Intial commit

This commit is contained in:
2026-02-28 09:23:11 +01:00
commit 72fe4e04d0
5 changed files with 645 additions and 0 deletions

65
.gitignore vendored Normal file
View File

@@ -0,0 +1,65 @@
# Python
__pycache__/
*.py[cod]
*$py.class
*.so
.Python
.pytest_cache/
.mypy_cache/
.ruff_cache/
# Virtual environments
.venv/
venv/
env/
ENV/
# IDE / editor
.vscode/
.idea/
*.swp
*.swo
# OS files
.DS_Store
Thumbs.db
# Build / packaging
build/
dist/
*.egg-info/
# Logs
*.log
# Project runtime temp/output
*_tmp/
frames_in/
frames_out/
/tmp/
# Video and media artifacts
*.mp4
*.mkv
*.avi
*.mov
*.m4a
*.wav
*.png
*.jpg
*.jpeg
# Keep model and tool binaries out of git
models/
*.bin
*.param
*.pth
# Local env/config
.env
.env.*
# Downloaded archives
*.zip
*.tar
*.tar.gz

3
.vscode/settings.json vendored Normal file
View File

@@ -0,0 +1,3 @@
{
"python-envs.defaultEnvManager": "ms-python.python:pyenv"
}

199
README.md Normal file
View File

@@ -0,0 +1,199 @@
# Local AI Video Upscaling on Ubuntu (RTX 4070 Super)
This project uses:
- `ffmpeg` / `ffprobe` for demux/remux and frame handling
- `realesrgan-ncnn-vulkan` for GPU upscaling (Vulkan backend, works well on NVIDIA)
- `upscale_video.py` as Python controller/orchestrator
## 1) System setup (Ubuntu)
### NVIDIA driver + Vulkan check
```bash
nvidia-smi
vulkaninfo | head
```
If missing tools:
```bash
sudo apt update
sudo apt install -y ffmpeg vulkan-tools mesa-vulkan-drivers unzip wget
```
Install/update NVIDIA driver with Ubuntu tooling if needed:
```bash
sudo ubuntu-drivers autoinstall
sudo reboot
```
## 2) Install Real-ESRGAN binary and models
### a) Download and install binary
Download `realesrgan-ncnn-vulkan` Linux release from the official Real-ESRGAN releases page,
extract it, and add the binary folder to your `PATH`.
Example:
```bash
mkdir -p ~/tools && cd ~/tools
# Replace URL with latest Linux release zip from official Real-ESRGAN releases
wget <REAL_ESRGAN_LINUX_RELEASE_ZIP_URL> -O realesrgan.zip
unzip realesrgan.zip -d realesrgan
REAL_ESRGAN_DIR="$(find "$HOME/tools/realesrgan" -maxdepth 2 -type f -name realesrgan-ncnn-vulkan -printf '%h\n' | head -n 1)"
echo "Found binary dir: $REAL_ESRGAN_DIR"
echo "export PATH=\"$REAL_ESRGAN_DIR:\$PATH\"" >> ~/.bashrc
source ~/.bashrc
command -v realesrgan-ncnn-vulkan
realesrgan-ncnn-vulkan -h
```
If the command is still not found, test directly with absolute path:
```bash
find "$HOME/tools/realesrgan" -maxdepth 3 -type f -name realesrgan-ncnn-vulkan
```
### b) Download model files
The binary needs model files to work. Download them from the official Real-ESRGAN repository:
```bash
cd ~/tools/realesrgan
mkdir -p models
cd models
# Download all common models
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-animevideov3.pth
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth
# For ncnn-vulkan, you need the converted .param and .bin files
# Download pre-converted models:
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-animevideov3-x2.pth -O realesr-animevideov3-x2.pth
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-animevideov3-x3.pth -O realesr-animevideov3-x3.pth
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesr-animevideov3-x4.pth -O realesr-animevideov3-x4.pth
```
**Important**: The ncnn-vulkan version needs specific model formats. If models still fail, download the complete model pack:
```bash
cd ~/tools/realesrgan
wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.5.0/realesrgan-ncnn-vulkan-20220424-ubuntu.zip
unzip -j realesrgan-ncnn-vulkan-20220424-ubuntu.zip "realesrgan-ncnn-vulkan/models/*" -d models/
```
## 3) Install Python dependencies (optional but recommended)
For a cleaner progress bar during upscaling:
```bash
pip install tqdm
# or
pip install -r requirements.txt
```
The script works without tqdm but will show a nicer single-line progress bar if it's installed.
## 4) Run the Python controller
From this project directory:
```bash
python3 upscale_video.py \
-i input.mp4 \
-o output_upscaled.mp4 \
--model realesr-animevideov3 \
--model-path ~/tools/realesrgan/models \
--scale 2
```
**Important**: If you get "find_blob_index_by_name" errors, the model files are missing.
Use `--model-path` to point to your models directory (see section 2b above).
By default, temporary working files are created on `/mnt/winsteam`.
Override if needed with `--temp-root /some/other/path`.
By default, GPU selection uses `--gpu-id auto`.
To force a specific Vulkan GPU, pass e.g. `--gpu-id 0`.
### Useful options
- `--model realesr-animevideov3` for animation/anime-like sources
- `--model realesrgan-x4plus` for natural/live-action footage
- `--model-path ~/tools/realesrgan/models` to specify model directory
- `--scale 2|3|4`
- `--tile-size 128` (or 256) if you hit VRAM limits
- `--jobs 2:2:2` to tune throughput
- `--crf 14` for higher output quality (bigger file)
- `--keep-temp` to keep extracted and processed frame directories
- `--temp-root /mnt/winsteam` for temp workspace location
- `--gpu-id auto` (or `--gpu-id 0`, `--gpu-id 1`, etc.)
- `--test-seconds 60` to process only first N seconds for validation
- `--pre-vf "hqdn3d=1.5:1.5:6:6"` to denoise/deblock before upscaling
During upscaling, the script prints live status every ~2 seconds:
- processed/total frames
- percentage
- current average fps
- ETA (remaining time)
**Note**: Audio is automatically re-encoded to AAC 192kbps for maximum compatibility.
**Progress display**: Real-ESRGAN's verbose output is suppressed. The script shows clean progress with tqdm (if installed) or simple periodic updates otherwise.
**Aspect ratio handling**: Input frames are normalized to square pixels before upscaling (`SAR=1`).
For anamorphic sources (for example 720x576 PAL 16:9), this avoids “squeezed” frame geometry.
The conversion uses non-cropping width expansion, so source frame content is preserved.
## 5) Typical tuning for RTX 4070 Super
Start with:
- `--scale 2`
- `--jobs 2:2:2`
- `--tile-size 0`
If you see memory errors, lower memory pressure using:
- `--tile-size 128`
- `6-jobs 1:2:2`
## 5) Optional quality upgrades
For best final quality, you can output with HEVC:
```bash
python3 upscale_video.py -i input.mp4 -o output_hevc.mp4 --codec libx265 --crf 18
```
## 7) GPU ID mapping (`nvidia-smi` vs Vulkan `-g`)
Real-ESRGAN uses Vulkan GPU IDs (`-g`), which may not match `nvidia-smi` index order.
Check NVIDIA GPUs:
```bash
nvidia-smi --query-gpu=index,name,uuid,pci.bus_id --format=csv,noheader
```
Check Vulkan devices:
```bash
vulkaninfo --summary
```
Match by GPU name or UUID:
- `nvidia-smi` UUID format: `GPU-xxxxxxxx-...`
- `vulkaninfo` UUID format: `xxxxxxxx-...` (same value without `GPU-` prefix)
Example from this setup:
- Vulkan `GPU0` = `NVIDIA GeForce RTX 4070 SUPER`
- `nvidia-smi` index `1` = `NVIDIA GeForce RTX 4070 SUPER`
So for the RTX 4070 SUPER here, use Vulkan ID `-g 0`.
With this script, that is:
```bash
python3 upscale_video.py -i input.mp4 -o output.mp4 --gpu-id 0
```
Quick test run example (first 60 seconds only):
```bash
python3 upscale_video.py -i input.mp4 -o output_test.mp4 --model realesrgan-x4plus --model-path ~/tools/realesrgan/models --scale 2 --gpu-id 0 --test-seconds 60
```
---
If you want, this can be extended with:
- batch-folder processing
- automatic model selection by content type
- optional frame interpolation (RIFE) for smoother motion

