nodes/avenstack-ComfyUI-AV-FunASR
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

1. 提升运行速度，2. 优化工作流，3. 字幕默认保存为utf-8编码

Browse Source
This commit is contained in:
aven
2025-06-12 17:00:51 +08:00
parent 6840820f8d
commit 9e252d92ef
5 changed files with 900 additions and 569 deletions
Download Patch File Download Diff File Expand all files Collapse all files
example_workflow/av-audio2subtitle-with-prompt-1.json
+423
View File
@@ -0,0 +1,423 @@
{
"id": "377e0192-2ad2-408a-871c-6a86cf058f73",
"revision": 0,
"last_node_id": 45,
"last_link_id": 77,
"nodes": [
{
"id": 43,
"type": "easy showAnything",
"pos": [
737.1539306640625,
51.4323844909668
],
"size": [
478.719970703125,
307.6799621582031
],
"flags": {},
"order": 5,
"mode": 0,
"inputs": [
{
"name": "anything",
"shape": 7,
"type": "*",
"link": 76
}
],
"outputs": [
{
"name": "output",
"type": "*",
"links": null
}
],
"properties": {
"cnr_id": "comfyui-easy-use",
"ver": "1.2.9",
"Node name for S&R": "easy showAnything"
},
"widgets_values": [
"{\"key\": \"737fb4ed-6c4a-4e66-b8ce-c2c0c416f45a\", \"text\": \"三 分 钟 吉 祥 q 弹 面 条 裹 满 浓 郁 汤 汁 每 一 口 都 香 到 上 头 无 论 是 深 夜 加 班 追 剧 嘴 馋 还 是 懒 人 素 食 x 泡 面 都 能 瞬 间 满 足 你 的 胃 经 典 红 烧 酸 辣 劲 爽 香 浓 豚 骨 多 种 口 味 总 有 一 款 让 你 欲 罢 不 能\", \"timestamp\": [[150, 350], [350, 530], [530, 710], [710, 870], [870, 1110], [1410, 1650], [1650, 1850], [1850, 2010], [2010, 2250], [2550, 2730], [2730, 2930], [2930, 3090], [3090, 3210], [3210, 3390], [3390, 3630], [3930, 4110], [4110, 4290], [4290, 4530], [4530, 4670], [4670, 4830], [4830, 5010], [5010, 5170], [5170, 5410], [5890, 6070], [6070, 6250], [6250, 6450], [6450, 6610], [6610, 6810], [6810, 6910], [6910, 7150], [7410, 7630], [7630, 7770], [7770, 7930], [7930, 8170], [8470, 8650], [8650, 8850], [8850, 8950], [8950, 9130], [9130, 9310], [9310, 9550], [9870, 10110], [10210, 10370], [10370, 10610], [10630, 10730], [10730, 10930], [10930, 11090], [11090, 11290], [11290, 11450], [11450, 11570], [11570, 11750], [11750, 11870], [11870, 12110], [12610, 12830], [12830, 13030], [13030, 13190], [13190, 13430], [13770, 14010], [14030, 14230], [14230, 14370], [14370, 14610], [14910, 15090], [15090, 15330], [15350, 15570], [15570, 15810], [16050, 16230], [16230, 16450], [16450, 16650], [16650, 16890], [17130, 17310], [17310, 17490], [17490, 17570], [17570, 17810], [17830, 17970], [17970, 18110], [18110, 18270], [18270, 18410], [18410, 18570], [18570, 19045]]}"
]
},
{
"id": 42,
"type": "easy showAnything",
"pos": [
1252.9384765625,
-201.37600708007812
],
"size": [
544.5440673828125,
575.8159790039062
],
"flags": {},
"order": 7,
"mode": 0,
"inputs": [
{
"name": "anything",
"shape": 7,
"type": "*",
"link": 73
}
],
"outputs": [
{
"name": "output",
"type": "*",
"links": [
77
]
}
],
"properties": {
"cnr_id": "comfyui-easy-use",
"ver": "1.2.9",
"Node name for S&R": "easy showAnything"
},
"widgets_values": [
