API Parameters
wz_encoder_config_t
- Reopen encoder: if updated, these parameters will cause the encoder to reopen itself.
- No: not safe or no effect
- Yes: safe to update
Basic configuration parameters are preset (fast, faster, etc.) and tune (camera/screen/adaptive). Available values are specifid in wzav1enc.h or in Presets and Tuning Options.
Recommended values for low complexity RTC applications are between superfast and ultrafast.
| type | name | Description | Safe for update config |
|---|---|---|---|
| const char * | preset | codec preset for speed level | Reopen encoder |
| const char * | tune | codec tune for scene | Reopen encoder |
| const char * | rc_type | rate control method type. | Reopen encoder |
| int32_t | adaptive_vq_bias | bias towards higher quality high motion/camera frames when using tune=adaptive [0-9] | No |
| int32_t | width | image origin width | Reopen encoder |
| int32_t | width_max | image maximum width | Reopen encoder |
| int32_t | height | image origin height | Reopen encoder |
| int32_t | height_max | image maximum height | Reopen encoder |
| int32_t | threads | encoder threads number | Reopen encoder |
| int32_t | frame_rate_num | frame rate numerator | Yes |
| int32_t | frame_rate_den | frame rate denominator | Yes |
| int32_t | time_base_num | time base numerator, default should be 1 | Yes |
| int32_t | time_base_den | time base denominator, time scale of pts | Yes |
| int32_t | bitratekbps | bit rate kbps per second, valid when rc type is "vbr", "cbr" | Yes |
| int32_t | vbv_buffer_size | buffer size of vbv, 0 means vbv not enabled,default 0 | Yes |
| int32_t | vbv_max_rate | max rate of vbv, 0 means vbv not enabled,default 0 | Yes |
| int32_t | frame_kbits_limit | frame max kbits limit for rc, if qpmax is also set,the actual frame bit may exceed this | Yes |
| int32_t | max_intra_kbits | maximum allowed bitrate(kbps) for any intra frame, default 0 | Yes |
| int32_t | crf | crf for video quality, only valid when rc_type is "crf" | No |
| int32_t | qp | qp , only valid when rc_type is "q" | No |
| int32_t | qpmin | minimal qp, valid when rc_type is not "q", 1~qpmax | No |
| int32_t | qpmax | maximal qp, valid when rc_type is not "q", 1~255 | No |
| int32_t | max_keyframe_dist | max dist between key frames(<=0: no periodically I frame insertion) | No |
| int32_t | min_keyframe_dist | min dist between key frames | No |
| int32_t | s_frame_dist | min dist between S frames | No |
| int32_t | gop_size | (number of b frames) + 1, [1, 16] | No |
| int32_t | b_pyramid | pyramid b-frames, default 5 | No |
| int32_t | lookahead | lookahead depth [0, 65] | No |
| int32_t | log_level | log level, [-1,2] debug info warn error, default is info | No |
| int32_t | enable_scalability | enable svc (will overwrite gop_size and lookahead) | Yes |
| wz_scalability_mode_t | scalability_mode | svc mode, [0, 1] | Yes |
| int32_t | enable_frame_skip | max continuous skipped frame for rate control,default 0 | Yes |
| int32_t | tile_cols | tile cols, [1, 64], default 1 | Reopen encoder |
| int32_t | tile_rows | tile rows, [1, 64], default 1 | Reopen encoder |
| int32_t | enable_frame_id | enable frame_id_numbers_present_flag | No |
| int32_t | scenecut | scenecut threshold | Yes |
| int32_t | fpp | frame parallel threads, default 0(disable) | Reopen encoder |
extra_cfg
The API allows a number of additional parameters to be passed through to the encoder in string format. This can be done at setup and during encoding as an update.
