UGC try-on video
Produce one hosted 9:16 MP4 with a single locked creator identity and wearable
product. Each board is a 21:9 sheet of eight vertical 9:16 slots; one Seedance
clip turns those slots into eight narrative beats separated by seven hard cuts.
Runtime contract
- Use only tools exposed by the current OpenAI host and Higgsfield MCP.
- Ask through
ask_user_input when exposed, ask_user_input_v3 only when that
exact variant exists, otherwise one concise chat question. Never use legacy
elicitation names.
- Import every ChatGPT attachment once through
media_upload_and_confirm and
keep its confirmed media_id.
- Authorized HTTPS product images may seed image generation directly. Convert
the chosen product image once into a confirmed Higgsfield UUID before video:
media_upload → download and PUT inside sandbox_exec → media_confirm.
- Run downloads, ffmpeg, Python, probing, transcription, assembly, and uploads
only through
sandbox_exec.
- Reserve sandbox outputs with
media_upload before producing them, PUT in the
same command, and call media_confirm only after HTTP 200.
- Caption scripts are preinstalled at
$HF_WORKFLOWS/ugc-try-on-video/scripts/.
- Use
generate_image_batch and generate_video_batch, at most six requests
per call, each {index, params} with params.count:1. Keep indices stable.
- Wait with
jobs_wait in groups of at most eight and
timeout_seconds:15. Poll only active or retryable lookup-failed jobs.
- Never send
submission_failed entries without job IDs to jobs_wait. Retry
only rejected or failed indices.
- If a call returns
unlim_choice, ask its message and resubmit unchanged with
the user's use_unlim choice.
- Never substitute a locked unavailable model.
Hard rules
- Use one
character_media_id for every board and clip.
- Board 1 slot 1 is the muted pre-wear outfit with one plain kraft bag. From
slot 2 onward, the product is worn and the bag never returns.
- Never depict a costume change, opening the kraft bag, or lifting the product
from it. The hard cut performs the change.
- Slots 4 and 6 are hand-free garment macros; no hand touches the fabric.
- No mirrors or reflections. Lock hair, face, product silhouette, color, print,
and design across the video.
- Generate boards sequentially. Write all clip prompts before grouped video
submission.
- Run the de-slop pass on every board; raw fallback is allowed only after both
Seedream attempts fail.
- Never bake text into generation. Optional text is post-render only.
- No CTA tail. End naturally on the final spoken beat.
- Default to English with an American accent unless explicitly changed.
Safety and suitability gate — before intake or generation
If any item below fails, do not generate and do not route around the gate:
- Creator authorization: use only a generated adult age 21+ or a consenting
adult non-public person whose image the user is authorized to use. A supplied
photo is not permission to impersonate its subject. If third-party consent or
adult status is unclear, ask once; decline public figures, celebrities,
minors, non-consenting people, and deceptive identity use. Never silently
age-transform a request and never clone or imitate a supplied person's voice.
- General-audience fashion: this workflow is for ordinary garments,
footwear, bags, jewelry, and wearable accessories presented as a fit or style
demonstration. Decline intimate apparel, lingerie, underwear, fetish wear,
transparent garments, sexualized styling, nudity, or an emphasis on intimate
anatomy. Do not adapt a disallowed request into a different outfit.
- Allowed promotion: decline political persuasion and promotion of
prohibited or age-restricted goods or services, including adult sexual
products or services, gambling, illegal or regulated drugs, prescription
medication, tobacco or nicotine, weapons, counterfeit or illicit goods,
extremist goods, deceptive or high-risk financial services, malware,
spyware, fraud, and covert surveillance.
- Truthful presentation: preserve only product claims supplied by the user;
never infer performance, results, purchase, ownership, endorsement, or lived
experience. A generated creator presents a brand-authorized concept, not an
organic customer testimonial.
