Nodes/Ultra MiniMax H3 Prompt/๐ŸŽฌ Ultra MiniMax H3 Prompt
ComfyUI Node

๐ŸŽฌ Ultra MiniMax H3 Prompt

MiniMax H3 prompt compiler with default Korean-to-English scene translation and timeline-based multi-person dialogue/action support.

By Yamoon-jjangยทCreated 22 days agoยทUpdated 9 days agoยท 16
๐ŸŽฌ Ultra MiniMax H3 Prompt
  • start_image
  • end_image
  • h3_prompt
  • info
โ—„scene_promptโ–บ
โ—„scene_translateko_to_enโ–บ
โ—„scene_input_rolescene_actionโ–บ
โ—„modeautoโ–บ
โ—„duration_seconds5โ–บ
โ—„shot_layoutauto_safeโ–บ
โ—„visual_styleautoโ–บ
โ—„narrator_voice_descโ–บ
โ—„camera_motionautoโ–บ
โ—„camera_amplitudedefaultโ–บ
โ—„camera_speeddefaultโ–บ
โ—„camera_sequence_enabledfalseโ–บ
โ—„camera_shot_count2โ–บ
โ—„camera_shot_1_targetcurrent_actionโ–บ
โ—„camera_shot_1_framingautoโ–บ
โ—„camera_shot_2_enabledtrueโ–บ
โ—„camera_shot_2_start_seconds0.5โ–บ
โ—„camera_shot_2_transitioncutโ–บ
โ—„camera_shot_2_targetcurrent_actionโ–บ
โ—„camera_shot_2_framingautoโ–บ
โ—„camera_shot_2_motionautoโ–บ
โ—„camera_shot_2_amplitudeinheritโ–บ
โ—„camera_shot_2_speedinheritโ–บ
โ—„camera_shot_3_enabledfalseโ–บ
โ—„camera_shot_3_start_seconds1.0โ–บ
โ—„camera_shot_3_transitioncutโ–บ
โ—„camera_shot_3_targetcurrent_actionโ–บ
โ—„camera_shot_3_framingautoโ–บ
โ—„camera_shot_3_motionautoโ–บ
โ—„camera_shot_3_amplitudeinheritโ–บ
โ—„camera_shot_3_speedinheritโ–บ
โ—„camera_shot_4_enabledfalseโ–บ
โ—„camera_shot_4_start_seconds1.5โ–บ
โ—„camera_shot_4_transitioncutโ–บ
โ—„camera_shot_4_targetcurrent_actionโ–บ
โ—„camera_shot_4_framingautoโ–บ
โ—„camera_shot_4_motionautoโ–บ
โ—„camera_shot_4_amplitudeinheritโ–บ
โ—„camera_shot_4_speedinheritโ–บ
โ—„camera_shot_5_enabledfalseโ–บ
โ—„camera_shot_5_start_seconds2.0โ–บ
โ—„camera_shot_5_transitioncutโ–บ
โ—„camera_shot_5_targetcurrent_actionโ–บ
โ—„camera_shot_5_framingautoโ–บ
โ—„camera_shot_5_motionautoโ–บ
โ—„camera_shot_5_amplitudeinheritโ–บ
โ—„camera_shot_5_speedinheritโ–บ
โ—„camera_shot_6_enabledfalseโ–บ
โ—„camera_shot_6_start_seconds2.5โ–บ
โ—„camera_shot_6_transitioncutโ–บ
โ—„camera_shot_6_targetcurrent_actionโ–บ
โ—„camera_shot_6_framingautoโ–บ
โ—„camera_shot_6_motionautoโ–บ
โ—„camera_shot_6_amplitudeinheritโ–บ
โ—„camera_shot_6_speedinheritโ–บ
โ—„reference_fidelitybalancedโ–บ
โ—„soundscape_modeautoโ–บ
โ—„sound_hintโ–บ
โ—„music_modenoneโ–บ
โ—„music_hintโ–บ
โ—„dialogue_textโ–บ
โ—„dialogue_languageautoโ–บ
โ—„dialogue_deliveryon_screenโ–บ
โ—„person_count1โ–บ
โ—„turn_count1โ–บ
โ—„person_a_genderunspecifiedโ–บ
โ—„person_a_ageunspecifiedโ–บ
โ—„person_a_advancedfalseโ–บ
โ—„person_a_appearance_presetnoneโ–บ
โ—„person_a_appearance_descโ–บ
โ—„person_a_clothing_presetnoneโ–บ
โ—„person_a_clothing_descโ–บ
