Make videos on the computer you already own. FreeVideo runs MiniMax H3 in as little as 8 GB of VRAM and 16 GB of RAM, and adapts its acceleration path to your hardware.
Agent rules: cut to a new shot by default, continue only what must stay unbroken
Bowen Xueffdaaf3fe4d9
diff --git a/freevideo_engine/agent/guide.py b/freevideo_engine/agent/guide.pyindex 8474c3c..d5f9905 100644--- a/freevideo_engine/agent/guide.py+++ b/freevideo_engine/agent/guide.py@@ -109,7 +109,8 @@ GUIDE = {for name, (method, path, summary) in SESSION_OPERATIONS.items()],'filming': ['Filming is for a film of several shots, each picked from takes. One video of up to 60 seconds is one generate '- 'job (longer requests continue and join automatically); to polish an idea first, run enhance, then generate.',+ 'job (longer requests continue and join automatically); to polish an idea first, run enhance, then generate. '+ 'A film is cut into shots: see the rule "Cut by default" before making one long continued video.','Use drama plan → start → get → sheet → choose/reshoot → finish; inspect todo after every change.','At start tell the user the estimated total time; on over_budget decide whether to ask them; ''without an estimate report only the current work and elapsed time.',@@ -158,6 +159,11 @@ RULES = ['otherwise make its reference pictures with POST /cast/{id}/generate. ''Before shooting, tell the user the casting in one sentence. Later shots use the card\'s current pictures. ''Keep cards you made so later films can reuse them.',+ 'Cut by default. Wherever the film can change shot, make the new shot from the reference pictures it should '+ 'have (cast cards, scene pictures) instead of continuing the previous video. Continue a video only for a '+ 'long take that cannot be cut, an action that must stay unbroken, or sound that must stay unbroken, and '+ 'avoid continuing again and again. One shot can be continued up to 5 parts, each at least 10 seconds, '+ 'about 70 seconds in all; do not fill that. Past it, cut to a new shot.',]@@ -177,6 +183,9 @@ RULES_ZH = ['先选角,再拍摄。先搜索角色库,使用用户指定的卡片或明确符合要求的卡片;拿不准时询问用户。''只有找不到合适卡片时才新建:用户提供了照片就用照片,否则通过 POST /cast/{id}/generate 生成参考图。''开拍前用一句话告诉用户选角结果。后续镜头使用卡片当前的图片。保留你创建的卡片,方便以后拍片复用。',+ '默认切镜头。能切镜头的地方就切:新镜头直接用该有的参考图(人物卡、场景图)生成,不接上一段视频。'+ '只有不能切镜的长镜头、必须连贯的动作、必须连贯的声音,才接着上一段生成,也不要一段接一段地连续接。'+ '一个镜头最多接到 5 段,每段不短于 10 秒,总长约 70 秒,不建议接满;超过就切镜头。',]diff --git a/freevideo_engine/agent/mcp.py b/freevideo_engine/agent/mcp.pyindex ea8dc38..a003728 100644--- a/freevideo_engine/agent/mcp.py+++ b/freevideo_engine/agent/mcp.py@@ -44,7 +44,7 @@ def tool_list():dict(name='freevideo_prompt_guide', title='H3 prompt guide', description='How to write prompts: sections, shots and cuts, camera motion, dialogue, references, short-drama tips, and a full example.', annotations=read),dict(name='freevideo_schema', title='Job JSON Schemas', description='JSON Schemas of every job kind and of image edit recipes.', annotations=read),dict(name='freevideo_plan', title='Plan a job', description="Validate a job without running it: actual frames/seconds, sampling passes, warnings and a time estimate from this computer's history. Same fields as freevideo_generate.", annotations=read),- dict(name='freevideo_generate', title='Generate a video', description='Generate one video with audio from an H3 prompt, up to 60 seconds: a longer request continues and joins into one file automatically. Use this for any single video, also after freevideo_enhance_prompt; the drama tools are only for a film of several separately chosen shots. Optionally from first/last frames or references (files, local paths, or frames of earlier jobs). Returns the job; with wait, its result (video path and URL).', annotations=write),+ dict(name='freevideo_generate', title='Generate a video', description='Generate one video with audio from an H3 prompt, up to 60 seconds: a longer request continues and joins into one file automatically. Prefer a new shot with its reference pictures wherever the film can cut; continue only a take, action or sound that must stay unbroken. Use this for any single video, also after freevideo_enhance_prompt; the drama tools are only for a film of several separately chosen shots. Optionally from first/last frames or references (files, local paths, or frames of earlier jobs). Returns the job; with wait, its result (video path and URL).', annotations=write),dict(name='freevideo_submit_batch', title='Submit a storyboard', description='Submit several jobs at once. Each row is a job object with "kind" (generate default, assemble, vlm, enhance) and an optional "ref"; later rows use "@ref" wherever a job id is expected (from_job, after, clips, refine_from). FreeVideo runs them in dependency order.', annotations=write),dict(name='freevideo_job', title='Job status', description='Status, progress, result or error of a job.', annotations=read),dict(name='freevideo_wait', title='Wait for a job or the user', description="Wait at most 25 seconds for a job (with id) or, without id, for the user's words in this session. Returns early when the user says something (inbox). If the job is still active, call it again with the same ID; do not resubmit.", annotations=read),