diff --git a/models/minimax/minimax-h3-extender-ref2va/README.md b/models/minimax/minimax-h3-extender-ref2va/README.md index d59a219..6d2622d 100644 --- a/models/minimax/minimax-h3-extender-ref2va/README.md +++ b/models/minimax/minimax-h3-extender-ref2va/README.md @@ -58,5 +58,5 @@ validated clip-by-clip reuse, and final joined video/audio decode. - Use this when you want the upstream extender logic itself. - For the maintained optimized lanes, use: - - `optimized/image-to-video/minimax-h3-best-current-r2v-master-480p-3x10s/` - - `optimized/image-to-video/minimax-h3-contex-loop-3070-safe-60s-awq-int8/` + - `optimized/image-to-video/minimax-h3-r2v-master/` + - `optimized/image-to-video/minimax-h3-extension-master/` diff --git a/optimized/README.md b/optimized/README.md index 6c008a9..856b91c 100644 --- a/optimized/README.md +++ b/optimized/README.md @@ -39,8 +39,8 @@ These workflows are intended to become the practical default choices for each ca - `text-to-image/zimage-turbo-3070-fast-start/` - `image-edit/flux2klein-9b-fp8-edit-master/` - `image-edit/zimage-turbo-3070-inpaint-start/` -- `image-to-video/minimax-h3-best-current-r2v-master-480p-3x10s/` -- `image-to-video/minimax-h3-contex-loop-3070-safe-60s-awq-int8/` +- `image-to-video/minimax-h3-r2v-master/` +- `image-to-video/minimax-h3-extension-master/` - `text-to-video/minimax-h3-3070-turbo-master/` ## Workflow format diff --git a/optimized/image-to-video/README.md b/optimized/image-to-video/README.md index 85a8559..beb7c49 100644 --- a/optimized/image-to-video/README.md +++ b/optimized/image-to-video/README.md @@ -4,14 +4,14 @@ Home for maintained category masters focused on image-to-video generation. This category is intentionally capped at two maintained workflows: -- `minimax-h3-best-current-r2v-master-480p-3x10s` -- `minimax-h3-contex-loop-3070-safe-60s-awq-int8` +- `minimax-h3-r2v-master` +- `minimax-h3-extension-master` ## Roles -- `minimax-h3-best-current-r2v-master-480p-3x10s` +- `minimax-h3-r2v-master` The current reference-first generation master when the job starts from still images and should grow into connected clips. -- `minimax-h3-contex-loop-3070-safe-60s-awq-int8` +- `minimax-h3-extension-master` The current extension master when the job starts from an existing video clip and should continue through the `Contex Loop` lane. ## Policy diff --git a/optimized/image-to-video/minimax-h3-contex-loop-3070-safe-60s-awq-int8/README.md b/optimized/image-to-video/minimax-h3-extension-master/README.md similarity index 95% rename from optimized/image-to-video/minimax-h3-contex-loop-3070-safe-60s-awq-int8/README.md rename to optimized/image-to-video/minimax-h3-extension-master/README.md index 758c316..2ccb823 100644 --- a/optimized/image-to-video/minimax-h3-contex-loop-3070-safe-60s-awq-int8/README.md +++ b/optimized/image-to-video/minimax-h3-extension-master/README.md @@ -1,4 +1,4 @@ -# MiniMax H3 Contex Loop 3070 Safe 60s AWQ INT8 +# MiniMax H3 Extension Master ## Summary @@ -24,7 +24,7 @@ to `6 x 10s` clips with review pauses disabled. ## Built from - successful ComfyUI run `cc74efa5-1a53-4b2a-9d7c-7170ea7eba5b` -- the maintained `MiniMax` `ref2v` stack used in `optimized/image-to-video/minimax-h3-best-current-r2v-master-480p-3x10s/` +- the maintained `MiniMax` `ref2v` stack used in `optimized/image-to-video/minimax-h3-r2v-master/` - earlier compact `Contex Loop` experiments that established the local-safe continuation budget ## Key model stack diff --git a/optimized/image-to-video/minimax-h3-contex-loop-3070-safe-60s-awq-int8/workflow.json b/optimized/image-to-video/minimax-h3-extension-master/workflow.json similarity index 99% rename from optimized/image-to-video/minimax-h3-contex-loop-3070-safe-60s-awq-int8/workflow.json rename to optimized/image-to-video/minimax-h3-extension-master/workflow.json index a7b17e8..9880039 100644 --- a/optimized/image-to-video/minimax-h3-contex-loop-3070-safe-60s-awq-int8/workflow.json +++ b/optimized/image-to-video/minimax-h3-extension-master/workflow.json @@ -390,7 +390,7 @@ }, "widgets_values": [ "plan", - "extended_video_recovered", + "optimized/minimax_h3_extension_master/minimax_h3_extension_master_recovered", 256 ], "color": "#59451e", @@ -1324,7 +1324,7 @@ }, "widgets_values": [ "plan", - "extended_video_final", + "optimized/minimax_h3_extension_master/minimax_h3_extension_master_final", 256 ], "color": "#24533a", @@ -1743,8 +1743,8 @@ }, "widgets_values": [ "{\n \"defaults\": {\n \"duration_seconds\": 10,\n \"steps\": 5\n },\n \"shots\": [\n {\n \"id\": \"existing_video_extension_01\",\n \"prompt\": \"integrated_multimodal_description: [Shot 1] Live-action, cinematic, begin inside the incoming video context with no establishing reset. Continue the exact final camera direction and speed from the imported ending. Preserve the principal subject's screen position, identity, body proportions, wardrobe, hair, accessories, pose, expression, contact points, clothing folds, hair movement, and ongoing action across the join. Complete the unfinished movement naturally during the first seconds. The camera then follows the subject forward into the nearest plausible adjoining area already implied by the source ending. Introduce new space through continuous parallax and physically connected geometry, never through a cut, dissolve, teleport, or sudden reframing. Maintain the established lens behavior, frame geometry, lighting direction, color response, motion cadence, anatomy, and environmental scale. End while the subject is still moving and the camera is beginning a gentle arc, leaving both motions visibly unresolved for the next scene.\\n\\noverall_soundscape:\\nContinue the incoming room tone, environmental texture, movement sounds, and acoustic perspective without a sudden change at the join. Keep new footsteps, fabric movement, impacts, and spatial reflections synchronized to the visible action.\\n\\nnon_diegetic_music:\\nIf music is already audible, continue its instrumentation, tempo, key, and dynamic level without a restart; otherwise use no non-diegetic music.\",\n \"seed\": \"8401\"\n },\n {\n \"id\": \"existing_video_extension_02\",\n \"prompt\": \"integrated_multimodal_description: [Shot 1] Continue seamlessly from the previous scene with the same subject and camera motion already in progress. Preserve exact identity, wardrobe, lighting direction, environment continuity, lens behavior, and spatial scale. Let the camera complete the gentle arc and reveal the next connected part of the environment through natural parallax. Keep the subject's action continuous and physically grounded, with no visual reset. End with momentum still alive for scene three.\\n\\noverall_soundscape:\\nMaintain continuous ambience and synchronized motion sounds with no audio restart at the join.\\n\\nnon_diegetic_music:\\nIf music was already present, continue it without resetting; otherwise use none.\",\n \"seed\": \"8402\"\n },\n {\n \"id\": \"existing_video_extension_03\",\n \"prompt\": \"integrated_multimodal_description: [Shot 1] Continue directly from the unresolved motion of the previous scene. Preserve identity, wardrobe, proportions, lighting direction, lens behavior, environmental continuity, and physical momentum inherited from the prior clip. Let the subject complete the next natural beat inside the same connected space, then carry that action deeper into the location with coherent parallax, reflections, contact points, and motion cadence. End with a visible unfinished motion into scene four.\\n\\noverall_soundscape:\\nPreserve continuous environmental ambience and synchronized movement sounds across the boundary.\\n\\nnon_diegetic_music:\\nContinue any established score naturally, otherwise none.\",\n \"seed\": \"8403\"\n },\n {\n \"id\": \"existing_video_extension_04\",\n \"prompt\": \"integrated_multimodal_description: [Shot 1] Continue from the previous clip without restarting the action. Preserve exact subject identity, proportions, wardrobe, lighting response, environment continuity, and lens behavior. Let the camera and subject travel through the same connected location with readable depth and grounded momentum. Introduce a small reactive beat that emerges from the visible environment while keeping motion coherent and continuous. End with slight unresolved momentum for scene five.\\n\\noverall_soundscape:\\nKeep ambience, footsteps, cloth movement, impacts, and room reflections continuous and synchronized.\\n\\nnon_diegetic_music:\\nContinue existing music if present; otherwise none.\",\n \"seed\": \"8404\"\n },\n {\n \"id\": \"existing_video_extension_05\",\n \"prompt\": \"integrated_multimodal_description: [Shot 1] Continue cleanly from the previous scene and preserve the exact same subject, environment, and camera language. Let the action transition toward a more stable beat while still feeling like one unbroken shot. Keep background continuity, surface detail, anatomy, and lighting consistent. During the final seconds, start easing the motion toward a readable landing without fully resolving it yet.