โ—„person_a_voice_descโ–บ
โ—„person_b_genderunspecifiedโ–บ
โ—„person_b_ageunspecifiedโ–บ
โ—„person_b_advancedfalseโ–บ
โ—„person_b_appearance_presetnoneโ–บ
โ—„person_b_appearance_descโ–บ
โ—„person_b_clothing_presetnoneโ–บ
โ—„person_b_clothing_descโ–บ
โ—„person_b_voice_descโ–บ
โ—„person_c_genderunspecifiedโ–บ
โ—„person_c_ageunspecifiedโ–บ
โ—„person_c_advancedfalseโ–บ
โ—„person_c_appearance_presetnoneโ–บ
โ—„person_c_appearance_descโ–บ
โ—„person_c_clothing_presetnoneโ–บ
โ—„person_c_clothing_descโ–บ
โ—„person_c_voice_descโ–บ
โ—„turn_1_speakerAโ–บ
โ—„turn_1_elapsed_seconds0.0โ–บ
โ—„turn_1_actionโ–บ
โ—„turn_1_dialogue_textโ–บ
โ—„turn_1_dialogue_languageautoโ–บ
โ—„turn_1_dialogue_deliveryon_screenโ–บ
โ—„turn_1_vocal_2_textโ–บ
โ—„turn_1_vocal_3_textโ–บ
โ—„turn_2_speakerAโ–บ
โ—„turn_2_elapsed_seconds0.0โ–บ
โ—„turn_2_actionโ–บ
โ—„turn_2_dialogue_textโ–บ
โ—„turn_2_dialogue_languageautoโ–บ
โ—„turn_2_dialogue_deliveryon_screenโ–บ
โ—„turn_2_vocal_2_textโ–บ
โ—„turn_2_vocal_3_textโ–บ
โ—„turn_3_speakerAโ–บ
โ—„turn_3_elapsed_seconds0.0โ–บ
โ—„turn_3_actionโ–บ
โ—„turn_3_dialogue_textโ–บ
โ—„turn_3_dialogue_languageautoโ–บ
โ—„turn_3_dialogue_deliveryon_screenโ–บ
โ—„turn_3_vocal_2_textโ–บ
โ—„turn_3_vocal_3_textโ–บ
โ—„turn_4_speakerAโ–บ
โ—„turn_4_elapsed_seconds0.0โ–บ
โ—„turn_4_actionโ–บ
โ—„turn_4_dialogue_textโ–บ
โ—„turn_4_dialogue_languageautoโ–บ
โ—„turn_4_dialogue_deliveryon_screenโ–บ
โ—„turn_4_vocal_2_textโ–บ
โ—„turn_4_vocal_3_textโ–บ
โ—„turn_5_speakerAโ–บ
โ—„turn_5_elapsed_seconds0.0โ–บ
โ—„turn_5_actionโ–บ
โ—„turn_5_dialogue_textโ–บ
โ—„turn_5_dialogue_languageautoโ–บ
โ—„turn_5_dialogue_deliveryon_screenโ–บ
โ—„turn_5_vocal_2_textโ–บ
โ—„turn_5_vocal_3_textโ–บ
โ—„turn_6_speakerAโ–บ
โ—„turn_6_elapsed_seconds0.0โ–บ
โ—„turn_6_actionโ–บ
โ—„turn_6_dialogue_textโ–บ
โ—„turn_6_dialogue_languageautoโ–บ
โ—„turn_6_dialogue_deliveryon_screenโ–บ
โ—„turn_6_vocal_2_textโ–บ
โ—„turn_6_vocal_3_textโ–บ
โ—„timeline_events_jsonโ–บ
โ—„variation_seed0โ–บ
โ—„i2va_preserve_reference_clothingtrueโ–บ
โ—„ultra_scene_validation_gateokโ–บ
โ—„ultra_ui_localeenโ–บ
โ—„ultra_extension_options_json{}โ–บ
CategoryUltra/MiniMax H3

Inputs (145)

NameTypeDefaultDescription
scene_promptSTRINGShared scene/backdrop only: location, mood, lighting, and initial character arrangement. Put changing actions and dialogue in Ordered Steps. Example: A wide meadow in warm late-afternoon light; a woman and a man stand facing each other. Text containing only spaces or line breaks is treated as empty. In T2VA, a completely empty draft remains queueable; the h3_prompt branch waits with status=waiting_for_content instead of becoming a native input error.
