Speech2text

Automatic speech recognition using Whisper

⚠ Warning: This service is provided as is, please read the disclaimer here before use as well as for more information.

Audio

Model

Larger models increase transcription quality, but reduce performance

Language

If the language is known upfront, always set it manually

Model

Model used for translation

Language

Language used for output translation

Translate using OpenAI Whisper's built-in module

Translate to English

Translate using Facebook's NLLB

Translate to selected language

Show preview with timestamps

Add timestamp to filenames

Beam Size

Beam size to use for decoding.

Log Probability Threshold

If the average log probability over sampled tokens is below this value, treat as failed.

No Speech Threshold

If the no speech probability is higher than this value AND the average log probability over sampled tokens is below 'Log Prob Threshold', consider the segment as silent.

Compute Type

Select the type of computation to perform.

Best Of

Number of candidates when sampling with non-zero temperature.

Patience

Beam search patience factor.

Condition on previous text during decoding.

Condition On Previous Text

Prompt Reset On Temperature

Resets prompt if temperature is above this value. Arg has effect only if 'Condition On Previous Text' is True.

0 1

Initial Prompt

Initial prompt to use for decoding.

Temperature

Temperature for sampling. It can be a tuple of temperatures, which will be successively used upon failures according to either Compression Ratio Threshold or Log Prob Threshold.

0 1

Compression Ratio Threshold

If the gzip compression ratio is above this value, treat as failed.

Chunk Length (s)

The length of audio segments. If it is not None, it will overwrite the default chunk_length of the FeatureExtractor.

Length Penalty

Exponential length penalty constant.

Repetition Penalty

Penalty applied to the score of previously generated tokens (set > 1 to penalize).

No Repeat N-gram Size

Prevent repetitions of n-grams with this size (set 0 to disable).

Prefix

Optional text to provide as a prefix for the first window.

Suppress blank outputs at the beginning of the sampling.

Suppress Blank

Suppress Tokens

List of token IDs to suppress. -1 will suppress a default set of symbols as defined in the model config.json file.

Max Initial Timestamp

The initial timestamp cannot be later than this.

Extract word-level timestamps using the cross-attention pattern and dynamic time warping, and include the timestamps for each word in each segment.

Word Timestamps

Prepend Punctuations

If 'Word Timestamps' is True, merge these punctuation symbols with the next word.

Append Punctuations

If 'Word Timestamps' is True, merge these punctuation symbols with the previous word.

Max New Tokens

Maximum number of new tokens to generate per-chunk. If not set, the maximum will be set by the default max_length.

Hallucination Silence Threshold (sec)

When 'Word Timestamps' is True, skip silent periods longer than this threshold (in seconds) when a possible hallucination is detected.

Hotwords

Hotwords/hint phrases to provide the model with. Has no effect if prefix is not None.

Language Detection Threshold

If the maximum probability of the language tokens is higher than this value, the language is detected.

Language Detection Segments

Number of segments to consider for the language detection.

Output preview (Always review output generated by AI models)

Output info

Output data