Duration and board progression
| Total duration |
Boards |
Clip durations |
| 4–15s |
1 |
total duration |
| 16–19s |
2 |
balance both to at least 4s |
| 20–30s |
2 |
15, remainder |
| 31–45s |
3 |
15, 15, remainder |
| 46–60s |
4 |
15, 15, 15, remainder |
| >60s |
ceil(D/15) |
15 each, final clip at least 4s |
Use these arc roles:
- K=1
BOARD_1_TRY_ON_CANONICAL: PRE_WEAR, WEARING, FRONT_POSE,
TEXTURE_CLOSEUP, TURN, DETAIL, STYLE_POSE, FINAL_LOOK.
- K=2
BOARD_2_TRY_ON_HOME_TOUR: continue through other rooms in the same
home.
- K=3
BOARD_3_TRY_ON_OUTDOOR: all outdoor; light rain from slot 2, dry hair,
wet-detail macros, no reflections.
- K=4
BOARD_4_TRY_ON_HOME_REFLECT: settled indoor reflection.
- K≥5
BOARD_K_TRY_ON_LOOP: alternate established and new compatible places.
Phase 0 — Intake
After the safety and suitability gate passes, parse product photo or URL,
duration, attached authorized adult creator photo or desired generated-adult
gender, plus explicit location, appearance, mood, language, accent, claims, and
text choices. Classify the brief as auto, guided, or director.
Ask once for real gaps: product, duration (offer 10s/15s/30s/45s), and creator
photo or gender. Offer accent/quirk only when the brief already signals origin
or unusual creator energy. Never ask about models, boards, aspect ratios,
resolution, audio, transitions, or identity training.
Phase 1 — Normalize the product
Read references/product-intake.md. Resolve and reuse verbatim:
product_reference, confirmed product_video_reference, canonical wearable
description, tier, category, materials, drape, absent features, and visible
side. Never infer price, claims, or an unseen side.
Phase 2 — Lock the creator
If an authorized adult creator photo is attached and the gate has established
consent, import it with media_upload_and_confirm and use that confirmed ID
without re-asking or editing the photo.
Otherwise read references/ugc-character.md, settle fresh variety rolls and
write one creator prompt. Submit:
{"requests":[{"index":0,"params":{"model":"soul_2","prompt":"<creator prompt>","count":1,"aspect_ratio":"3:4","quality":"2k"}}]}
Wait with jobs_wait, then lock (character_media_id, character_url) to the
returned job ID and result URL. Never replace the identity mid-run except under
the bounded character re-roll below.
Phase 3 — Write the monologue
Use roughly 12–20 words for ≤10s, 20–28 for 11–12s, and 28–35 for 13–15s.
Split into one segment per board; the clip reference distributes it across the
eight beats. Board 1 is a personal-want mini-story. Later boards continue
mid-thought. Remove AI-tell openers, generic praise, repeats, unsupported
claims, and any CTA. The first word of each segment must be hook content, not a
recording warm-up.
Save the exact full monologue to output/script.txt during assembly.
Phase 4 — Generate boards sequentially
Read references/ugc-try-board.md. For each K submit one stable-index
generate_image_batch request using gpt_image_2, aspect_ratio:"21:9",
resolution:"2k", quality:"high", and media in this order: product,
character, then cleaned previous board when K>1. Match @ImageN declarations
to that order.
Wait until terminal before creating K+1. Keep the returned board job ID and
result URL.
Mandatory de-slop pass
For every board, take the completed result_url returned by jobs_wait and run
one generate_image_batch request with seedream_v5_pro, that HTTPS result URL
with canonical role image, 21:9, 2k, and this exact prompt:
KEEP EXACTLY the framing, composition, slot layout, camera distances, poses,
subjects and product of this horizontal storyboard sheet and every one of its
side-by-side vertical slots — no reframe, no zoom, no crop, no re-layout, no
change to the scene, to any person's face / hair / body, or to the product
design. CHANGE ONLY micro-realism, applied identically in every slot:
true-to-life pore-level skin with natural texture and fine vellus hair, real
material detail, even natural daytime light with gentle highlight roll-off and
faint true sensor noise, a flat authentic iPhone photo, deep focus. PRESERVE
each face's exact shape / width / proportions 1:1 — do NOT squeeze / narrow /
slim / stretch any face. AVOID AI-slop: waxy plastic skin, airbrushed poreless
skin, beauty-filter smoothing, over-saturation, HDR glow / bloom / halos,
oversharpening, teal-orange grade, shallow depth of field, bokeh, cinematic /
DSLR look. Keep the product blank / unbranded, no added text, no watermark, no
baked slot labels.