scene_translateCOMBOko_to_enGlobal non-dialogue descriptive-text translation control. Automatic: translate Korean scene/action/profile/reference/sound/music descriptions to English while leaving already-English text unchanged. Off: preserve every descriptive field exactly as entered and make no online translation request. Dialogue is never translated.
scene_input_roleCOMBOscene_actionLegacy serialized workflow field retained only for backward-compatible widget alignment. Current runtime always uses Timeline mode and ignores this value.
modeCOMBOautoAuto follows the actual image connections: none=T2VA, start only=I2VA, end only=L2VA, start+end=FL2VA. Manual T2VA stays T2VA and ignores connected images for prompt-mode resolution. Manual I2VA requires a start image, L2VA requires an end image, and FL2VA requires both; when a required image is missing, the node shows the guidance badge and quietly waits instead of changing to another mode.
duration_secondsINT54โ€“30Reference duration for H3 timing. This node accepts 4-30 whole seconds. MiniMax H3 officially supports up to 15 seconds; 16-30 seconds are an advanced/local-use range and may increase generation time and memory use sharply. Match this to the actual H3 video-generation duration.
shot_layoutCOMBOauto_safeLegacy serialized workflow field retained only for backward-compatible widget alignment. Current runtime ignores this value; use Multiple Camera Cuts for explicit shots.
visual_styleCOMBOauto22 options: auto, cinematic, family_film, warm_cinematic, natural, live_action, +16
narrator_voice_descSTRINGOptional custom Narrator voice characteristics. This is a normal ComfyUI STRING widget/input and can be connected from an external STRING source.
camera_motionCOMBOautoBase camera motion for the shot. In multiple-camera mode, this remains the motion for Shot 1 because Shot 1 begins the video and has no transition-in. Tracking Shot follows the subject smoothly; camera shake emphasizes shake; Handheld Tracking follows the action with natural operator movement and irregular micro-corrections.
camera_amplitudeCOMBOdefaultSingle-camera mode: applies to the selected camera motion. Multiple-camera mode: this is Shot 1 amplitude and the inherited default for later Shots whose Camera Amplitude is set to Follow Base Settings.
camera_speedCOMBOdefaultSingle-camera mode: applies to the selected camera motion. Multiple-camera mode: this is Shot 1 speed and the inherited default for later Shots whose Camera Speed is set to Follow Base Settings.
camera_sequence_enabledBOOLEANfalseEnable explicit H3 shot changes. Shot 1 always exists. When enabled, Shot 2-6 can be enabled in order and each Shot gets its own start time in seconds, independent of Timeline Progress timing.
camera_shot_countINT22โ€“6Internal derived active-shot count. The UI derives this from the enabled Shot 2-6 blocks; users do not set it directly.
camera_shot_1_targetCOMBOcurrent_actionShot 1 is the default opening shot and always starts at 0.0 seconds. Choose what the opening shot frames. Its timing is independent of Timeline Progress timing.
camera_shot_1_framingCOMBOautoFraming for the default Shot 1. Shot 1 always exists even when Multiple Camera Cuts is OFF.
camera_shot_2_enabledBOOLEANtrueEnable Shot 2. When ON, set Shot 2 Start Time (seconds). Shot timing is independent of Timeline Progress timing. Shots must be enabled in order without gaps.
camera_shot_2_start_secondsFLOAT0.50.5โ€“29.5Video time in seconds when Shot 2 starts. Use 0.5-second steps. This timing is independent of Timeline Progress start times; the UI dynamically limits the maximum to video duration minus 0.5 seconds.
camera_shot_2_transitionCOMBOcutTransition used when Shot 2 starts.
camera_shot_2_targetCOMBOcurrent_actionWhat Shot 2 should frame.
camera_shot_2_framingCOMBOautoFraming for Shot 2. Available choices follow the selected target.
camera_shot_2_motionCOMBOautoOptional camera movement inside Shot 2 after its transition.
camera_shot_2_amplitudeCOMBOinheritAmplitude for Shot 2 camera motion. Follow Base Settings inherits the top Camera Amplitude; Default ยท Omit suppresses the amplitude phrase for this Shot; Small/Large override it only for this Shot.