"1\n00:00:00,150 --> 00:00:00,710\n3分钟即享\n\n2\n00:00:00,710 --> 00:00:03,210\nQ弹面条 裹满浓郁\n\n3\n00:00:03,210 --> 00:00:05,410\n汤汁 每一口都香到上头\n\n4\n00:00:05,890 --> 00:00:08,170\n无论是深夜加班、追剧嘴馋\n\n5\n00:00:08,470 --> 00:00:10,730\n还是懒人速食 X泡面都\n\n6\n00:00:10,730 --> 00:00:11,870\n能瞬间满足你的胃\n\n7\n00:00:11,870 --> 00:00:15,330\n经典红烧、酸辣劲爽、香浓\n\n8\n00:00:15,350 --> 00:00:17,810\n豚骨 多种口味 总有一款\n\n9\n00:00:17,830 --> 00:00:19,045\n让你欲罢不能\n\n"
]
},
{
"id": 38,
"type": "AVASRTimestamp",
"pos": [
343.9599304199219,
-215.4831085205078
],
"size": [
270,
122
],
"flags": {},
"order": 3,
"mode": 0,
"inputs": [
{
"name": "audio",
"type": "AUDIO",
"link": 74
}
],
"outputs": [
{
"name": "text",
"type": "STRING",
"links": [
75
]
},
{
"name": "json_result",
"type": "STRING",
"links": [
76
]
},
{
"name": "asr_result",
"type": "ASRRESULT",
"links": [
71
]
}
],
"properties": {
"aux_id": "ComfyUI-AV-FunASR.git",
"ver": "6840820f8d34ffb273e8776b8ba73a8fb0e89d17",
"Node name for S&R": "AVASRTimestamp"
},
"widgets_values": [
300,
false
]
},
{
"id": 41,
"type": "AVFormat2Subtitle",
"pos": [
773.8641967773438,
-325.087890625
],
"size": [
270,
58
],
"flags": {},
"order": 6,
"mode": 0,
"inputs": [
{
"name": "asr_result",
"type": "ASRRESULT",
"link": 71
},
{
"name": "text",
"type": "STRING",
"widget": {
"name": "text"
},
"link": 72
}
],
"outputs": [
{
"name": "subtitle",
"type": "STRING",
"links": [
73
]
}
],
"properties": {
"aux_id": "ComfyUI-AV-FunASR.git",
"ver": "6840820f8d34ffb273e8776b8ba73a8fb0e89d17",
"Node name for S&R": "AVFormat2Subtitle"
},
"widgets_values": [
""
]
},
{
"id": 2,
"type": "LoadAudio",
"pos": [
51.75614929199219,
-214.3138885498047
],
"size": [
270,
136
],
"flags": {},
"order": 0,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "AUDIO",
"type": "AUDIO",
"links": [
74
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.32",
"Node name for S&R": "LoadAudio"
},
"widgets_values": [
"22_jidaohuman-megatts3_0.flac",
null,
null
]
},
{
"id": 39,
"type": "easy showAnything",
"pos": [
734.82470703125,
-204.00804138183594
],
"size": [
489.3119812011719,
201.248046875
],
"flags": {},
"order": 4,
"mode": 0,
"inputs": [
{
"name": "anything",
"shape": 7,
"type": "*",
"link": 75
}
],
"outputs": [
{
"name": "output",
"type": "*",
"links": []
}
],
"properties": {
"cnr_id": "comfyui-easy-use",
"ver": "1.2.9",
"Node name for S&R": "easy showAnything"
},
"widgets_values": [
"三 分 钟 吉 祥 q 弹 面 条 裹 满 浓 郁 汤 汁 每 一 口 都 香 到 上 头 无 论 是 深 夜 加 班 追 剧 嘴 馋 还 是 懒 人 素 食 x 泡 面 都 能 瞬 间 满 足 你 的 胃 经 典 红 烧 酸 辣 劲 爽 香 浓 豚 骨 多 种 口 味 总 有 一 款 让 你 欲 罢 不 能"
]
},
{
"id": 44,
"type": "AVSaveSubtitles",
"pos": [
1844.3179931640625,
-202.0580291748047
],
"size": [
270,
82
],
"flags": {},
"order": 8,
"mode": 0,
"inputs": [
{
"name": "subtitles",
"type": "STRING",
"widget": {
"name": "subtitles"
},
"link": 77
}
],
"outputs": [],
"title": "保存字幕文件到output",
"properties": {
"aux_id": "ComfyUI-AV-FunASR.git",
"ver": "6840820f8d34ffb273e8776b8ba73a8fb0e89d17",
"Node name for S&R": "AVSaveSubtitles"
},
"widgets_values": [
"",
"subtitles"
]
},
{
"id": 25,
"type": "PrimitiveStringMultiline",
"pos": [
346.7919006347656,
-10.424053192138672
],
"size": [