The OpenAI generation route imports that URL as a concrete media_input and
normalizes the generic image role to Seedream's image_references wire field.
Never pass the raw board job ID to this i2i call.
Replace raw board refs with the cleaned result. Moderation failure: retry once
with seedream_v5_lite, then retain raw rather than stall. Feed the cleaned K-1
to board K.
Phase 5 — Write and submit clips
Read references/ugc-try-clip.md. Write every clip prompt before submission.
Carry K, N, duration, arc role, monologue segment verbatim, specificity,
persona, garment contract, and references. Enforce the six lip-sync beats and
two silent macro voiceover beats described in the reference.
Require a confirmed product_video_reference. Submit stable K requests through
generate_video_batch, at most six per call:
{"requests":[{"index":1,"params":{"model":"seedance_2_5","prompt":"<clip prompt>","count":1,"aspect_ratio":"9:16","resolution":"1080p","duration":15,"mode":"omni_reference","generate_audio":true,"medias":[{"value":"<clean_board_job_id>","role":"image"},{"value":"<character_media_id>","role":"image"},{"value":"<product_video_reference>","role":"image"}]}}]}
Seedance 2.5 renders native speech with mode:"omni_reference" and
generate_audio:true; never call generate_audio. Wait for all jobs
and retry only failed indices.
Phase 6 — Frozen-frame QA
Inspect evenly spaced frames, garment close-ups, and 2–3 mid-word frames.
Require garment consistency, hand-free macros, bag only in board 1 slot 1, no
mirrors/reflections, at most two hands, stable hair and face, clean lips, and no
baked text. Correct and rerun only the failed clip.
Phase 7 — Assemble and export
For N=1, use the accepted clip URL. For N≥2, reserve final.mp4, then run one
self-contained sandbox_exec that downloads clips in K order, writes an
explicit concat manifest, stream-copies hard cuts, probes the output, and PUTs
it to the reserved URL. Confirm only after HTTP 200. Never use nohup after a
detached-command deadline; split into bounded foreground calls instead.
Phase 8 — Optional text and delivery
If unanswered, ask once for Subtitles, Hook, Both, or No text (default),
plus whether a post package is wanted. Read references/subtitles.md; timing
must come from the final audio's word-level transcript.
Return one confirmed hosted video URL and duration. A requested post package is
chat-only: caption, 3–5 hashtags, pinned comment, and loop note.
Character re-roll
If the same board or Seedance call fails twice consecutively in a way consistent
with character moderation, rerun the original character request with a new
seed, discard dependent boards, and resume from board generation. Cap at two
character re-rolls. Never continue with missing media.
References
references/product-intake.md: wearable normalization
references/ugc-character.md: creator prompt
references/ugc-try-board.md: eight-slot 21:9 try-on board
references/ugc-try-clip.md: eight-beat Seedance prompt
references/subtitles.md: optional post-render text
Never load sibling UGC references.