camera_shot_2_speedCOMBOinheritSpeed for Shot 2 camera motion. Follow Base Settings inherits the top Camera Speed; Default ยท Omit suppresses the speed phrase for this Shot; Slow/Fast override it only for this Shot.
camera_shot_3_enabledBOOLEANfalseEnable Shot 3. When ON, set Shot 3 Start Time (seconds). Shot timing is independent of Timeline Progress timing. Shots must be enabled in order without gaps.
camera_shot_3_start_secondsFLOAT1.00.5โ€“29.5Video time in seconds when Shot 3 starts. Use 0.5-second steps. This timing is independent of Timeline Progress start times; the UI dynamically limits the maximum to video duration minus 0.5 seconds.
camera_shot_3_transitionCOMBOcutTransition used when Shot 3 starts.
camera_shot_3_targetCOMBOcurrent_actionWhat Shot 3 should frame.
camera_shot_3_framingCOMBOautoFraming for Shot 3. Available choices follow the selected target.
camera_shot_3_motionCOMBOautoOptional camera movement inside Shot 3 after its transition.
camera_shot_3_amplitudeCOMBOinheritAmplitude for Shot 3 camera motion. Follow Base Settings inherits the top Camera Amplitude; Default ยท Omit suppresses the amplitude phrase for this Shot; Small/Large override it only for this Shot.
camera_shot_3_speedCOMBOinheritSpeed for Shot 3 camera motion. Follow Base Settings inherits the top Camera Speed; Default ยท Omit suppresses the speed phrase for this Shot; Slow/Fast override it only for this Shot.
camera_shot_4_enabledBOOLEANfalseEnable Shot 4. When ON, set Shot 4 Start Time (seconds). Shot timing is independent of Timeline Progress timing. Shots must be enabled in order without gaps.
camera_shot_4_start_secondsFLOAT1.50.5โ€“29.5Video time in seconds when Shot 4 starts. Use 0.5-second steps. This timing is independent of Timeline Progress start times; the UI dynamically limits the maximum to video duration minus 0.5 seconds.
camera_shot_4_transitionCOMBOcutTransition used when Shot 4 starts.
camera_shot_4_targetCOMBOcurrent_actionWhat Shot 4 should frame.
camera_shot_4_framingCOMBOautoFraming for Shot 4. Available choices follow the selected target.
camera_shot_4_motionCOMBOautoOptional camera movement inside Shot 4 after its transition.
camera_shot_4_amplitudeCOMBOinheritAmplitude for Shot 4 camera motion. Follow Base Settings inherits the top Camera Amplitude; Default ยท Omit suppresses the amplitude phrase for this Shot; Small/Large override it only for this Shot.
camera_shot_4_speedCOMBOinheritSpeed for Shot 4 camera motion. Follow Base Settings inherits the top Camera Speed; Default ยท Omit suppresses the speed phrase for this Shot; Slow/Fast override it only for this Shot.
camera_shot_5_enabledBOOLEANfalseEnable Shot 5. When ON, set Shot 5 Start Time (seconds). Shot timing is independent of Timeline Progress timing. Shots must be enabled in order without gaps.
camera_shot_5_start_secondsFLOAT2.00.5โ€“29.5Video time in seconds when Shot 5 starts. Use 0.5-second steps. This timing is independent of Timeline Progress start times; the UI dynamically limits the maximum to video duration minus 0.5 seconds.
camera_shot_5_transitionCOMBOcutTransition used when Shot 5 starts.
camera_shot_5_targetCOMBOcurrent_actionWhat Shot 5 should frame.
camera_shot_5_framingCOMBOautoFraming for Shot 5. Available choices follow the selected target.
camera_shot_5_motionCOMBOautoOptional camera movement inside Shot 5 after its transition.
camera_shot_5_amplitudeCOMBOinheritAmplitude for Shot 5 camera motion. Follow Base Settings inherits the top Camera Amplitude; Default ยท Omit suppresses the amplitude phrase for this Shot; Small/Large override it only for this Shot.
camera_shot_5_speedCOMBOinheritSpeed for Shot 5 camera motion. Follow Base Settings inherits the top Camera Speed; Default ยท Omit suppresses the speed phrase for this Shot; Slow/Fast override it only for this Shot.
camera_shot_6_enabledBOOLEANfalseEnable Shot 6. When ON, set Shot 6 Start Time (seconds). Shot timing is independent of Timeline Progress timing. Shots must be enabled in order without gaps.