263.8999938964844,
168.8079833984375
],
"flags": {
"collapsed": false
},
"order": 1,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "STRING",
"type": "STRING",
"links": [
72
]
}
],
"title": "校准文字（如果没有请删除此结点）",
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.32",
"Node name for S&R": "PrimitiveStringMultiline"
},
"widgets_values": [
"3分钟即享，Q弹面条，裹满浓郁汤汁，每一口都香到上头！无论是深夜加班、追剧嘴馋，还是懒人速食，X泡面都能瞬间满足你的胃！经典红烧、酸辣劲爽、香浓豚骨，多种口味，总有一款让你欲罢不能。"
]
},
{
"id": 45,
"type": "Note",
"pos": [
118.52088165283203,
204.437744140625
],
"size": [
574.4384155273438,
157.96160888671875
],
"flags": {},
"order": 2,
"mode": 0,
"inputs": [],
"outputs": [],
"properties": {},
"widgets_values": [
"原理介绍：https://mp.weixin.qq.com/s/ilGpgS5i_ZbE3SdA6godyw\n\nBy Aven，公众号：智效笔谈"
],
"color": "#432",
"bgcolor": "#653"
}
],
"links": [
[
71,
38,
2,
41,
0,
"ASRRESULT"
],
[
72,
25,
0,
41,
1,
"STRING"
],
[
73,
41,
0,
42,
0,
"*"
],
[
74,
2,
0,
38,
0,
"AUDIO"
],
[
75,
38,
0,
39,
0,
"*"
],
[
76,
38,
1,
43,
0,
"*"
],
[
77,
42,
0,
44,
0,
"STRING"
]
],
"groups": [],
"config": {},
"extra": {
"frontendVersion": "1.18.9",
"VHS_latentpreview": false,
"VHS_latentpreviewrate": 0,
"VHS_MetadataImage": true,
"VHS_KeepIntermediate": true
},
"version": 0.4
}
example_workflow/av-audio2subtitle-with-prompt-2.json
+440
View File
@@ -0,0 +1,440 @@
{
"id": "377e0192-2ad2-408a-871c-6a86cf058f73",
"revision": 0,
"last_node_id": 46,
"last_link_id": 82,
"nodes": [
{
"id": 43,
"type": "easy showAnything",
"pos": [
737.1539306640625,
51.4323844909668
],
"size": [
478.719970703125,
307.6799621582031
],
"flags": {},
"order": 5,
"mode": 0,
"inputs": [
{
"name": "anything",
"shape": 7,
"type": "*",
"link": 80
}
],
"outputs": [
{
"name": "output",
"type": "*",
"links": null
}
],
"properties": {
"cnr_id": "comfyui-easy-use",
"ver": "1.2.9",
"Node name for S&R": "easy showAnything"
},
"widgets_values": [
"{\"key\": \"rand_key_gUe52RvEJgwBu\", \"text\": \"3 分 钟 即 享 q 弹 面 条 裹 满 浓 郁 汤 汁 每 一 口 都 香 到 上 头 无 论 是 深 夜 加 班 追 剧 嘴 馋 还 是 懒 人 速 食 x 泡 面 都 能 瞬 间 满 足 你 的 胃 经 典 红 烧 酸 辣 劲 爽 香 浓 豚 骨 多 种 口 味 总 有 一 款 让 你 欲 罢 不 能\", \"timestamp\": [[170, 330], [330, 510], [510, 670], [670, 910], [910, 1150], [1410, 1470], [1470, 1670], [1670, 1830], [1830, 2050], [2050, 2290], [2590, 2730], [2730, 2850], [2850, 2930], [2930, 3090], [3090, 3250], [3250, 3470], [3470, 3710], [3930, 4070], [4070, 4130], [4130, 4310], [4310, 4530], [4530, 4690], [4690, 4810], [4810, 4970], [4970, 5190], [5190, 5250], [5250, 5490], [5970, 6130], [6130, 6290], [6290, 6430], [6430, 6590], [6590, 6650], [6650, 6830], [6830, 7010], [7010, 7250], [7450, 7630], [7630, 7790], [7790, 8010], [8010, 8070], [8070, 8310], [8550, 8670], [8670, 8830], [8830, 8950], [8950, 9150], [9150, 9350], [9350, 9430], [9430, 9670], [9910, 10150], [10230, 10410], [10410, 10610], [10610, 10730], [10730, 10790], [10790, 10970], [10970, 11130], [11130, 11290], [11290, 11470], [11470, 11590], [11590, 11690], [11690, 11770], [11770, 11950], [11950, 12190], [12690, 12890], [12890, 13070], [13070, 13310], [13310, 13370], [13370, 13610], [13910, 14070], [14070, 14230], [14230, 14470], [14470, 14710], [14990, 15170], [15170, 15230], [15230, 15430], [15430, 15670], [15670, 15910], [16129, 16309], [16309, 16490], [16490, 16550]]}"