1
requirements.txt Normal file
View File

@@ -0,0 +1 @@
tqdm>=4.66.0

377
upscale_video.py Normal file
View File

@@ -0,0 +1,377 @@
#!/usr/bin/env python3
import argparse
import shutil
import subprocess
import sys
import tempfile
import time
from pathlib import Path
try:
from tqdm import tqdm
HAS_TQDM = True
except ImportError:
HAS_TQDM = False
def run(cmd: list[str]) -> None:
print("\n$", " ".join(cmd))
completed = subprocess.run(cmd)
if completed.returncode != 0:
raise RuntimeError(f"Command failed ({completed.returncode}): {' '.join(cmd)}")
def command_exists(name: str) -> bool:
return shutil.which(name) is not None
def assert_prerequisites(realesrgan_bin: str) -> None:
missing = []
if not command_exists("ffmpeg"):
missing.append("ffmpeg")
if not command_exists("ffprobe"):
missing.append("ffprobe")
real_esrgan_ok = Path(realesrgan_bin).exists() or command_exists(realesrgan_bin)
if not real_esrgan_ok:
missing.append(realesrgan_bin)
if missing:
items = ", ".join(missing)
raise RuntimeError(
f"Missing required tools: {items}.\n"
"Install them first (see README.md)."
)
def has_audio_stream(input_video: Path) -> bool:
cmd = [
"ffprobe",
"-v",
"error",
"-select_streams",
"a:0",
"-show_entries",
"stream=codec_type",
"-of",
"csv=p=0",
str(input_video),
]
result = subprocess.run(cmd, capture_output=True, text=True)
return result.returncode == 0 and "audio" in result.stdout
def count_png_frames(folder: Path) -> int:
return sum(1 for _ in folder.glob("*.png"))
def run_upscale_with_progress(cmd: list[str], input_frames: Path, output_frames: Path) -> None:
total_frames = count_png_frames(input_frames)
if total_frames == 0:
raise RuntimeError("No extracted frames found before upscaling.")
started = time.time()
# Suppress Real-ESRGAN's verbose output by redirecting stdout/stderr
process = subprocess.Popen(cmd, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
if HAS_TQDM:
pbar = tqdm(total=total_frames, unit="frames", desc="Upscaling",
bar_format="{l_bar}{bar}| {n_fmt}/{total_fmt} [{elapsed}<{remaining}, {rate_fmt}]")
last_count = 0
else:
print(f"Upscaling: 0/{total_frames} frames (0.0%) | ETA --:--:--")
last_print = 0.0
while True:
return_code = process.poll()
now = time.time()
done_frames = count_png_frames(output_frames)
if HAS_TQDM:
delta = done_frames - last_count
if delta > 0:
pbar.update(delta)
last_count = done_frames
else:
if now - last_print >= 2.0:
progress = min(100.0, (done_frames / total_frames) * 100)
elapsed = max(now - started, 1e-6)
fps = done_frames / elapsed
if done_frames > 0 and fps > 0:
remaining_frames = max(total_frames - done_frames, 0)
eta_seconds = int(remaining_frames / fps)
eta_h, rem = divmod(eta_seconds, 3600)
eta_m, eta_s = divmod(rem, 60)
eta_str = f"{eta_h:02d}:{eta_m:02d}:{eta_s:02d}"
else:
eta_str = "--:--:--"
print(
f"Upscaling: {done_frames}/{total_frames} "
f"({progress:.1f}%) | {fps:.2f} fps | ETA {eta_str}"
)
last_print = now
if return_code is not None:
done_frames = count_png_frames(output_frames)
if HAS_TQDM:
delta = done_frames - last_count
if delta > 0:
pbar.update(delta)
pbar.close()
elapsed = max(time.time() - started, 1e-6)
fps = done_frames / elapsed
print(
f"Upscaling complete: {done_frames}/{total_frames} frames | "
f"avg {fps:.2f} fps | total time {elapsed:.1f}s"
)
if return_code != 0:
raise RuntimeError(f"Command failed ({return_code}): {' '.join(cmd)}")
break
time.sleep(0.2)
def upscale_video(
input_video: Path,
output_video: Path,
realesrgan_bin: str,
model: str,
model_path: str | None,
scale: int,
tile_size: int,
jobs: str,
fps: str | None,
codec: str,
crf: int,
preset: str,
keep_temp: bool,
temp_root: Path,
gpu_id: str,
test_seconds: float | None,
pre_vf: str | None,
) -> None:
if not input_video.exists():
raise FileNotFoundError(f"Input video does not exist: {input_video}")
output_video.parent.mkdir(parents=True, exist_ok=True)
temp_root.mkdir(parents=True, exist_ok=True)
with tempfile.TemporaryDirectory(prefix="video_upscale_", dir=str(temp_root)) as tmp_dir_str:
tmp_dir = Path(tmp_dir_str)
frames_in = tmp_dir / "frames_in"
frames_out = tmp_dir / "frames_out"
audio_file = tmp_dir / "audio.m4a"
frames_in.mkdir(parents=True, exist_ok=True)
frames_out.mkdir(parents=True, exist_ok=True)
