> ## Documentation Index
> Fetch the complete documentation index at: https://mintlify.com/xinntao/Real-ESRGAN/llms.txt
> Use this file to discover all available pages before exploring further.

# Inference Scripts

> Command-line interfaces for image and video upsampling

## inference\_realesrgan.py

Command-line script for upsampling images using Real-ESRGAN models.

### Usage

```bash theme={null}
python inference_realesrgan.py -i inputs -o results -n RealESRGAN_x4plus
```

### Arguments

<ParamField path="-i, --input" type="str" default="inputs">
  Input image or folder path
</ParamField>

<ParamField path="-n, --model_name" type="str" default="RealESRGAN_x4plus">
  Model name to use for inference

  **Available models:**

  * `RealESRGAN_x4plus` - General 4x upsampling
  * `RealESRNet_x4plus` - 4x upsampling without GAN
  * `RealESRGAN_x4plus_anime_6B` - Anime images 4x
  * `RealESRGAN_x2plus` - General 2x upsampling
  * `realesr-animevideov3` - Anime video 4x
  * `realesr-general-x4v3` - General purpose 4x with denoise control
</ParamField>

<ParamField path="-o, --output" type="str" default="results">
  Output folder path
</ParamField>

<ParamField path="-dn, --denoise_strength" type="float" default="0.5">
  Denoise strength. Range: 0 (weak denoise, keep noise) to 1 (strong denoise ability)

  Only used for the `realesr-general-x4v3` model
</ParamField>

<ParamField path="-s, --outscale" type="float" default="4">
  The final upsampling scale of the image
</ParamField>

<ParamField path="--model_path" type="str" default="None">
  Optional custom model path. Usually not needed as models are downloaded automatically
</ParamField>

<ParamField path="--suffix" type="str" default="out">
  Suffix of the restored image filename
</ParamField>

<ParamField path="-t, --tile" type="int" default="0">
  Tile size for processing. 0 means no tiling. Use tiling to avoid out-of-memory errors with large images
</ParamField>

<ParamField path="--tile_pad" type="int" default="10">
  Tile padding size to reduce border artifacts
</ParamField>

<ParamField path="--pre_pad" type="int" default="0">
  Pre-padding size at each border
</ParamField>

<ParamField path="--face_enhance" type="bool">
  Use GFPGAN to enhance faces in the image (flag, no value needed)
</ParamField>

<ParamField path="--fp32" type="bool">
  Use fp32 (full) precision during inference. Default is fp16 (half precision)
</ParamField>

<ParamField path="--alpha_upsampler" type="str" default="realesrgan">
  The upsampler for alpha channels (transparency)

  **Options:** `realesrgan` | `bicubic`
</ParamField>

<ParamField path="--ext" type="str" default="auto">
  Output image extension

  **Options:** `auto` | `jpg` | `png`

  `auto` uses the same extension as the input
</ParamField>

<ParamField path="-g, --gpu-id" type="int" default="None">
  GPU device to use. Can be 0, 1, 2, etc. for multi-GPU systems
</ParamField>

### Examples

<CodeGroup>
  ```bash Basic Usage theme={null}
  # Upscale a single image
  python inference_realesrgan.py -i input.jpg -o output_folder -n RealESRGAN_x4plus
  ```

  ```bash Batch Processing theme={null}
  # Upscale all images in a folder
  python inference_realesrgan.py -i inputs/ -o results/ -n RealESRGAN_x4plus -s 4
  ```

  ```bash Anime Images theme={null}
  # Use anime-specific model
  python inference_realesrgan.py -i anime.png -o results/ -n RealESRGAN_x4plus_anime_6B
  ```

  ```bash With Tiling theme={null}
  # Process large images with tiling to avoid OOM
  python inference_realesrgan.py -i large_image.png -o results/ -t 400 --tile_pad 10
  ```

  ```bash Face Enhancement theme={null}
  # Enhance faces using GFPGAN
  python inference_realesrgan.py -i portrait.jpg -o results/ -n RealESRGAN_x4plus --face_enhance
  ```

  ```bash Custom Scale theme={null}
  # Upscale to 2x instead of 4x
  python inference_realesrgan.py -i input.jpg -o results/ -n RealESRGAN_x4plus -s 2
  ```

  ```bash Denoise Control theme={null}
  # Use general model with custom denoise strength
  python inference_realesrgan.py -i noisy.jpg -o results/ -n realesr-general-x4v3 -dn 0.8
  ```

  ```bash Specify GPU theme={null}
  # Use specific GPU device
  python inference_realesrgan.py -i input.jpg -o results/ -g 1
  ```
</CodeGroup>

<Note>
  **Memory Management:**
  If you encounter CUDA out of memory errors, try using the `--tile` option with a smaller tile size (e.g., `-t 200` or `-t 400`). Tiling processes the image in smaller chunks at the cost of slightly slower processing.
</Note>

<Note>
  **File Formats:**
  RGBA images (with transparency) are automatically saved as PNG regardless of the `--ext` setting to preserve the alpha channel.
</Note>

## inference\_realesrgan\_video.py

Command-line script for upsampling videos using Real-ESRGAN models. Optimized for anime videos.