]
},
{
"id": 42,
"type": "easy showAnything",
"pos": [
1252.9384765625,
-201.37600708007812
],
"size": [
544.5440673828125,
575.8159790039062
],
"flags": {},
"order": 7,
"mode": 0,
"inputs": [
{
"name": "anything",
"shape": 7,
"type": "*",
"link": 73
}
],
"outputs": [
{
"name": "output",
"type": "*",
"links": [
77
]
}
],
"properties": {
"cnr_id": "comfyui-easy-use",
"ver": "1.2.9",
"Node name for S&R": "easy showAnything"
},
"widgets_values": [
"1\n00:00:00,170 --> 00:00:00,910\n3分钟即享\n\n2\n00:00:00,910 --> 00:00:02,930\nQ弹面条 裹满浓郁\n\n3\n00:00:02,930 --> 00:00:04,810\n汤汁 每一口都香到上头\n\n4\n00:00:04,810 --> 00:00:07,010\n无论是深夜加班、追剧嘴馋\n\n5\n00:00:07,010 --> 00:00:09,150\n还是懒人速食 X泡面都\n\n6\n00:00:09,150 --> 00:00:10,730\n能瞬间满足你的胃\n\n7\n00:00:10,730 --> 00:00:12,890\n经典红烧、酸辣劲爽、香浓\n\n8\n00:00:12,890 --> 00:00:15,230\n豚骨 多种口味 总有一款\n\n9\n00:00:15,230 --> 00:00:16,550\n让你欲罢不能\n\n"
]
},
{
"id": 39,
"type": "easy showAnything",
"pos": [
734.82470703125,
-204.00804138183594
],
"size": [
489.3119812011719,
201.248046875
],
"flags": {},
"order": 4,
"mode": 0,
"inputs": [
{
"name": "anything",
"shape": 7,
"type": "*",
"link": 79
}
],
"outputs": [
{
"name": "output",
"type": "*",
"links": []
}
],
"properties": {
"cnr_id": "comfyui-easy-use",
"ver": "1.2.9",
"Node name for S&R": "easy showAnything"
},
"widgets_values": [
"3 分 钟 即 享 q 弹 面 条 裹 满 浓 郁 汤 汁 每 一 口 都 香 到 上 头 无 论 是 深 夜 加 班 追 剧 嘴 馋 还 是 懒 人 速 食 x 泡 面 都 能 瞬 间 满 足 你 的 胃 经 典 红 烧 酸 辣 劲 爽 香 浓 豚 骨 多 种 口 味 总 有 一 款 让 你 欲 罢 不 能"
]
},
{
"id": 44,
"type": "AVSaveSubtitles",
"pos": [
1844.3179931640625,
-202.0580291748047
],
"size": [
270,
82
],
"flags": {},
"order": 8,
"mode": 0,
"inputs": [
{
"name": "subtitles",
"type": "STRING",
"widget": {
"name": "subtitles"
},
"link": 77
}
],
"outputs": [],
"title": "保存字幕文件到output",
"properties": {
"aux_id": "ComfyUI-AV-FunASR.git",
"ver": "6840820f8d34ffb273e8776b8ba73a8fb0e89d17",
"Node name for S&R": "AVSaveSubtitles"
},
"widgets_values": [
"",
"subtitles"
]
},
{
"id": 2,
"type": "LoadAudio",
"pos": [
51.75614929199219,
-214.3138885498047
],
"size": [
270,
136
],
"flags": {},
"order": 0,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "AUDIO",
"type": "AUDIO",
"links": [
78
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.32",
"Node name for S&R": "LoadAudio"
},
"widgets_values": [
"22_jidaohuman-megatts3_0.flac",
null,
null
]
},
{
"id": 41,
"type": "AVFormat2Subtitle",
"pos": [
741.9202880859375,
-337.8655090332031
],
"size": [
270,
58
],
"flags": {},
"order": 6,
"mode": 0,
"inputs": [
{
"name": "asr_result",
"type": "ASRRESULT",
"link": 81
},
{
"name": "text",
"type": "STRING",
"widget": {
"name": "text"
},
"link": 72
}
],
"outputs": [
{
"name": "subtitle",
"type": "STRING",
"links": [
73
]
}
],
"properties": {
"aux_id": "ComfyUI-AV-FunASR.git",