camera_shot_6_start_secondsFLOAT2.50.5โ€“29.5Video time in seconds when Shot 6 starts. Use 0.5-second steps. This timing is independent of Timeline Progress start times; the UI dynamically limits the maximum to video duration minus 0.5 seconds.
camera_shot_6_transitionCOMBOcutTransition used when Shot 6 starts.
camera_shot_6_targetCOMBOcurrent_actionWhat Shot 6 should frame.
camera_shot_6_framingCOMBOautoFraming for Shot 6. Available choices follow the selected target.
camera_shot_6_motionCOMBOautoOptional camera movement inside Shot 6 after its transition.
camera_shot_6_amplitudeCOMBOinheritAmplitude for Shot 6 camera motion. Follow Base Settings inherits the top Camera Amplitude; Default ยท Omit suppresses the amplitude phrase for this Shot; Small/Large override it only for this Shot.
camera_shot_6_speedCOMBOinheritSpeed for Shot 6 camera motion. Follow Base Settings inherits the top Camera Speed; Default ยท Omit suppresses the speed phrase for this Shot; Slow/Fast override it only for this Shot.
reference_fidelityCOMBObalanced3 options: strict, balanced, creative
soundscape_modeCOMBOautoControls ambience, physical-action sounds, and non-verbal human sounds. 'silent' suppresses those sounds only; dialogue and non-diegetic music remain separate. If both are also absent, overall_soundscape uses N/A for complete silence.
sound_hintSTRINGOptional ambience, physical-action sound, or non-verbal human sound. Do not put spoken dialogue or music here. English is recommended; Korean input is translated internally to English before H3 prompt assembly.
music_modeCOMBOnone33 options: none, sparse_piano_strings, acoustic_guitar_bass, classical_guitar, soft_light_music, gentle_classical, +27
music_hintSTRINGOptional audience-only non-diegetic music description. Prefer instrumentation, tempo/speed, rhythm, texture, and dynamic changes; avoid abstract mood/emotional-function wording. Korean input is translated internally to English.
dialogue_textSTRINGLegacy serialized workflow field retained only for backward-compatible widget alignment. Current runtime ignores this field; use Timeline dialogue instead.
dialogue_languageCOMBOautoLegacy serialized workflow field retained only for backward-compatible widget alignment. Current runtime ignores this value; use Timeline dialogue language instead.
dialogue_deliveryCOMBOon_screenLegacy serialized workflow field retained only for backward-compatible widget alignment. Current runtime ignores this value; use Timeline dialogue delivery instead.
person_countINT11โ€“3How many main character profiles are available to the ordered action/dialogue steps.
turn_countINT11โ€“6Number of timeline events shown in the UI. Each visible row is one real event with one target person/group, one start time, one action, one dialogue field, one dialogue language, and one delivery mode. Final prompt order still follows start time.
person_a_genderCOMBOunspecifiedOptional profile detail for Main Character 1.
person_a_ageCOMBOunspecifiedOptional age group for Main Character 1. Final prompts use natural phrases such as a woman in her thirties. Adult-only built-in presets are selectable only with an explicit age of 20s or older; changing to another age temporarily hides an incompatible preset and restores it when an adult age is selected again.
person_a_advancedBOOLEANfalseShow or hide advanced per-person profile options for Main Character 1. Hidden values are preserved and restored when shown again or when person count changes.
person_a_appearance_presetCOMBOnoneOptional appearance preset for Main Character 1. Combines naturally with Appearance Additional Input when both are used. Adult-only built-in choices are available only with an explicit age of 20s or older.
person_a_appearance_descSTRINGOptional additional appearance/body description for Main Character 1. English is recommended; Korean is translated internally. Prefer concrete body/build, silhouette, hair, or distinctive physical traits over vague beauty adjectives. This is combined with the appearance preset when both are provided.
person_a_clothing_presetCOMBOnoneOptional clothing preset for Main Character 1. Combines naturally with Clothing Additional Input when both are used. Adult-only built-in choices are available only with an explicit age of 20s or older.
person_a_clothing_descSTRINGOptional additional clothing for Main Character 1. English is recommended; Korean is translated internally. Examples: color, material, or a specific garment refinement. This is combined with the clothing preset when both are provided.