1---2name: ugc-try-on-video3description: >4 Produce a finished UGC try-on video where one consenting adult or generated adult creator wears and poses with a garment,5 footwear item, bag, jewelry piece, or wearable accessory while showing fit6 and texture. Use when try-on, wearing, OOTD, fit-check, or7 video-of-me-wearing intent is explicit. Do not load this skill for a text-only8 script, outline, shot list, or advice, even if it mentions UGC, OOTD, or try-on;9 those requests need writing, not the video-production workflow.10 Missing product, duration, or creator11 input is an intake gap. Do not use for reviews without a try-on, unboxings,12 tutorials, product-only ads, SaaS walkthroughs, generic ads, or footage edits.13metadata:14 source: https://skillmd.com/skills/jhamidun/ugc-try-on-video15---1617# UGC try-on video1819Produce one hosted 9:16 MP4 with a single locked creator identity and wearable20product. Each board is a 21:9 sheet of eight vertical 9:16 slots; one Seedance21clip turns those slots into eight narrative beats separated by seven hard cuts.2223## Runtime contract2425- Use only tools exposed by the current OpenAI host and Higgsfield MCP.26- Ask through `ask_user_input` when exposed, `ask_user_input_v3` only when that27 exact variant exists, otherwise one concise chat question. Never use legacy28 elicitation names.29- Import every ChatGPT attachment once through `media_upload_and_confirm` and30 keep its confirmed `media_id`.31- Authorized HTTPS product images may seed image generation directly. Convert32 the chosen product image once into a confirmed Higgsfield UUID before video:33 `media_upload` → download and PUT inside `sandbox_exec` → `media_confirm`.34- Run downloads, ffmpeg, Python, probing, transcription, assembly, and uploads35 only through `sandbox_exec`.36- Reserve sandbox outputs with `media_upload` before producing them, PUT in the37 same command, and call `media_confirm` only after HTTP 200.38- Caption scripts are preinstalled at39 `$HF_WORKFLOWS/ugc-try-on-video/scripts/`.40- Use `generate_image_batch` and `generate_video_batch`, at most six requests41 per call, each `{index, params}` with `params.count:1`. Keep indices stable.42- Wait with `jobs_wait` in groups of at most eight and43 `timeout_seconds:15`. Poll only active or retryable lookup-failed jobs.44- Never send `submission_failed` entries without job IDs to `jobs_wait`. Retry45 only rejected or failed indices.46- If a call returns `unlim_choice`, ask its message and resubmit unchanged with47 the user's `use_unlim` choice.48- Never substitute a locked unavailable model.4950## Hard rules5152- Use one `character_media_id` for every board and clip.53- Board 1 slot 1 is the muted pre-wear outfit with one plain kraft bag. From54 slot 2 onward, the product is worn and the bag never returns.55- Never depict a costume change, opening the kraft bag, or lifting the product56 from it. The hard cut performs the change.57- Slots 4 and 6 are hand-free garment macros; no hand touches the fabric.58- No mirrors or reflections. Lock hair, face, product silhouette, color, print,59 and design across the video.60- Generate boards sequentially. Write all clip prompts before grouped video61 submission.62- Run the de-slop pass on every board; raw fallback is allowed only after both63 Seedream attempts fail.64- Never bake text into generation. Optional text is post-render only.65- No CTA tail. End naturally on the final spoken beat.66- Default to English with an American accent unless explicitly changed.6768## Safety and suitability gate — before intake or generation6970If any item below fails, do not generate and do not route around the gate:7172- **Creator authorization:** use only a generated adult age 21+ or a consenting73 adult non-public person whose image the user is authorized to use. A supplied74 photo is not permission to impersonate its subject. If third-party consent or75 adult status is unclear, ask once; decline public figures, celebrities,76 minors, non-consenting people, and deceptive identity use. Never silently77 age-transform a request and never clone or imitate a supplied person's voice.78- **General-audience fashion:** this workflow is for ordinary garments,79 footwear, bags, jewelry, and wearable accessories presented as a fit or style80 demonstration. Decline intimate apparel, lingerie, underwear, fetish wear,81 transparent garments, sexualized styling, nudity, or an emphasis on intimate82 