Skip to content

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.

Wording: replace the unclear 'FreeVideo tests' phrasing with 'tested' in setup and repair text

Bowen Xue3aa2435d15ff
diff --git a/freevideo_engine/bootstrap.py b/freevideo_engine/bootstrap.py
index 4c8134a..bf04997 100644
--- a/freevideo_engine/bootstrap.py
+++ b/freevideo_engine/bootstrap.py
@@ -531,7 +531,7 @@ def plan(args, *, local_progress=None, stage=None):
errors.append('Native Windows uses the unified environment. Rerun with --environment unified.')
layout = 'unified'
if windows_target and getattr(args, 'rebuild_sage', False):
- errors.append('Windows uses the Sage2 wheel FreeVideo tests; --rebuild-sage is a Linux source-build option.')
+ errors.append('Windows uses the tested Sage2 wheel; --rebuild-sage is a Linux source-build option.')
driver = str(snapshot['selected_gpu']['driver_version']).strip()
# An unreadable version is left to the later checks instead of blocking setup.
major = re.match(r'\d+', driver)
@@ -811,9 +811,9 @@ def plan(args, *, local_progress=None, stage=None):
'first_generation_note': 'Portable AdaLN tables follow model weights and the exact schedule, across GPUs and Torch/CUDA versions. New schedules or modulation-changing LoRAs require preparation. Existing environments stay on disk when switching.',
'estimates_are_not_capacity_guarantees': True, 'errors': errors, 'warnings': warnings,
'steps': ['Install isolated uv/Python and ' + ('one shared CUDA environment' if layout == 'unified' else 'two CUDA environments'),
- 'Clone the VDN, patched Diffusers and native H3 text-encoder library versions FreeVideo tests',
+ 'Clone the tested versions of VDN, patched Diffusers and the native H3 text-encoder library',
'Install verified Windows Triton / Sage2 wheels' if windows_target else 'Install a local CUDA compiler and build Sage2 for the detected GPU only',
- 'Download and verify the base, eight-step checkpoint and encoder weights FreeVideo tests',
+ 'Download and verify the tested base, eight-step checkpoint and encoder weights',
'Prepare or verify the architecture-compatible FP8 cache',
'Execute small attention/linear kernel probes and save machine configuration'],
'licenses': ['https://huggingface.co/OpenVDN/vdn-minimax-h3/blob/751739ee5b9e3ac802dca5d5111075fdaeb47885/LICENSE',
@@ -2272,15 +2272,15 @@ def run_main(argv, shown):
help='With --prepared-format: download and verify that model beside the installed one, which stays '
'in use. No other setup step runs and machine.json is unchanged')
parser.add_argument('--model-source', choices=('prepared', 'source'), default='prepared',
- help='Default: download a slim model FreeVideo tests matching the GPU format. source: explicitly download original weights and convert locally')
+ help='Default: download the tested slim model matching the GPU format. source: explicitly download original weights and convert locally')
parser.add_argument('--encoder-models', type=Path, help='Directory containing text_encoders/')
- parser.add_argument('--reuse-models', type=Path, help='Scan an existing model library, verify content against the version FreeVideo tests and import matching files; only missing models are downloaded')
+ parser.add_argument('--reuse-models', type=Path, help='Scan an existing model library, verify content against the tested version and import matching files; only missing models are downloaded')
parser.add_argument('--reuse-models-manifest', type=Path, help='Versioned JSON containing model library roots, including ComfyUI extra model paths')
parser.add_argument('--copy-existing-models', action='store_true', help='Copy verified local models instead of using same-disk hardlinks; source files are kept')