### Usage

```bash theme={null}
python inference_realesrgan_video.py -i video.mp4 -o results -n realesr-animevideov3
```

### Arguments

<ParamField path="-i, --input" type="str" default="inputs">
  Input video, image, or folder path
</ParamField>

<ParamField path="-n, --model_name" type="str" default="realesr-animevideov3">
  Model name to use for inference

  **Available models:**

  * `realesr-animevideov3` - Optimized for anime videos (default)
  * `RealESRGAN_x4plus_anime_6B` - Anime images/videos 4x
  * `RealESRGAN_x4plus` - General 4x upsampling
  * `RealESRNet_x4plus` - 4x upsampling without GAN
  * `RealESRGAN_x2plus` - General 2x upsampling
  * `realesr-general-x4v3` - General purpose 4x with denoise control
</ParamField>

<ParamField path="-o, --output" type="str" default="results">
  Output folder path
</ParamField>

<ParamField path="-dn, --denoise_strength" type="float" default="0.5">
  Denoise strength. Range: 0 (weak denoise, keep noise) to 1 (strong denoise ability)

  Only used for the `realesr-general-x4v3` model
</ParamField>

<ParamField path="-s, --outscale" type="float" default="4">
  The final upsampling scale of the video
</ParamField>

<ParamField path="--suffix" type="str" default="out">
  Suffix of the restored video filename
</ParamField>

<ParamField path="-t, --tile" type="int" default="0">
  Tile size for processing. 0 means no tiling. Use tiling to avoid out-of-memory errors
</ParamField>

<ParamField path="--tile_pad" type="int" default="10">
  Tile padding size to reduce border artifacts
</ParamField>

<ParamField path="--pre_pad" type="int" default="0">
  Pre-padding size at each border
</ParamField>

<ParamField path="--face_enhance" type="bool">
  Use GFPGAN to enhance faces in the video (flag, no value needed)

  Note: Automatically disabled for anime models
</ParamField>

<ParamField path="--fp32" type="bool">
  Use fp32 (full) precision during inference. Default is fp16 (half precision)
</ParamField>

<ParamField path="--fps" type="float" default="None">
  FPS of the output video. If not specified, uses the input video's FPS
</ParamField>

<ParamField path="--ffmpeg_bin" type="str" default="ffmpeg">
  Path to the ffmpeg binary
</ParamField>

<ParamField path="--extract_frame_first" type="bool">
  Extract frames to disk before processing (flag, no value needed). Can be useful for certain workflows
</ParamField>

<ParamField path="--num_process_per_gpu" type="int" default="1">
  Number of processes to spawn per GPU for parallel processing
</ParamField>

<ParamField path="--alpha_upsampler" type="str" default="realesrgan">
  The upsampler for alpha channels (transparency)

  **Options:** `realesrgan` | `bicubic`
</ParamField>

<ParamField path="--ext" type="str" default="auto">
  Image extension when processing image folders

  **Options:** `auto` | `jpg` | `png`
</ParamField>

### Examples

<CodeGroup>
  ```bash Basic Video Upscaling theme={null}
  # Upscale anime video with default settings
  python inference_realesrgan_video.py -i anime_video.mp4 -o results/
  ```

  ```bash Custom Model theme={null}
  # Use RealESRGAN_x4plus for general videos
  python inference_realesrgan_video.py -i video.mp4 -o results/ -n RealESRGAN_x4plus
  ```

  ```bash With Tiling theme={null}
  # Process with tiling to manage memory
  python inference_realesrgan_video.py -i video.mp4 -o results/ -t 400
  ```

  ```bash Custom FPS theme={null}
  # Set output FPS
  python inference_realesrgan_video.py -i video.mp4 -o results/ --fps 60
  ```

  ```bash Multi-GPU Processing theme={null}
  # Use 2 processes per GPU for faster processing
  python inference_realesrgan_video.py -i video.mp4 -o results/ --num_process_per_gpu 2
  ```

  ```bash Extract Frames First theme={null}
  # Extract frames before processing
  python inference_realesrgan_video.py -i video.mp4 -o results/ --extract_frame_first
  ```

  ```bash Custom Scale theme={null}
  # Upscale to 2x instead of 4x
  python inference_realesrgan_video.py -i video.mp4 -o results/ -s 2
  ```

  ```bash Full Precision theme={null}
  # Use FP32 for maximum quality
  python inference_realesrgan_video.py -i video.mp4 -o results/ --fp32
  ```
</CodeGroup>

<Note>
  **Performance Warning:**
  If you are generating videos larger than 4K resolution, processing will be very slow due to I/O speed limitations. It is highly recommended to decrease the `--outscale` value.
</Note>

<Note>
  **Multi-GPU Processing:**
  The script automatically detects available GPUs and can process video segments in parallel. Use `--num_process_per_gpu` to control how many processes run per GPU. For example, with 2 GPUs and `--num_process_per_gpu 2`, a total of 4 processes will run in parallel.
</Note>

<Note>
  **FLV Format:**
  If the input is a `.flv` file, the script automatically converts it to `.mp4` using ffmpeg before processing.
</Note>

<Note>
  **Audio Preservation:**
  The script automatically preserves the audio track from the input video in the output video.
</Note>

### Helper Classes

The video inference script includes two helper classes:

#### Reader

Handles reading frames from videos, images, or folders. Supports streaming from video files using ffmpeg.

**Methods:**

* `get_resolution()` - Returns (height, width) of the input
* `get_fps()` - Returns the FPS of the video
* `get_audio()` - Returns the audio stream
* `get_frame()` - Returns the next frame
* `close()` - Closes the stream reader

#### Writer

Handles writing frames to output video using ffmpeg with H.264 encoding.

**Methods:**

* `write_frame(frame)` - Writes a frame to the output video
* `close()` - Finalizes and closes the output video