"ver": "6840820f8d34ffb273e8776b8ba73a8fb0e89d17",
"Node name for S&R": "AVFormat2Subtitle"
},
"widgets_values": [
""
]
},
{
"id": 45,
"type": "AVSpeechTimestamp",
"pos": [
350.81390380859375,
-202.5904083251953
],
"size": [
270,
122
],
"flags": {},
"order": 3,
"mode": 0,
"inputs": [
{
"name": "audio",
"type": "AUDIO",
"link": 78
},
{
"name": "text",
"type": "STRING",
"widget": {
"name": "text"
},
"link": 82
}
],
"outputs": [
{
"name": "text",
"type": "STRING",
"links": [
79
]
},
{
"name": "json_result",
"type": "STRING",
"links": [
80
]
},
{
"name": "asr_result",
"type": "ASRRESULT",
"links": [
81
]
}
],
"properties": {
"aux_id": "ComfyUI-AV-FunASR.git",
"ver": "6840820f8d34ffb273e8776b8ba73a8fb0e89d17",
"Node name for S&R": "AVSpeechTimestamp"
},
"widgets_values": [
"",
false
]
},
{
"id": 25,
"type": "PrimitiveStringMultiline",
"pos": [
58.23122024536133,
-22.136831283569336
],
"size": [
259.0320129394531,
168.8079833984375
],
"flags": {
"collapsed": false
},
"order": 1,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "STRING",
"type": "STRING",
"links": [
72,
82
]
}
],
"title": "校准文字",
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.32",
"Node name for S&R": "PrimitiveStringMultiline"
},
"widgets_values": [
"3分钟即享，Q弹面条，裹满浓郁汤汁，每一口都香到上头！无论是深夜加班、追剧嘴馋，还是懒人速食，X泡面都能瞬间满足你的胃！经典红烧、酸辣劲爽、香浓豚骨，多种口味，总有一款让你欲罢不能。"
]
},
{
"id": 46,
"type": "Note",
"pos": [
81.41957092285156,
204.1630096435547
],
"size": [
574.4384155273438,
157.96160888671875
],
"flags": {},
"order": 2,
"mode": 0,
"inputs": [],
"outputs": [],
"properties": {},
"widgets_values": [
"原理介绍：https://mp.weixin.qq.com/s/ilGpgS5i_ZbE3SdA6godyw\n\nBy Aven，公众号：智效笔谈"
],
"color": "#432",
"bgcolor": "#653"
}
],
"links": [
[
72,
25,
0,
41,
1,
"STRING"
],
[
73,
41,
0,
42,
0,
"*"
],
[
77,
42,
0,
44,
0,
"STRING"
],
[
78,
2,
0,
45,
0,
"AUDIO"
],
[
79,
45,
0,
39,
0,
"*"
],
[
80,
45,
1,
43,
0,
"*"
],
[
81,
45,
2,
41,
0,
"ASRRESULT"
],
[
82,
25,
0,
45,
1,
"STRING"
]
],
"groups": [],
"config": {},
"extra": {
"frontendVersion": "1.18.9",
"VHS_latentpreview": false,
"VHS_latentpreviewrate": 0,
"VHS_MetadataImage": true,
"VHS_KeepIntermediate": true
},
"version": 0.4
}
example_workflow/av-audio2subtitle-with-prompt.json
-311
View File
@@ -1,311 +0,0 @@
{
"id": "377e0192-2ad2-408a-871c-6a86cf058f73",
"revision": 0,
"last_node_id": 33,
"last_link_id": 50,
"nodes": [
{
"id": 2,
"type": "LoadAudio",
"pos": [
182.18014526367188,
389.2699890136719
],
"size": [
270,
136
],
"flags": {},
"order": 0,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "AUDIO",
"type": "AUDIO",
"links": [
24
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.32",
"Node name for S&R": "LoadAudio"
},
"widgets_values": [
"4_jidaohuman-megatts3_0.flac",
null,
null
]
},
{
"id": 25,
"type": "PrimitiveStringMultiline",
"pos": [
186.9199676513672,
591.9599609375
],
"size": [
260,
163
],
"flags": {
"collapsed": false
},
"order": 1,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "STRING",
"type": "STRING",
"links": [
47,
48