print(f"Working directory: {tmp_dir}")
if test_seconds is not None:
print(f"Test mode: processing first {test_seconds:.2f} seconds")
test_duration_args = ["-t", str(test_seconds)] if test_seconds is not None else []
filter_chain = []
if pre_vf:
filter_chain.append(pre_vf)
filter_chain.extend([
"scale=ceil(iw*sar/2)*2:ih",
"setsar=1",
])
extract_cmd = [
"ffmpeg",
"-y",
"-i",
str(input_video),
*test_duration_args,
"-vf",
",".join(filter_chain),
str(frames_in / "%08d.png"),
]
run(extract_cmd)
audio_present = has_audio_stream(input_video)
if audio_present:
extract_audio_cmd = [
"ffmpeg",
"-y",
"-i",
str(input_video),
*test_duration_args,
"-vn",
"-c:a",
"aac",
"-b:a",
"192k",
str(audio_file),
]
run(extract_audio_cmd)
upscale_cmd = [
realesrgan_bin,
"-i",
str(frames_in),
"-o",
str(frames_out),
"-n",
model,
"-s",
str(scale),
"-f",
"png",
"-t",
str(tile_size),
"-j",
jobs,
"-g",
gpu_id,
]
if model_path:
upscale_cmd.extend(["-m", model_path])
run_upscale_with_progress(upscale_cmd, frames_in, frames_out)
fps_args = ["-r", fps] if fps else []
encode_cmd = [
"ffmpeg",
"-y",
*fps_args,
"-i",
str(frames_out / "%08d.png"),
]
if audio_present and audio_file.exists():
encode_cmd.extend(["-i", str(audio_file), "-c:a", "copy"])
encode_cmd.extend(
[
"-c:v",
codec,
"-crf",
str(crf),
"-preset",
preset,
"-pix_fmt",
"yuv420p",
str(output_video),
]
)
run(encode_cmd)
if keep_temp:
kept = output_video.parent / f"{output_video.stem}_tmp"
if kept.exists():
shutil.rmtree(kept)
shutil.copytree(tmp_dir, kept)
print(f"Temporary files copied to: {kept}")
def parse_args() -> argparse.Namespace:
parser = argparse.ArgumentParser(
description="Upscale a video locally with Real-ESRGAN (RTX GPU via Vulkan)."
)
parser.add_argument("-i", "--input", required=True, help="Input video path")
parser.add_argument("-o", "--output", required=True, help="Output video path")
parser.add_argument(
"--realesrgan-bin",
default="realesrgan-ncnn-vulkan",
help="Path or command name of realesrgan-ncnn-vulkan",
)
parser.add_argument(
"--model",
default="realesr-animevideov3",
help="Model name (e.g. realesr-animevideov3, realesrgan-x4plus)",
)
parser.add_argument(
"--model-path",
default=None,
help="Path to models directory (required if models not in default location)",
)
parser.add_argument("--scale", type=int, default=2, choices=[2, 3, 4])
parser.add_argument(
"--tile-size",
type=int,
default=0,
help="Tile size for VRAM-limited cases (0 = auto)",
)
parser.add_argument(
"--jobs",
default="2:2:2",
help="NCNN worker threads as load:proc:save",
)
parser.add_argument(
"--fps",
default=None,
help="Override output FPS (default: keep source timing)",
)
parser.add_argument("--codec", default="libx264", help="Output video codec")
parser.add_argument("--crf", type=int, default=16, help="Quality (lower = better)")
parser.add_argument("--preset", default="medium", help="Encoder preset")
parser.add_argument(
"--keep-temp",
action="store_true",
help="Keep extracted/upscaled frame files next to output",
)
parser.add_argument(
"--temp-root",
default="/mnt/winsteam",
help="Directory used for temporary working files (default: /mnt/winsteam)",
)
parser.add_argument(
"--gpu-id",
default="auto",
help="Vulkan GPU id for Real-ESRGAN (e.g. 0, 1, 0,1). Use 'auto' by default",
)
parser.add_argument(
"--test-seconds",
type=float,
default=None,
help="Only process first N seconds (for quick test runs)",
)
parser.add_argument(
"--pre-vf",
default=None,
help="Optional ffmpeg video filter(s) applied before upscaling (e.g. hqdn3d=1.5:1.5:6:6)",
)
return parser.parse_args()
def main() -> int:
args = parse_args()
try:
assert_prerequisites(args.realesrgan_bin)
upscale_video(
input_video=Path(args.input),
output_video=Path(args.output),
realesrgan_bin=args.realesrgan_bin,
model=args.model,
model_path=args.model_path,
scale=args.scale,
tile_size=args.tile_size,
jobs=args.jobs,
fps=args.fps,
codec=args.codec,
crf=args.crf,
preset=args.preset,
keep_temp=args.keep_temp,
temp_root=Path(args.temp_root),
gpu_id=args.gpu_id,
test_seconds=args.test_seconds,
pre_vf=args.pre_vf,
)
except Exception as exc:
print(f"Error: {exc}", file=sys.stderr)
return 1
print("\nDone.")
return 0
if __name__ == "__main__":
raise SystemExit(main())