ComfyUI Node
TS Whisper
Transcribe speech to text, with optional SRT subtitles, on the local OpenAI Whisper engine.
TS Whisper
- audio
- srt_content
- text_content
- ttml_content
◄modellarge-v3►
◄tasktranscribe►
◄source_languageru►
◄timestampssegment►
◄precisionfp16►
◄beam_size5►
◄temperature0.0►
◄temperature_fallbacks0.0,0.2,0.4,0.6,0.8,1.0►
◄condition_on_previous_texttrue►
◄compression_ratio_threshold2.4►
◄logprob_threshold-1.0►
◄no_speech_threshold0.60►
◄initial_prompt►
◄save_srt_filetrue►
◄output_filename_prefixtranscribed_audio►
◄output_dir/tmp/ComfyUI/output►
CategoryTS/Audio
Inputs (17)
| Name | Type | Default | Description |
|---|---|---|---|
| audio | AUDIO | ComfyUI audio input (waveform + sample_rate). | |
| model | COMBO | large-v3 | Whisper model: large-v3 (best quality) or turbo (faster). Weights and cache are shared with TS Super Prompt. |
| task | COMBO | transcribe | transcribe = recognize in the source language; translate_to_english = translate to English (turbo cannot translate and falls back to transcription). |
| source_language | COMBO | ru | Language of the input audio. auto = auto-detect. Use ru for Russian. |
| timestamps | COMBO | segment | segment = per-segment timecodes; word = per-word; none = no timecodes (SRT/TTML disabled). |
| precision | COMBO | fp16 | fp16 is faster on GPU; fp32 is more stable. CPU always uses fp32. |
| beam_size | INT | 51–10 | Beam search width (used when temperature=0). 1 = greedy. Higher = more accurate but slower. |
| temperature | FLOAT | 0.00–1 | Decoding temperature. 0 = deterministic. |
| temperature_fallbacks | STRING | 0.0,0.2,0.4,0.6,0.8,1.0 | Comma-separated temperature ladder (e.g. 0.0,0.2,0.4,0.6,0.8,1.0). Overrides temperature. |
| condition_on_previous_text | BOOLEAN | true | Feed context from previous segments (more coherent, but can amplify drift). |
| compression_ratio_threshold | FLOAT | 2.40–10 | zlib compression threshold for the hallucination detector (whisper default 2.4). |
| logprob_threshold | FLOAT | -1.0-10–0 | Average log-probability threshold (whisper default -1.0). |
| no_speech_threshold | FLOAT | 0.600–1 | No-speech probability threshold (whisper default 0.6). |
| initial_prompt | STRING | Initial prompt to bias vocabulary, names, and style. | |
| save_srt_file | BOOLEAN | true | Save SRT/TTML files to the output/subtitles folder. |
| output_filename_prefix | STRING | transcribed_audio | Filename prefix for SRT/TTML. |
| output_diropt | STRING | /tmp/ComfyUI/output | Output folder for SRT/TTML. Empty = default output directory. |
Outputs (3)
| Name | Type | Description |
|---|---|---|
| srt_content | STRING | Subtitles in SRT format (empty when timestamps=none). |
| text_content | STRING | Full transcription as plain text. |
| ttml_content | STRING | Subtitles in TTML format (empty when timestamps=none). |