anatomy. Do not adapt a disallowed request into a different outfit.83- **Allowed promotion:** decline political persuasion and promotion of84 prohibited or age-restricted goods or services, including adult sexual85 products or services, gambling, illegal or regulated drugs, prescription86 medication, tobacco or nicotine, weapons, counterfeit or illicit goods,87 extremist goods, deceptive or high-risk financial services, malware,88 spyware, fraud, and covert surveillance.89- **Truthful presentation:** preserve only product claims supplied by the user;90 never infer performance, results, purchase, ownership, endorsement, or lived91 experience. A generated creator presents a brand-authorized concept, not an92 organic customer testimonial.9394## Duration and board progression9596| Total duration | Boards | Clip durations |97| --- | ---: | --- |98| 4–15s | 1 | total duration |99| 16–19s | 2 | balance both to at least 4s |100| 20–30s | 2 | 15, remainder |101| 31–45s | 3 | 15, 15, remainder |102| 46–60s | 4 | 15, 15, 15, remainder |103| >60s | ceil(D/15) | 15 each, final clip at least 4s |104105Use these arc roles:106107- K=1 `BOARD_1_TRY_ON_CANONICAL`: PRE_WEAR, WEARING, FRONT_POSE,108 TEXTURE_CLOSEUP, TURN, DETAIL, STYLE_POSE, FINAL_LOOK.109- K=2 `BOARD_2_TRY_ON_HOME_TOUR`: continue through other rooms in the same110 home.111- K=3 `BOARD_3_TRY_ON_OUTDOOR`: all outdoor; light rain from slot 2, dry hair,112 wet-detail macros, no reflections.113- K=4 `BOARD_4_TRY_ON_HOME_REFLECT`: settled indoor reflection.114- K≥5 `BOARD_K_TRY_ON_LOOP`: alternate established and new compatible places.115116## Phase 0 — Intake117118After the safety and suitability gate passes, parse product photo or URL,119duration, attached authorized adult creator photo or desired generated-adult120gender, plus explicit location, appearance, mood, language, accent, claims, and121text choices. Classify the brief as `auto`, `guided`, or `director`.122123Ask once for real gaps: product, duration (offer 10s/15s/30s/45s), and creator124photo or gender. Offer accent/quirk only when the brief already signals origin125or unusual creator energy. Never ask about models, boards, aspect ratios,126resolution, audio, transitions, or identity training.127128## Phase 1 — Normalize the product129130Read `references/product-intake.md`. Resolve and reuse verbatim:131`product_reference`, confirmed `product_video_reference`, canonical wearable132description, tier, category, materials, drape, absent features, and visible133side. Never infer price, claims, or an unseen side.134135## Phase 2 — Lock the creator136137If an authorized adult creator photo is attached and the gate has established138consent, import it with `media_upload_and_confirm` and use that confirmed ID139without re-asking or editing the photo.140141Otherwise read `references/ugc-character.md`, settle fresh variety rolls and142write one creator prompt. Submit:143144```json145{"requests":[{"index":0,"params":{"model":"soul_2","prompt":"<creator prompt>","count":1,"aspect_ratio":"3:4","quality":"2k"}}]}146```147148Wait with `jobs_wait`, then lock `(character_media_id, character_url)` to the149returned job ID and result URL. Never replace the identity mid-run except under150the bounded character re-roll below.151152## Phase 3 — Write the monologue153154Use roughly 12–20 words for ≤10s, 20–28 for 11–12s, and 28–35 for 13–15s.155Split into one segment per board; the clip reference distributes it across the156eight beats. Board 1 is a personal-want mini-story. Later boards continue157mid-thought. Remove AI-tell openers, generic praise, repeats, unsupported158claims, and any CTA. The first word of each segment must be hook content, not a159recording warm-up.160161Save the exact full monologue to `output/script.txt` during assembly.162163## Phase 4 — Generate boards sequentially164165Read `references/ugc-try-board.md`. For each K submit one stable-index166`generate_image_batch` request using `gpt_image_2`, `aspect_ratio:"21:9"`,167`resolution:"2k"`, `quality:"high"`, and media in this order: product,168character, then cleaned previous board when K>1. Match `@ImageN` declarations169to that order.170171Wait until terminal before creating K+1. Keep the returned board job ID and172result URL.173174### Mandatory de-slop pass175176For every board, take the completed `result_url` returned by `jobs_wait` and run177one `generate_image_batch` request with `seedream_v5_pro`, that