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.32",
"Node name for S&R": "PrimitiveStringMultiline"
},
"widgets_values": [
"直观来说，测试时试错的步骤只有当能带来信息增益的情况下才有益，然而人们并没有在RL训练中告诉模型试错和反思带来的信息增益。"
]
},
{
"id": 18,
"type": "AVSpeechTimestamp",
"pos": [
489.56988525390625,
374.6600341796875
],
"size": [
210,
128
],
"flags": {
"collapsed": false
},
"order": 2,
"mode": 0,
"inputs": [
{
"name": "audio",
"type": "AUDIO",
"link": 24
},
{
"name": "text",
"type": "STRING",
"widget": {
"name": "text"
},
"link": 47
}
],
"outputs": [
{
"name": "text",
"type": "STRING",
"links": []
},
{
"name": "json_result",
"type": "STRING",
"links": []
},
{
"name": "asr_result",
"type": "ASRRESULT",
"links": [
43
]
}
],
"properties": {
"Node name for S&R": "AVSpeechTimestamp"
},
"widgets_values": [
"",
false
]
},
{
"id": 30,
"type": "AVFormat2Subtitle",
"pos": [
518.5700073242188,
561.3599853515625
],
"size": [
210,
88
],
"flags": {
"collapsed": false
},
"order": 3,
"mode": 0,
"inputs": [
{
"name": "asr_result",
"type": "ASRRESULT",
"link": 43
},
{
"name": "text",
"type": "STRING",
"widget": {
"name": "text"
},
"link": 48
}
],
"outputs": [
{
"name": "subtitle",
"type": "STRING",
"links": [
45,
50
]
}
],
"properties": {
"Node name for S&R": "AVFormat2Subtitle"
},
"widgets_values": [
""
]
},
{
"id": 31,
"type": "AVSaveSubtitles",
"pos": [
778.79541015625,
357.02178955078125
],
"size": [
270,
82
],
"flags": {
"collapsed": true
},
"order": 4,
"mode": 0,
"inputs": [
{
"name": "subtitles",
"type": "STRING",
"widget": {
"name": "subtitles"
},
"link": 45
}
],
"outputs": [],
"properties": {
"Node name for S&R": "AVSaveSubtitles"
},
"widgets_values": [
"",
"subtitles"
]
},
{
"id": 33,
"type": "Bjornulf_ShowText",
"pos": [
814.5744018554688,
427.3533935546875
],
"size": [
650.6199951171875,
437.7900085449219
],
"flags": {},
"order": 5,
"mode": 0,
"inputs": [
{
"name": "text_value",
"type": "STRING",
"link": 50
}
],
"outputs": [],
"properties": {
"cnr_id": "bjornulf_custom_nodes",
"ver": "1.0.0",
"Node name for S&R": "Bjornulf_ShowText"
},
"widgets_values": [
"1\n00:00:00,170 --> 00:00:02,330\n直观来说 测试时试错的步骤\n\n2\n00:00:02,330 --> 00:00:04,590\n只有当能带来信息增益的情况下才有益\n\n3\n00:00:04,590 --> 00:00:06,870\n然而人们并没有在RL训练\n\n4\n00:00:06,870 --> 00:00:09,330\n中告诉模型试错和反思带来的信息增益\n\n5\n00:00:09,330 --> 00:00:09,470\n\n\n"
]
}
],
"links": [
[
24,
2,
0,
18,
0,
"AUDIO"
],
[
43,
18,
2,
30,
0,
"ASRRESULT"
],
[
45,
30,
0,
31,
0,
"STRING"
],
[
47,
25,
0,
18,
1,
"STRING"
],
[
48,
25,
0,
30,
1,
"STRING"
],
[
50,
30,
0,
33,
0,
"STRING"
]
],
"groups": [],
"config": {},
"extra": {
"frontendVersion": "1.18.9",
"VHS_latentpreview": false,
"VHS_latentpreviewrate": 0,
"VHS_MetadataImage": true,
"VHS_KeepIntermediate": true
},
"version": 0.4
}
example_workflow/av-audio2subtitle-without-prompt.json
-241
View File
@@ -1,241 +0,0 @@
{
"id": "377e0192-2ad2-408a-871c-6a86cf058f73",
"revision": 0,
"last_node_id": 35,
"last_link_id": 53,
"nodes": [
{
"id": 2,
"type": "LoadAudio",
"pos": [
187.18014526367188,
368.2699890136719
],
