THP (File Format)
THP files are video files used in many Wii and GameCube games.
File Format
Header
The file starts with a header.
| Offset | Type | Description |
|---|---|---|
| 0x00 | Char[4] | File magic. Always THP\0 in ASCII (0x54485000). |
| 0x04 | UInt32 | Version number. Mario Kart Wii uses 0x11000. |
| 0x08 | UInt32 | Max buffer size. Maximum buffer size needed for one frame, which includes header, video and audio data. |
| 0x0C | UInt32 | Max audio samples. Maximum number of samples used in one frame. |
| 0x10 | Float | Frames per second. Frame rates between 1.0 and 59.94 are supported. |
| 0x14 | UInt32 | Number of frames in this file. |
| 0x18 | UInt32 | Length of the first frame in bytes. |
| 0x1C | UInt32 | Length of all frames in bytes. |
| 0x20 | UInt32 | Component Container offset. |
| 0x24 | UInt32 | Frame table offset. Only used if a THP file is intended to not start on the first frame. |
| 0x28 | UInt32 | First frame offset. |
| 0x2C | UInt32 | Last frame offset. |
Component Container
The component container specifies what component the file contains for each frame. Following this structure, comes info structures which specifies data for each component. It has the following structure:
| Offset | Type | Description | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 0x00 | UInt32 | Number of used components. | ||||||||
| 0x04 | SByte[16] | Component types.
|
Video Info
The video info structure specifies info about the video.
| Offset | Type | Description | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| 0x00 | UInt32 | Width of the video in pixels. | ||||||||
| 0x04 | UInt32 | Height of the video in pixels. | ||||||||
| If version ≥ 0x11000 | ||||||||||
| 0x08 | UInt32 | Video format.
| ||||||||
Audio Info
The audio info structure specifies info about the audio.
| Offset | Type | Description |
|---|---|---|
| 0x00 | UInt32 | Number of audio channels. |
| 0x04 | UInt32 | Sample rate in hertz. |
| 0x08 | UInt32 | Number of samples. |
| If version ≥ 0x11000 | ||
| 0x0C | UInt32 | Number of tracks. Allows for more than 2 audio channels in a THP file. |
Frame
A frame consist of a frame header, video data and optionally audio data, defined by the component container. The size of a frame must be a multiple of 32. A frame starts with the following structure:
| Offset | Type | Description |
|---|---|---|
| 0x00 | UInt32 | Next frame size. For the last frame this value is set to first frame size. |
| 0x04 | UInt32 | Previous frame size. For the first frame this value is set to last frame size. |
After the data for each component, which is specified in the component container, is stored. First an UInt32 is stored for each component, which is the size of the component data. The data is padded by 4. Afterwards the actual data is stored.
Video Data
The video data structure contains the actual still image. The size of the image is specified above, but since the video data is padded to be a multiple of 4, it may contain padding. The video data will always end with FF D9 (EndOfImage).
The image data is basically a JPEG. An JPEG consist of different markers, which always start with a FF byte, followed by a non-zero byte. Following the 2-byte marker ID is some data, which differs between marker. Some markers will store no data, some will store a 2-byte length, which specifies the length of the data which follows (including the length bytes), and some markers store other data. JPEG data will always start with a StartOfImage (SOI) marker and end with an EndOfImage (EOI) marker. It will also contain a StartOfScan (SOS) marker. The JPEG specification says that if the value FF needs to be stored, it needs to be stored as FF 00. This is not to confuse the byte with a marker. However, in Nintendo's implementation, this is not the case within the StartOfScan marker's data. In order for other tools to decode the JPEG, all FF needs to be replaced with FF 00 in the StartOfScan's data.
| Offset | Type | Description |
|---|---|---|
| 0x00 | Byte[S] | Image data, where S is the non-padded size of the image. |
Audio Data
The audio data structure contains the actual audio data. The audio data is stored as DSP ADPCM data.
If version ≥ 0x11000, multiple audio datas are allowed to be stored. This allows for more than 2 channels to be stored in a file. The number of audio data structures are defined by number of tracks. The size of the audio data structure stored in the frame header stores the size of one and not the total size.
| Offset | Type | Description |
|---|---|---|
| 0x00 | UInt32 | Channel size in bytes. Specifies the size of each audio channel. |
| 0x04 | UInt32 | Number of samples. |
| 0x08 | Int16[8][2] | Channel 1 coefficients. |
| 0x28 | Int16[8][2] | Channel 2 coefficients. |
| 0x48 | Int16 | Channel 1 yn 1. |
| 0x4A | Int16 | Channel 1 yn 2. |
| 0x4C | Int16 | Channel 2 yn 1. |
| 0x4E | Int16 | Channel 2 yn 2. |
The raw audio data follows.
| Offset | Type | Description |
|---|---|---|
| 0x00 | Byte[N][S] | Audio channel data, where N is the number of channels and S is the size of each channel. |
Tools
The following tools can handle THP files:
- BrawlBox, by Kryal & BlackJax (only playable)
- Wexos's Toolbox, by Wexos (only playable)