person_a_voice_descSTRINGOptional voice identity for Main Character 1. English is recommended; Korean description is translated internally. Its H3 S ID is assigned at this character's first spoken line and reused afterward.
person_b_genderCOMBOunspecifiedOptional profile detail for Main Character 2.
person_b_ageCOMBOunspecifiedOptional age group for Main Character 2. Final prompts use natural phrases such as a man in his fifties. Adult-only built-in presets are selectable only with an explicit age of 20s or older; changing to another age temporarily hides an incompatible preset and restores it when an adult age is selected again.
person_b_advancedBOOLEANfalseShow or hide advanced per-person profile options for Main Character 2. Hidden values are preserved and restored when shown again or when person count changes.
person_b_appearance_presetCOMBOnoneOptional appearance preset for Main Character 2. Combines naturally with Appearance Additional Input when both are used. Adult-only built-in choices are available only with an explicit age of 20s or older.
person_b_appearance_descSTRINGOptional additional appearance/body description for Main Character 2. English is recommended; Korean is translated internally. This is combined with the appearance preset when both are provided.
person_b_clothing_presetCOMBOnoneOptional clothing preset for Main Character 2. Combines naturally with Clothing Additional Input when both are used. Adult-only built-in choices are available only with an explicit age of 20s or older.
person_b_clothing_descSTRINGOptional additional clothing for Main Character 2. English is recommended; Korean is translated internally. This is combined with the clothing preset when both are provided.
person_b_voice_descSTRINGOptional voice identity for Main Character 2. English is recommended; Korean description is translated internally. Its H3 S ID is assigned at this character's first spoken line and reused afterward.
person_c_genderCOMBOunspecifiedOptional profile detail for Main Character 3.
person_c_ageCOMBOunspecifiedOptional age group for Main Character 3. Final prompts use natural English age wording. Adult-only built-in presets are selectable only with an explicit age of 20s or older; changing to another age temporarily hides an incompatible preset and restores it when an adult age is selected again.
person_c_advancedBOOLEANfalseShow or hide advanced per-person profile options for Main Character 3. Hidden values are preserved and restored when shown again or when person count changes.
person_c_appearance_presetCOMBOnoneOptional appearance preset for Main Character 3. Combines naturally with Appearance Additional Input when both are used. Adult-only built-in choices are available only with an explicit age of 20s or older.
person_c_appearance_descSTRINGOptional additional appearance/body description for Main Character 3. English is recommended; Korean is translated internally. This is combined with the appearance preset when both are provided.
person_c_clothing_presetCOMBOnoneOptional clothing preset for Main Character 3. Combines naturally with Clothing Additional Input when both are used. Adult-only built-in choices are available only with an explicit age of 20s or older.
person_c_clothing_descSTRINGOptional additional clothing for Main Character 3. English is recommended; Korean is translated internally. This is combined with the clothing preset when both are provided.
person_c_voice_descSTRINGOptional voice identity for Main Character 3. English is recommended; Korean description is translated internally. Its H3 S ID is assigned at this character's first spoken line and reused afterward.
turn_1_speakerCOMBOATarget person(s) for this timeline event. This row is one real event. If you want another person to act or speak at a different moment, use the next Timeline row. Combined targets act or speak together.
turn_1_elapsed_secondsFLOAT0.00โ€“29.5Approximate start-time target for this timeline event, not a frame-exact trigger. Natural sequential motion and normal playback speed take priority. The UI uses 0.5-second steps and dynamically limits the maximum to video duration minus 0.5 seconds (for example, 10 seconds โ†’ 9.5, 30 seconds โ†’ 29.5).
turn_1_actionSTRINGDescribe what happens at this timeline event. English is recommended. Korean action text is translated internally to English before H3 prompt assembly and always remains an action; dialogue text is never auto-translated.
turn_1_dialogue_textSTRINGOptional spoken line for this step. The original dialogue text is preserved and is never auto-translated.
turn_1_dialogue_languageCOMBOautoLanguage tag written inside H3 <d>[Language] ...</d>. Auto prioritizes Korean vs English and also recognizes obvious supported writing systems; choose a language manually when another language needs an exact tag.
turn_1_dialogue_deliveryCOMBOon_screenChoose direct on-screen speech or off-screen voiceover for this line.
turn_1_vocal_2_textSTRINGProgress 1 Vocal 2 dialogue. Preserved verbatim and connectable from an external STRING source.