parser.add_argument('--allow-model-restart', action='store_true', help='Permit a fresh download when an incomplete model cannot resume; retains old data and may use extra download/disk space')
cache_options = parser.add_mutually_exclusive_group()
cache_options.add_argument('--cache', type=Path, help='Verify and reuse an existing FP8 cache')
- cache_options.add_argument('--rebuild-cache', action='store_true', help='Recreate FP8 from the source weights FreeVideo tests, downloading missing sources; preserve failed caches')
+ cache_options.add_argument('--rebuild-cache', action='store_true', help='Recreate FP8 from the tested source weights, downloading missing sources; preserve failed caches')
parser.add_argument('--storage', choices=('compact', 'retain'), help='Default compact: stream FP8 and remove installer-owned conversion sources after verification; retain keeps sources for re-quantization')
parser.add_argument('--frontend-root', type=Path, help=argparse.SUPPRESS)
parser.add_argument('--frontend-separate', action='store_true', help=argparse.SUPPRESS)
diff --git a/freevideo_engine/model_repair.py b/freevideo_engine/model_repair.py
index 8e3ef9f..c0b84f0 100644
--- a/freevideo_engine/model_repair.py
+++ b/freevideo_engine/model_repair.py
@@ -1,4 +1,4 @@
-"""Review and retain local files that differ from the model FreeVideo tests."""
+"""Review and retain local files that differ from the tested model."""
import ctypes
import errno
import json
@@ -16,15 +16,15 @@ from .long_paths import io_path, plain
# Shared copy for setup, the launcher and the browser plan.
REVIEW = (
- '{file} ({size}) differs from the version FreeVideo tests: it will be renamed to {keep_as} and kept, '
+ '{file} ({size}) differs from the tested version: it will be renamed to {keep_as} and kept, '
'and the tested version ({expected_size}) will be downloaded.',
'{file}({size})与 FreeVideo 测试过的版本不同:将改名为 {keep_as} 保留,'
'再下载测试过的版本({expected_size})。',
)
PROGRESS = ('Keeping the original file: ', '正在保留原文件:')
-FAILURE_TITLE = ('An existing model differs from the version FreeVideo tests', '已有模型与 FreeVideo 测试过的版本不同')
+FAILURE_TITLE = ('An existing model differs from the tested version', '已有模型与 FreeVideo 测试过的版本不同')
FAILURE = (
- '{file} differs from the version FreeVideo tests and was left unchanged. Click {entry} again: '
+ '{file} differs from the tested version and was left unchanged. Click {entry} again: '
'the confirmation page will list this file, and after you confirm, it will be renamed and kept, '
'and the tested version will be downloaded.',
'{file} 与 FreeVideo 测试过的版本不同,这次没有改动它。请再次点击“{entry}”:'
@@ -35,7 +35,7 @@ MISMATCH = 'MODEL_MISMATCH_REPAIR'
MISMATCH_PATH = 'MODEL_MISMATCH_PATH'
# Where the launcher offers no Repair: the first sentence of FAILURE, then one way out
# for each reason (failure_details.WITHOUT_REPAIR).
-UNCHANGED = ('{file} differs from the version FreeVideo tests and was left unchanged.',
+UNCHANGED = ('{file} differs from the tested version and was left unchanged.',
'{file} 与 FreeVideo 测试过的版本不同,这次没有改动它。')
BUSY = ('When the current task ends, a Repair button appears here; after you click it and confirm, '
'FreeVideo renames and keeps this file, then downloads the tested version.',
@@ -44,7 +44,7 @@ OFFLINE = ('This offline-package installation cannot switch back in one step yet
'then import the Prompt model package (FreeVideo-文本编码器.zip) again on the Installation page.',
'离线包安装暂时不能一键换回。请把下面这个文件移走或改名,再在“安装”页重新导入“提示词模型(FreeVideo-文本编码器.zip)”包。')
FIRST_INSTALL = (
- '{file} differs from the version FreeVideo tests and was left unchanged. Click Retry installation below: '
+ '{file} differs from the tested version and was left unchanged. Click Retry installation below: '
'the confirmation page will list this file, and after you confirm, it is renamed and kept, '
'and the tested version is downloaded.',
'{file} 与 FreeVideo 测试过的版本不同,这次没有改动它。请点击下方的“重试安装”:'