The following table lists the encoding tools that can be configured through "extra_cfg", and you can use "key=value" to set these tools, with colon delimiters (":"). For example, psnr=1:summary=1
| Params | DataType,{Value} | Description | Safe for update config |
|---|---|---|---|
| psnr | Int, {0, 1} | Show PSNR | Yes |
| ssim | Int, {0, 1} | Show SSIM | Yes |
| summary | Int, {0, 1} | Show some information after encoding | Yes |
| refs | Int, {0 - 7} | Maximum reference frames to be used , 0:auto | Yes |
| rr | Int, {0,1,2,3} | Rate control level, 0: frame level, 1: constrained frame level, 2: block level, 3: hard frame block level; default is 3. | No |
| strict_vbv | Int, {0,1} | Strict VBV based rate control default is 0. | No |
| error_resilient_mode | Int, {0,1} | Turn error resilient mode on/off. Default is 0. | No |
| imaxq | Int, {1 - 255} | Maximal qp value for key frames, default is -1(means equal to qpmax) | No |
| iminq | Int, {1 - 255} | minimal qp value for key frames, default is -1(means equal to qpmin) | No |
| layer_maxq | Int, {1 - 255} | Maximal qp value for each scalable layer, comma-separated | No |
| layer_maxq | Int, {1 - 255} | Maximal qp value for each scalable layer, comma-separated | No |
| tsbr | string, {value1, value2,...} | Target bitrate(kbps) for each temporal layer, comma-separated | Yes |
| lry | int, {0 - 2} | Loop restoration level for luma | No |
| lruv | int, {0 - 2} | Loop restoration level for chroma | No |
| sfrs | Int | Static frame refresh level | Yes |
| init_frame_id | Int, {0 - 2^15} | Init value of frame_id | No |
| sc_level | Int, {0,2} | Screen content detect level | Yes |
| force_ref | Int | Set a specific frame_id to be used for reference when only 1 reference frame is used | Yes |
| td_obu | Int,{0, 1} | Enable temporal_delimiter OBU, default is 1 | No |
| sbrow_tx_thr | Int | Reduce complexity for huge frames, default is INT_MAX | Yes |
| obu_extension | Int,{0, 1} | Enable obu header extension, default is 1 | No |
| svc_meta_obu | Int,{0, 1} | Enable svc metadata OBU, default is 0 | No |
| max_qindex_reduce | Int,{0 - 255} | Max qindex reduction from last layer 0 frame, default is 40 | Yes |
| dup_frame_detect | Int,{0, 1, 2} | Detect duplicate frames and re-show instead of coding 0:off, 1: strong test 2: weaker test default is 2 | Yes |
| aq_mode | Int, {0,3} | Adaptive Quantization model, default is speed/preset-dependent | Yes |
| aq_strength | Int | Adaptive Quantization strength, default is speed/preset-dependent | Yes |
| adaptive_resolution | Int {0-15} | Adaptive resolution control flags, default 0 | No |
| qindex_switch_up | Int, {0 - 255} | QIndex at which resolution is switched up | Yes |
| qindex_switch_down | Int, {0 - 255} | QIndex at which resolution is switched down: | Yes |
| gfi | Golden frame interval | Frame distance between golden frames, default is speed/preset dependent | Yes |
| gfb / gfbtf | Golden frame boost | Boost for golden frames or temporally filtered golden frames | Yes |
| cbr_decay | Int, {1 - 99} | Speed of rate adaptation | Yes |
Flags
Flags can also be passed in each encoder call to set frame-specific features (e.g. force a keyframe). The available flags are:
WZ_FORCE_KEYFRAME_FLAG ..... Force this frame to be a key frame
WZ_FORCE_SFRAME_FLAG ..... Force this frame to be a switch frame
WZ_FORCE_INTRAONLY_FLAG ..... Force this frame to be an intra-only frame
WZ_SET_SVC_TID_FLAG ..... Set this frame to have a specific temporal ID (must be in range for the scalability preset - see below)
WZ_SET_SVC_SID_FLAG ..... Set this frame to have a specific spatial ID (must be in range for the scalability preset - see below)
The layer ID values are set within the respective bit positions e.g. (tid << WZ_SVC_TID_SHIFT). Note there is potential conflicts between the set layer IDs and other flags and settings (such as frame skip) which may cause the encoded layer ID to be different. The output layer ID should always be checked.
Region of Interest (ROI) encoding
ROI rectangles may be passed with the frame structure with each call to the encoder. This is an experimental feature.
Film Grain Support
Aurora1-RTC does not support Film Grain Synthesis.
Screen Content Coding
Aurora1-RTC has extensive support for encoding optimised for screen content. It can be accessed via the screen preset.