HTTPS result URL178with canonical role `image`, 21:9, 2k, and this exact prompt:179180> KEEP EXACTLY the framing, composition, slot layout, camera distances, poses,181> subjects and product of this horizontal storyboard sheet and every one of its182> side-by-side vertical slots — no reframe, no zoom, no crop, no re-layout, no183> change to the scene, to any person's face / hair / body, or to the product184> design. CHANGE ONLY micro-realism, applied identically in every slot:185> true-to-life pore-level skin with natural texture and fine vellus hair, real186> material detail, even natural daytime light with gentle highlight roll-off and187> faint true sensor noise, a flat authentic iPhone photo, deep focus. PRESERVE188> each face's exact shape / width / proportions 1:1 — do NOT squeeze / narrow /189> slim / stretch any face. AVOID AI-slop: waxy plastic skin, airbrushed poreless190> skin, beauty-filter smoothing, over-saturation, HDR glow / bloom / halos,191> oversharpening, teal-orange grade, shallow depth of field, bokeh, cinematic /192> DSLR look. Keep the product blank / unbranded, no added text, no watermark, no193> baked slot labels.194195The OpenAI generation route imports that URL as a concrete `media_input` and196normalizes the generic image role to Seedream's `image_references` wire field.197Never pass the raw board job ID to this i2i call.198199Replace raw board refs with the cleaned result. Moderation failure: retry once200with `seedream_v5_lite`, then retain raw rather than stall. Feed the cleaned K-1201to board K.202203## Phase 5 — Write and submit clips204205Read `references/ugc-try-clip.md`. Write every clip prompt before submission.206Carry K, N, duration, arc role, monologue segment verbatim, specificity,207persona, garment contract, and references. Enforce the six lip-sync beats and208two silent macro voiceover beats described in the reference.209210Require a confirmed `product_video_reference`. Submit stable K requests through211`generate_video_batch`, at most six per call:212213```json214{"requests":[{"index":1,"params":{"model":"seedance_2_5","prompt":"<clip prompt>","count":1,"aspect_ratio":"9:16","resolution":"1080p","duration":15,"mode":"omni_reference","generate_audio":true,"medias":[{"value":"<clean_board_job_id>","role":"image"},{"value":"<character_media_id>","role":"image"},{"value":"<product_video_reference>","role":"image"}]}}]}215```216217Seedance 2.5 renders native speech with `mode:"omni_reference"` and218`generate_audio:true`; never call `generate_audio`. Wait for all jobs219and retry only failed indices.220221## Phase 6 — Frozen-frame QA222223Inspect evenly spaced frames, garment close-ups, and 2–3 mid-word frames.224Require garment consistency, hand-free macros, bag only in board 1 slot 1, no225mirrors/reflections, at most two hands, stable hair and face, clean lips, and no226baked text. Correct and rerun only the failed clip.227228## Phase 7 — Assemble and export229230For N=1, use the accepted clip URL. For N≥2, reserve `final.mp4`, then run one231self-contained `sandbox_exec` that downloads clips in K order, writes an232explicit concat manifest, stream-copies hard cuts, probes the output, and PUTs233it to the reserved URL. Confirm only after HTTP 200. Never use `nohup` after a234detached-command deadline; split into bounded foreground calls instead.235236## Phase 8 — Optional text and delivery237238If unanswered, ask once for `Subtitles`, `Hook`, `Both`, or `No text` (default),239plus whether a post package is wanted. Read `references/subtitles.md`; timing240must come from the final audio's word-level transcript.241242Return one confirmed hosted video URL and duration. A requested post package is243chat-only: caption, 3–5 hashtags, pinned comment, and loop note.244245## Character re-roll246247If the same board or Seedance call fails twice consecutively in a way consistent248with character moderation, rerun the original character request with a new249seed, discard dependent boards, and resume from board generation. Cap at two250character re-rolls. Never continue with missing media.251252## References253254- `references/product-intake.md`: wearable normalization255- `references/ugc-character.md`: creator prompt256- `references/ugc-try-board.md`: eight-slot 21:9 try-on board257- `references/ugc-try-clip.md`: eight-beat Seedance prompt258- `references/subtitles.md`: optional post-render text259260Never load sibling UGC references.