"size": [
270,
136
],
"flags": {},
"order": 0,
"mode": 0,
"inputs": [],
"outputs": [
{
"name": "AUDIO",
"type": "AUDIO",
"links": [
51
]
}
],
"properties": {
"cnr_id": "comfy-core",
"ver": "0.3.32",
"Node name for S&R": "LoadAudio"
},
"widgets_values": [
"4_jidaohuman-megatts3_0.flac",
null,
null
]
},
{
"id": 31,
"type": "AVSaveSubtitles",
"pos": [
819.79541015625,
213.02178955078125
],
"size": [
270,
82
],
"flags": {
"collapsed": true
},
"order": 3,
"mode": 0,
"inputs": [
{
"name": "subtitles",
"type": "STRING",
"widget": {
"name": "subtitles"
},
"link": 45
}
],
"outputs": [],
"properties": {
"Node name for S&R": "AVSaveSubtitles"
},
"widgets_values": [
"",
"subtitles"
]
},
{
"id": 34,
"type": "AVASRTimestamp",
"pos": [
467.7745361328125,
206.14337158203125
],
"size": [
270,
122
],
"flags": {},
"order": 1,
"mode": 0,
"inputs": [
{
"name": "audio",
"type": "AUDIO",
"link": 51
}
],
"outputs": [
{
"name": "text",
"type": "STRING",
"links": null
},
{
"name": "json_result",
"type": "STRING",
"links": null
},
{
"name": "asr_result",
"type": "ASRRESULT",
"links": [
52
]
}
],
"properties": {
"Node name for S&R": "AVASRTimestamp"
},
"widgets_values": [
300,
false
]
},
{
"id": 30,
"type": "AVFormat2Subtitle",
"pos": [
485.57000732421875,
415.3599853515625
],
"size": [
210,
88
],
"flags": {
"collapsed": false
},
"order": 2,
"mode": 0,
"inputs": [
{
"name": "asr_result",
"type": "ASRRESULT",
"link": 52
}
],
"outputs": [
{
"name": "subtitle",
"type": "STRING",
"links": [
45,
53
]
}
],
"properties": {
"Node name for S&R": "AVFormat2Subtitle"
},
"widgets_values": [
""
]
},
{
"id": 35,
"type": "Bjornulf_ShowText",
"pos": [
804.8345336914062,
269.0433654785156
],
"size": [
398,
387
],
"flags": {},
"order": 4,
"mode": 0,
"inputs": [
{
"name": "text_value",
"type": "STRING",
"link": 53
}
],
"outputs": [],
"properties": {
"cnr_id": "bjornulf_custom_nodes",
"ver": "1.0.0",
"Node name for S&R": "Bjornulf_ShowText"
},
"widgets_values": [
"1\n00:00:00,150 --> 00:00:02,610\n直观来说测试时试错的步骤只有\n\n2\n00:00:02,610 --> 00:00:05,110\n当能带来信息增益的情况下才有益\n\n3\n00:00:05,410 --> 00:00:07,130\n然而人们并没有在all训练中\n\n4\n00:00:07,130 --> 00:00:10,225\n告诉模型试错和反思带来的信息增益\n\n"
]
}
],
"links": [
[
45,
30,
0,
31,
0,
"STRING"
],
[
51,
2,
0,
34,
0,
"AUDIO"
],
[
52,
34,
2,
30,
0,
"ASRRESULT"
],
[
53,
30,
0,
35,
0,
"STRING"
]
],
"groups": [],
"config": {},
"extra": {
"frontendVersion": "1.18.9",
"VHS_latentpreview": false,
"VHS_latentpreviewrate": 0,
"VHS_MetadataImage": true,
"VHS_KeepIntermediate": true
},
"version": 0.4
}
nodes.py
+37 -17
View File
@@ -6,6 +6,7 @@ import torchaudio
import torchvision.utils as vutils
import torch
import json
import uuid
from modelscope.pipelines import pipeline
from modelscope.utils.constant import Tasks
@@ -59,19 +60,31 @@ class AVSpeechTimestamp:
if AVSpeechTimestamp.infer_ins_cache is None:
model_root = os.path.join(folder_paths.models_dir, "ASR/FunASR")
model_dir = os.path.join(model_root, name_maps_ms["fa-zh"])
