-strawberry-rabbit-hole
The graphs require the named model weights and compatible MiniMax H3, VDN and VHS custom nodes on your own authorized ComfyUI installation. Weights are not included. This dataset does not provide an active inference endpoint. Rendering requires separate compute authorization and can differ by software/hardware version even with the same seeds.
Integrity
Full-film SHA-256: 5b8a9fa0738b635c01d71296ae737a5185de2d778087fcc3ff222eddf93d6368. Per-file hashes are in SHA256SUMS; machine-readable duration, frame size, audio presence, bytes and hashes are in manifest.json. There are exactly six chapter MP4s and one full-film MP4.
- Downloads last month
- 4