turn_1_vocal_3_textSTRINGProgress 1 Vocal 3 dialogue. Preserved verbatim and connectable from an external STRING source.
turn_2_speakerCOMBOATarget person(s) for this timeline event. This row is one real event. If you want another person to act or speak at a different moment, use the next Timeline row. Combined targets act or speak together.
turn_2_elapsed_secondsFLOAT0.00โ€“29.5Approximate start-time target for this timeline event, not a frame-exact trigger. Natural sequential motion and normal playback speed take priority. The UI uses 0.5-second steps and dynamically limits the maximum to video duration minus 0.5 seconds (for example, 10 seconds โ†’ 9.5, 30 seconds โ†’ 29.5).
turn_2_actionSTRINGDescribe what happens at this timeline event. English is recommended. Korean action text is translated internally to English before H3 prompt assembly and always remains an action; dialogue text is never auto-translated.
turn_2_dialogue_textSTRINGOptional spoken line for this step. The original dialogue text is preserved and is never auto-translated.
turn_2_dialogue_languageCOMBOautoLanguage tag written inside H3 <d>[Language] ...</d>. Auto prioritizes Korean vs English and also recognizes obvious supported writing systems; choose a language manually when another language needs an exact tag.
turn_2_dialogue_deliveryCOMBOon_screenChoose direct on-screen speech or off-screen voiceover for this line.
turn_2_vocal_2_textSTRINGProgress 2 Vocal 2 dialogue. Preserved verbatim and connectable from an external STRING source.
turn_2_vocal_3_textSTRINGProgress 2 Vocal 3 dialogue. Preserved verbatim and connectable from an external STRING source.
turn_3_speakerCOMBOATarget person(s) for this timeline event. This row is one real event. If you want another person to act or speak at a different moment, use the next Timeline row. Combined targets act or speak together.
turn_3_elapsed_secondsFLOAT0.00โ€“29.5Approximate start-time target for this timeline event, not a frame-exact trigger. Natural sequential motion and normal playback speed take priority. The UI uses 0.5-second steps and dynamically limits the maximum to video duration minus 0.5 seconds (for example, 10 seconds โ†’ 9.5, 30 seconds โ†’ 29.5).
turn_3_actionSTRINGDescribe what happens at this timeline event. English is recommended. Korean action text is translated internally to English before H3 prompt assembly and always remains an action; dialogue text is never auto-translated.
turn_3_dialogue_textSTRINGOptional spoken line for this step. The original dialogue text is preserved and is never auto-translated.
turn_3_dialogue_languageCOMBOautoLanguage tag written inside H3 <d>[Language] ...</d>. Auto prioritizes Korean vs English and also recognizes obvious supported writing systems; choose a language manually when another language needs an exact tag.
turn_3_dialogue_deliveryCOMBOon_screenChoose direct on-screen speech or off-screen voiceover for this line.
turn_3_vocal_2_textSTRINGProgress 3 Vocal 2 dialogue. Preserved verbatim and connectable from an external STRING source.
turn_3_vocal_3_textSTRINGProgress 3 Vocal 3 dialogue. Preserved verbatim and connectable from an external STRING source.
turn_4_speakerCOMBOATarget person(s) for this timeline event. This row is one real event. If you want another person to act or speak at a different moment, use the next Timeline row. Combined targets act or speak together.
turn_4_elapsed_secondsFLOAT0.00โ€“29.5Approximate start-time target for this timeline event, not a frame-exact trigger. Natural sequential motion and normal playback speed take priority. The UI uses 0.5-second steps and dynamically limits the maximum to video duration minus 0.5 seconds (for example, 10 seconds โ†’ 9.5, 30 seconds โ†’ 29.5).
turn_4_actionSTRINGDescribe what happens at this timeline event. English is recommended. Korean action text is translated internally to English before H3 prompt assembly and always remains an action; dialogue text is never auto-translated.
turn_4_dialogue_textSTRINGOptional spoken line for this step. The original dialogue text is preserved and is never auto-translated.
turn_4_dialogue_languageCOMBOautoLanguage tag written inside H3 <d>[Language] ...</d>. Auto prioritizes Korean vs English and also recognizes obvious supported writing systems; choose a language manually when another language needs an exact tag.
turn_4_dialogue_deliveryCOMBOon_screenChoose direct on-screen speech or off-screen voiceover for this line.
turn_4_vocal_2_textSTRINGProgress 4 Vocal 2 dialogue. Preserved verbatim and connectable from an external STRING source.