\\n\\noverall_soundscape:\\nMaintain continuous environmental ambience and synchronized movement sounds, gradually softening intensity as the sequence approaches resolution.\\n\\nnon_diegetic_music:\\nIf a score exists, continue it with restrained forward motion; otherwise none.\",\n \"seed\": \"8405\"\n },\n {\n \"id\": \"existing_video_extension_06\",\n \"prompt\": \"integrated_multimodal_description: [Shot 1] Continue seamlessly from the prior scene and resolve the one-take continuation into a confident final composition. Preserve exact identity, wardrobe, lighting direction, environment continuity, lens behavior, and spatial scale. Let the subject complete the final natural action beat and settle into a readable closing frame without freezing abruptly or breaking continuity. Keep secondary environmental motion alive long enough to make the ending feel real and complete.\\n\\noverall_soundscape:\\nCarry the established ambience and synchronized action sounds through the ending, letting the acoustic energy resolve naturally without a restart.\\n\\nnon_diegetic_music:\\nIf music was established, let it resolve the same instrumentation, tempo, and key with a restrained ending; otherwise none.\",\n \"seed\": \"8406\"\n }\n ]\n}", - "minimax_h3_loop_safe_60s_awq_int8", - "minimax-h3-loop-safe-awq-int8-60s-v1", + "minimax_h3_extension_master", + "minimax-h3-extension-master-v1", 864, 480, 22, diff --git a/optimized/image-to-video/minimax-h3-best-current-r2v-master-480p-3x10s/README.md b/optimized/image-to-video/minimax-h3-r2v-master/README.md similarity index 96% rename from optimized/image-to-video/minimax-h3-best-current-r2v-master-480p-3x10s/README.md rename to optimized/image-to-video/minimax-h3-r2v-master/README.md index 46d456e..b28e383 100644 --- a/optimized/image-to-video/minimax-h3-best-current-r2v-master-480p-3x10s/README.md +++ b/optimized/image-to-video/minimax-h3-r2v-master/README.md @@ -1,4 +1,4 @@ -# MiniMax H3 Best Current R2V Master 480p 3x10s +# MiniMax H3 R2V Master ## Summary @@ -68,7 +68,7 @@ subject-definition prompts for continuity across the chain. - use this first when the goal is: keep one or more reference images stable and extend the action across several connected clips -- use `optimized/image-to-video/minimax-h3-contex-loop-3070-safe-60s-awq-int8/` +- use `optimized/image-to-video/minimax-h3-extension-master/` instead when the main job is extending an already existing source video rather than generating a new reference-driven chain diff --git a/optimized/image-to-video/minimax-h3-best-current-r2v-master-480p-3x10s/workflow.json b/optimized/image-to-video/minimax-h3-r2v-master/workflow.json similarity index 99% rename from optimized/image-to-video/minimax-h3-best-current-r2v-master-480p-3x10s/workflow.json rename to optimized/image-to-video/minimax-h3-r2v-master/workflow.json index ce128d5..7b860b5 100644 --- a/optimized/image-to-video/minimax-h3-best-current-r2v-master-480p-3x10s/workflow.json +++ b/optimized/image-to-video/minimax-h3-r2v-master/workflow.json @@ -550,7 +550,7 @@ }, "widgets_values": [ 24, - "MiniMax_H3_cached", + "optimized/minimax_h3_r2v_master/minimax_h3_r2v_master", "", "H.264", 17, diff --git a/optimized/notes/minimax.md b/optimized/notes/minimax.md index b0f919c..bcc86a3 100644 --- a/optimized/notes/minimax.md +++ b/optimized/notes/minimax.md @@ -35,8 +35,8 @@ - Keep `optimized/text-to-video/minimax-h3-3070-turbo-master/` as the first MiniMax default. - Keep the INT8 reference-video workflow as the next donor for future low-VRAM optimization. - Keep only two maintained `optimized/image-to-video/` MiniMax lanes: - - `optimized/image-to-video/minimax-h3-best-current-r2v-master-480p-3x10s/` as the reference-first generation master - - `optimized/image-to-video/minimax-h3-contex-loop-3070-safe-60s-awq-int8/` as the imported-video extension master + - `optimized/image-to-video/minimax-h3-r2v-master/` as the reference-first generation master + - `optimized/image-to-video/minimax-h3-extension-master/` as the imported-video extension master - Use the new same-day `javawock7618` `INT8 R2V` import as the main donor for future `best-current-r2v` revisions if the current master needs another pass. ## Ref2VA prompt guidance