# vad_model = os.path.join(model_root, name_maps_ms["fsmn-vad"])
os.makedirs(model_dir, exist_ok=True)
AVSpeechTimestamp.infer_ins_cache = pipeline(
task=Tasks.speech_timestamp,
device = "cuda" if torch.cuda.is_available() else "cpu"
AVSpeechTimestamp.infer_ins_cache = AutoModel(
model=model_dir,
model_revision="v2.0.4",
output_dir=temp_dir
device=device, # GPU加速
disable_update=True
)
# save audio
audio_save_path = os.path.join(temp_dir, f"{int(time.time())}.wav")
torchaudio.save(audio_save_path, audio['waveform'].squeeze(0), audio["sample_rate"])
rec_result = AVSpeechTimestamp.infer_ins_cache(input=(audio_save_path,text), data_type=("sound", "text"))
# save audio
uuidv4 = str(uuid.uuid4())
audio_save_path = os.path.join(temp_dir, f"{uuidv4}.wav")
waveform = audio['waveform']
sr = audio["sample_rate"]
waveform = torchaudio.functional.resample(waveform, sr, 16000)
torchaudio.save(audio_save_path, waveform.squeeze(0), 16000)
rec_result = AVSpeechTimestamp.infer_ins_cache.generate(
input=(audio_save_path, text),
data_type=("sound", "text"),
)
# print(rec_result)
if rec_result:
rec_result = rec_result[0]
@@ -86,7 +99,7 @@ class AVSpeechTimestamp:
# jr = json.dumps(rec_result, indent=4)
text = rec_result.get("text")
jr = json.dumps(rec_result, ensure_ascii=False)
print(text, jr)
# print(text, jr)
return (text, jr, rec_result)
@@ -122,28 +135,35 @@ class AVASRTimestamp:
model=model_dir,
vad_model=vad_model,
punc_model=None, #"ct-punc",
device=device # GPU加速
device=device, # GPU加速
disable_update=True
)
# save audio
audio_save_path = os.path.join(temp_dir, f"{int(time.time())}.wav")
torchaudio.save(audio_save_path, audio['waveform'].squeeze(0), audio["sample_rate"])
# save
uuidv4 = str(uuid.uuid4())
audio_save_path = os.path.join(temp_dir, f"{uuidv4}.wav")
# 重新采样为16k
waveform = audio['waveform']
sr = audio["sample_rate"]
waveform = torchaudio.functional.resample(waveform, sr, 16000)
torchaudio.save(audio_save_path, waveform.squeeze(0), 16000)
rec_result = AVASRTimestamp.infer_ins_cache.generate(input=audio_save_path, batch_size_s=batch_size_s)
# print(rec_result)
if rec_result:
rec_result = rec_result[0]
# infer
if unload_model:
import gc
if AVSpeechTimestamp.infer_ins_cache is not None:
AVSpeechTimestamp.infer_ins_cache = None
if AVASRTimestamp.infer_ins_cache is not None:
AVASRTimestamp.infer_ins_cache = None
gc.collect()
torch.cuda.empty_cache()
print("AVSpeechTimestamp memory cleanup successful")
print("AVASRTimestamp memory cleanup successful")
# jr = json.dumps(rec_result, indent=4)
text = rec_result.get("text")
jr = json.dumps(rec_result, ensure_ascii=False)
print((text, jr, rec_result))
# print((text, jr, rec_result))
return (text, jr, rec_result)
@@ -197,7 +217,7 @@ class AVSaveSubtitles:
def save_subtitles(self, subtitles, filename_prefix="subtitles"):
filename = f"{filename_prefix}-{time.time()}.srt"
save_path = os.path.join(self.output_dir, filename)
with open(save_path, "w") as f:
with open(save_path, "w", encoding='utf-8') as f:
f.write(subtitles)
return { "ui": { "text": save_path } }