turn_4_vocal_3_textSTRINGProgress 4 Vocal 3 dialogue. Preserved verbatim and connectable from an external STRING source.
turn_5_speakerCOMBOATarget person(s) for this timeline event. This row is one real event. If you want another person to act or speak at a different moment, use the next Timeline row. Combined targets act or speak together.
turn_5_elapsed_secondsFLOAT0.00โ€“29.5Approximate start-time target for this timeline event, not a frame-exact trigger. Natural sequential motion and normal playback speed take priority. The UI uses 0.5-second steps and dynamically limits the maximum to video duration minus 0.5 seconds (for example, 10 seconds โ†’ 9.5, 30 seconds โ†’ 29.5).
turn_5_actionSTRINGDescribe what happens at this timeline event. English is recommended. Korean action text is translated internally to English before H3 prompt assembly and always remains an action; dialogue text is never auto-translated.
turn_5_dialogue_textSTRINGOptional spoken line for this step. The original dialogue text is preserved and is never auto-translated.
turn_5_dialogue_languageCOMBOautoLanguage tag written inside H3 <d>[Language] ...</d>. Auto prioritizes Korean vs English and also recognizes obvious supported writing systems; choose a language manually when another language needs an exact tag.
turn_5_dialogue_deliveryCOMBOon_screenChoose direct on-screen speech or off-screen voiceover for this line.
turn_5_vocal_2_textSTRINGProgress 5 Vocal 2 dialogue. Preserved verbatim and connectable from an external STRING source.
turn_5_vocal_3_textSTRINGProgress 5 Vocal 3 dialogue. Preserved verbatim and connectable from an external STRING source.
turn_6_speakerCOMBOATarget person(s) for this timeline event. This row is one real event. If you want another person to act or speak at a different moment, use the next Timeline row. Combined targets act or speak together.
turn_6_elapsed_secondsFLOAT0.00โ€“29.5Approximate start-time target for this timeline event, not a frame-exact trigger. Natural sequential motion and normal playback speed take priority. The UI uses 0.5-second steps and dynamically limits the maximum to video duration minus 0.5 seconds (for example, 10 seconds โ†’ 9.5, 30 seconds โ†’ 29.5).
turn_6_actionSTRINGDescribe what happens at this timeline event. English is recommended. Korean action text is translated internally to English before H3 prompt assembly and always remains an action; dialogue text is never auto-translated.
turn_6_dialogue_textSTRINGOptional spoken line for this step. The original dialogue text is preserved and is never auto-translated.
turn_6_dialogue_languageCOMBOautoLanguage tag written inside H3 <d>[Language] ...</d>. Auto prioritizes Korean vs English and also recognizes obvious supported writing systems; choose a language manually when another language needs an exact tag.
turn_6_dialogue_deliveryCOMBOon_screenChoose direct on-screen speech or off-screen voiceover for this line.
turn_6_vocal_2_textSTRINGProgress 6 Vocal 2 dialogue. Preserved verbatim and connectable from an external STRING source.
turn_6_vocal_3_textSTRINGProgress 6 Vocal 3 dialogue. Preserved verbatim and connectable from an external STRING source.
timeline_events_jsonSTRINGโ€”
variation_seedINT00โ€“9223372036854776000Legacy serialized workflow field retained only for backward-compatible widget alignment. Current Timeline runtime ignores this value.
i2va_preserve_reference_clothingBOOLEANtrueI2VA only. Keep clothing, colors, materials, and accessories from the start image stable throughout the entire video. While enabled, per-person clothing presets/free text are preserved in the workflow but ignored for prompt generation. Ordered Steps and shot count never toggle this value.
ultra_scene_validation_gateCOMBOok2 options: ok, missing
ultra_ui_localeCOMBOen2 options: en, ko
ultra_extension_options_jsonSTRING{}โ€”
start_imageoptIMAGEAuto-mode topology signal and the required reference for manual I2VA/FL2VA. Manual T2VA ignores it for prompt-mode resolution. Pixels are never inspected.
end_imageoptIMAGEAuto-mode topology signal and the required reference for manual L2VA/FL2VA. Manual T2VA and manual I2VA ignore it for prompt-mode resolution. Pixels are never inspected.

Outputs (2)

NameTypeDescription
h3_promptSTRINGโ€”
infoSTRINGโ€”